Climate Scenarios for Olkiluoto on a Time-Scale of 120000 ... - Posiva

1 downloads 156 Views 3MB Size Report
Dec 4, 2011 - 4.3 Scenario with ice-free conditions at Olkiluoto during the MIS 3 . ...... Center, (http://www.ngdc.noaa
POSIVA 2011-04

Climate Scenarios for Olkiluoto on a Time-Scale of 120,000 Years Natalia Pimenoff Ari Venäläinen Heikki Järvinen

December 2011

POSIVA OY Olkiluoto FIN-27160 EURAJOKI, FINLAND Phone (02) 8372 31 (nat.), (+358-2-) 8372 31 (int.) Fax (02) 8372 3809 (nat.), (+358-2-) 8372 3809 (int.)

POSIVA 2011-04

Climate Scenarios for Olkiluoto on a Time-Scale of 120,000 Years Natalia Pimenoff Ari Venäläinen Heikki Järvinen Finnish Meteorological Institute

December 2011 POSIVA OY Olkiluoto FI-27160 EURAJOKI, FINLAND Phone (02) 8372 31 (nat.), (+358-2-) 8372 31 (int.) Fax (02) 8372 3809 (nat.), (+358-2-) 8372 3809 (int.)

ISBN 978-951-652-181-0 ISSN 1239-3096

Posiva-raportti – Posiva Report

Raportin tunnus – Report code

POSIVA 2011-04

Posiva Oy Olkiluoto FI-27160 EURAJOKI, FINLAND Puh. 02-8372 (31) – Int. Tel. +358 2 8372 (31) Tekijä(t) – Author(s)

Julkaisuaika – Date

December 2011

Toimeksiantaja(t) – Commissioned by

Natalia Pimenoff, Ari Venäläinen & Heikki Järvinen, Finnish Meteorological Institute

Posiva Oy

Nimeke – Title

CLIMATE SCENARIOS FOR OLKILUOTO ON A TIME-SCALE OF 120,000 YEARS Tiivistelmä – Abstract

Posiva Oy is planning to dispose of spent nuclear fuel in a repository, to be constructed at a depth of 400 m in the crystalline bedrock at Olkiluoto, Finland. Planning the storage requires careful consideration of many aspects, including an assessment of long-term repository safety. For estimating possible climate states at Olkiluoto on a time-scale of 120,000 years, we analyze climate simulations of an Earth System Model of Intermediate Complexity (CLIMBER-2) coupled with an ice sheet model (SICOPOLIS). The simulations into the future clearly show that the onset of the next glaciation is strongly dependent on the Earth’s orbital variations and the atmospheric CO2 concentration. It is evident that due to global warming, the climate of the next centuries will be warmer and wetter than at present. Most likely, due to global warming and low variations in the Earth’s orbit around the sun, the present interglacial will last for at least the next 30,000 years. Further, the future simulations showed that the insolation minima on the Northern Hemisphere 50,000–60,000 and 90,000–100,000 years after the present hold a potential for the onset of the next glaciation. Hence, on a time-scale of 120,000 years, one must take into account climate periods lasting several thousand years having the following features: an interglacial climate, a periglacial climate, a climate with an ice sheet margin near Olkiluoto, a glacial climate with an ice sheet covering Olkiluoto, and a climate with Olkiluoto being depressed below sea level after glaciation due to isostatic depression. Due to the uncertainties related to the evolution of the future climate, it is recommended the simulations into the far future to be used only qualitatively. Quantitative information about glacial climate is achieved from the reconstructions and simulations of the past climate.

Avainsanat - Keywords

climate scenarios, regional scale, next glaciation, future climate ISBN

ISSN

ISBN 978-951-652-181-0 Sivumäärä – Number of pages

2

ISSN 1239-3096 Kieli – Language

English

Posiva-raportti – Posiva Report

Raportin tunnus – Report code

POSIVA 2011-04

Posiva Oy Olkiluoto FI-27160 EURAJOKI, FINLAND Puh. 02-8372 (31) – Int. Tel. +358 2 8372 (31)

Julkaisuaika – Date

Joulukuu 2011

Tekijä(t) – Author(s)

Toimeksiantaja(t) – Commissioned by

Natalia Pimenoff, Ari Venäläinen & Heikki Järvinen, Ilmatieteen laitos

Posiva Oy

Nimeke – Title

ILMASTOSKENAARIOITA OLKILUOTOON 120 000 VUODEN AIKASKAALASSA Tiivistelmä – Abstract

Posiva Oy suunnittelee käytetyn ydinpolttoaineen loppusijoitusta loppusijoitustilaan Olkiluodon kallioperään 400 m syvyyteen. Loppusijoitustilan suunnittelu vaatii laajaa tutkimusta, mukaan lukien arvion loppusijoitustilan pitkäaikaisturvallisuudesta. Arvioidaksemme mahdollisia ilmastonvaihteluita Olkiluodossa seuraavien 120 000 vuoden aikana, tarkastelemme ilmastosimulaatioita, jotka on tehty CLIMBER-2 ilmastomallilla kytkettynä SICOPOLIS jäätikkömalliin. Tulevaisuuteen tehtyjen ilmastosimulaatioiden perusteella seuraavan jääkauden alkaminen riippuu suuresti maapallon kiertoradan vaihteluista sekä ilmakehän hiilidioksidipitoisuudesta. Näyttää ilmeiseltä, että ihmisen aiheuttaman ilmaston lämpenemisen seurauksena maapallon ilmasto tulee olemaan nykyistä lämpimämpi ja sateisempi seuraavien vuosisatojen aikana. Ilmaston lämpeneminen ja vähäiset vaihtelut maan kiertoradassa auringon ympäri johtavat todennäköisesti siihen, että nykyinen interglasiaali jatkuu vielä ainakin seuraavat 30 000 vuotta. Tulevaisuuteen tehtyjen ilmastosimulaatioiden perusteella seuraava jääkausi voisi alkaa pohjoisen pallonpuoliskon negatiivisen säteilypoikkeamien aikana 50 000 – 60 000 vuoden tai 90 000 - 100 000 vuoden kuluttua. 120 000 vuoden aikaskaalassa onkin otettava huomioon seuraavanlaisia usean tuhannen vuoden pituisia ilmastojaksoja: interglasiaalinen, periglasiaalinen, ilmasto jonka aikana jään reuna on Olkiluodon lähellä, jääkauden ilmasto jonka aikana mannerjäätikkö peittää Olkiluodon sekä jääkauden jälkeinen ilmasto jolloin Olkiluoto on painuneena merenpinnan alapuolelle. Johtuen kaukaiseen tulevaisuuteen tehtyjen simulaatioiden epävarmuudesta, suosittelemme tämän raportin tulevaisuussimulaatioita käytettävän ainoastaan kvalitatiivisesti. Kvantitatiivista tietoa jääkauden ilmastosta saadaan menneiden ilmastojen rekonstruktioista ja uudelleen simulaatioista.

Avainsanat - Keywords

ilmastoskenaariot, alueellinen mittakaava, seuraava jääkausi, tulevaisuuden ilmasto ISBN

ISSN

ISBN 978-951-652-181-0 Sivumäärä – Number of pages

2

ISSN 1239-3096 Kieli – Language

Englanti

PREFACE This study is prepared to support the formulation of scenarios in the Safety Case. The report was prepared at the Finnish Meteorological Institute and funded by Posiva Oy. The simulations with CLIMBER-2 coupled to SICOPOLIS were performed by the Potsdam Institute for Climate Impact Research. The simulations with the RCA3 model were performed by the Swedish Meteorological and Hydrological Institute. The contact persons at Posiva Oy are Ari Ikonen and Anne Lehtinen; at the Finnish Meteorological Institute they are Natalia Pimenoff, Heikki Järvinen and Ari Venäläinen.

1

TABLE OF CONTENTS ABSTRACT TIIVISTELMÄ PREFACE LIST OF ABBREVIATIONS AND KEY WORDS............................................................. 3 EXECUTIVE SUMMARY ................................................................................................ 5 1

INTRODUCTION .................................................................................................... 9

2

CLIMATE AND CLIMATE CHANGE ..................................................................... 11 2.1 The climate system ........................................................................................ 11 2.2 Past climate changes ..................................................................................... 13 2.2.1 Glacial-interglacial variability and dynamics ............................................ 14 2.2.2 Past sea level variations ......................................................................... 17 2.2.3 The last glaciation in Fennoscandia and in Olkiluoto .............................. 18 2.3 Atmospheric CO2 concentration ..................................................................... 22 2.3.1 The natural carbon cycle ......................................................................... 22 2.3.2 Changes in the atmospheric CO2 concentration before the industrialized . era ........................................................................................................... 24 2.3.3 Human interference in the carbon cycle.................................................. 25 2.4 Future climate scenarios ................................................................................ 27 2.4.1 Future climate scenarios on a time-scale of centuries ............................ 28 2.4.2 Future climate scenarios on a time-scale of 120,000 years .................... 29

3

MATERIALS AND METHODS FOR THE 120,000 YEAR TIME-SCALE CLIMATE SCENARIOS FOR OLKILUOTO........................................................................... 35 3.1 Global model simulations with CLIMBER-2 .................................................... 35 3.2 Statistical downscaling – Generalized Additive models ................................. 38 3.2.1 Time-slice simulations with the regional model RCA3 ............................ 39 3.2.2 Climate Research Unit observational data .............................................. 44 3.2.3 NOAA terrain data ................................................................................... 44 3.2.4 Fitting the statistical model ...................................................................... 45

4

LAST GLACIAL CYCLE SIMULATIONS .............................................................. 49 4.1 Simulated ice sheet and climate evolution in the baseline experiment .......... 49 4.2 Sensitivity studies as a basis for the model parameter settings ..................... 54 4.3 Scenario with ice-free conditions at Olkiluoto during the MIS 3 ..................... 55 4.4 Simulated ice sheet retreat............................................................................. 56

5

SIMULATIONS OF THE FUTURE USING CLIMBER-2 ....................................... 59 5.1 Simulation into the future with a constant CO2 concentration of 280 ppm ..... 61 5.2 Simulation into the future with a constant CO2 concentration of 400 ppm ..... 63

6

DISCUSSION AND CONCLUSIONS.................................................................... 65

7

ACKNOWLEDGEMENTS ..................................................................................... 69

REFERENCES ............................................................................................................. 71 Appendix 1

RCA3 model setup and results ............................................................. 83

2

Appendix 2 Appendix 3 Appendix 4

Evaluation of the statistical model ........................................................ 87 Deliverables and data ........................................................................... 99 Publication resulting from the project.................................................. 101

3

LIST OF ABBREVIATIONS AND KEY WORDS AD AP BP CCSM3 CLIMBER CMAP CPC CRU DIC EMIC EPICA FMI GAM GCM GIS Gton C IPCC kyr LGM LLN 2-D (NH) LPJ-GUESS MIS MoBidiC MPM NGDC NGRIP NH NOAA PETM PIK ppb ppm RCA3 SEMI SICOPOLIS SMHI SRES WAIS

Anno Domini After Present Before Present Community Climate System Model version 3 CLIMate-BiosphERe model Climate prediction center Merged Analysis of Precipitation Climate Prediction Center Climate Research Unit Dissolved Inorganic Carbon Earth system Model of Intermediate Complexity European Project for Ice Coring in Antarctica Finnish Meteorological Institute Generalized Additive Model General Circulation Model Greenland Ice Sheet Gigatonnes of carbon Intergovernmental Panel on Climate Change 1,000 years Last Glacial Maximum The Louvain-la-Neuve climate model Lund-Potsdam-Jena - General Ecosystem Simulator Marine Isotope Stage an improved version of the LLN 2D NH climate model green McGill paleoclimate model National Geophysical Data Center North Greenland Ice Core Project Northern Hemisphere National Oceanic and Atmospheric Administration Palaeocene-Eocene Thermal Maximum Potsdam Institute for Climate Impact Research parts per billion (10-9) parts per million (10-6) Rossby Centre Regional Climate Model Surface Energy Mass-balance Interface SImulation COde for POLythermal Ice Sheets Swedish Meteorological and Hydrological Institute Special Report on Emission Scenarios West Antarctic Ice Sheet

4

5

EXECUTIVE SUMMARY This report aims to estimate plausible future climate conditions in Olkiluoto, Western Finland, on a time-scale of 120,000 years. In general, climate modeling is the only available method to accomplish this task. The strength of these simulation models is that they are based on the fundamental geophysical principles of the Earth’s climate system. Their weakness is that the external forcing factors driving the evolution of the climate system need to be known sufficiently accurately. Some of these are well known, such as the future insolation, while some others, such as the atmospheric chemical composition, are much less accurately known. Thus, the report refers to climate simulations of the future which are, in fact, projections of the future forcing factors into climate conditions. The projections are accurate within the modeling uncertainty, and uncertainty in the specification of the forcing factors. The former can be estimated with model validation, while little can be done about the latter. An important model validation technique is to assess the models’ capability to simulate the past and present climate conditions. Our knowledge about the past climate in the timescale of tens of thousands of years ago is not very detailed. This makes the validation of past climate variations somewhat challenging. Nevertheless, confidence emerged towards the models such that the modeling uncertainty is very likely smaller than the forcing uncertainty regarding the aggregate uncertainty of the climate projections of the future. Past and present climate conditions Instrumental records of the present and recent past climate cover only the last 150 years. Information of the past climate prior to instrumental records is obtained from proxy data. The principal data sources for paleoclimatic reconstructions are glaciological (ice cores), geological (ocean and terrestrial sediments), biological (tree rings, pollen) and historical (written and phenological records) proxies. Uncertainties related to the paleoclimatological proxy data are, for instance, timing uncertainties, sediment disturbances, and uncertain relationships between the climate variables and the proxies. The proxy data from the ice cores and ocean sediments document that, during the last 650,000 years, Earth’s climate has varied between glacial and interglacial conditions with a strong periodicity of approximately 100,000 years. Within the glaciations, the climate varied between cold periods, stadials, and relatively warm periods, interstadials, that lasted several hundreds or thousands of years. During the cold stadials, ice sheets typically grew and during the warmer interstadials they shrank. The area of Olkiluoto has been covered by ice sheets several times during the past glaciations, the latest having been about 9,500 years ago. Currently, the Earth’s climate is in an interglacial. Quaternary studies provide evidence that past glacial-interglacial variations have been largely driven by the Earth’s orbital changes. The orbital theory states that glaciations are triggered by minima in summer insolation in the Northern high latitudes. This enables the snow aggregated during the winter to stay over the summer and therefore to gradually accumulate, generating the Northern Hemisphere continental ice sheets.

6

During the Last Glacial Maximum (LGM) about 20,000 years ago, large amounts of fresh water was stored in the ice sheets such that the global mean sea level was about 120 m lower than at present. During the LGM, the global mean temperature is estimated to have been 4 to 7 ÛC lower than the pre-industrial climate. The well-known insolation variations do not alone explain these variations. They have indeed been amplified by the interactions and feedbacks of the Earth’s climate system. These include foremost the responses of the carbon cycle (redistribution of carbon between the atmosphere and ocean), the hydrological cycle (ice-albedo feedback, changes in the ocean circulation) and the terrestrial biosphere (albedo effect, and CO2 fertilization effect). Proxy data from ice cores reveal that during the glacial cycles the atmospheric CO2 concentration has varied between 180 ppm and 280 ppm. It has been estimated, that during the LGM, about half of the global cooling resulted from the presence of the Northern Hemisphere ice sheets and the rest from to the decreased atmospheric CO2 concentration, shrunken vegetation cover and increased atmospheric dust content. During the recent 250 years, the anthropogenic carbon release into the atmosphere has been about 300 Giga tonnes (Gton C) and the atmospheric CO2 concentration has increased from the pre-industrial value of about 280 ppm to 388 ppm (in year 2010). This is considered to be the single largest factor contributing to the observed global mean temperature rise of ca. 0.8 ÛC during the last 150 years. The potential for future release from all known conventional fossil fuel reserves is estimated to be at least 5,000 Gton C. The Earth’s climate is projected to warm further during the next centuries, as the anthropogenic emissions continue to rise the atmospheric CO2 concentration. Climate modeling For simulating climate on a time-scale of 100,000 years, Earth system models of intermediate complexity (EMICs) are utilized. These models lie (in terms of their complexity) between simple energy-balance models and comprehensive general circulation models (GCMs). EMICs have a rather low spatial resolution enabling very long simulations with reasonable computing time. These models include simplified component models for the atmosphere, ocean, sea ice, land surface, terrestrial vegetation, and ice sheets. These interact under prescribed solar forcing (insolation). The main limitation is related to the lack of fully interactive carbon cycle. Thus, the atmospheric CO2 concentration has to be specified, too. Instead of just one simulation, many simulations with varying forcing factors are considered. This reveals the sensitivity of the simulated climate (or, climate projection) on the specification of the forcing factors. Uncertainties of the climate projections Temporal variations of the Earth’s orbital parameters are known with a high degree of certainty. In contrast, the future evolution of the atmospheric CO2 concentration is highly uncertain. First, there is no way of knowing the future anthropogenic influence on the Earth’s climate system (carbon emissions, land use, “climate engineering”). Second, modelling of the carbon cycle still contains large uncertainties. Therefore, the only realistic approach at this stage is to prescribe the atmospheric CO2 concentration based on, e.g., emission scenarios derived from socioeconomic studies. This enables us

7

to study a limited set of possible scenarios of the future as a base for climate projections of the future. All climate and Earth system models are idealizations of the real climate system. The models do not fully reproduce all interactions and feedbacks. Sub-grid scale phenomena are dealt with parameterizations, necessarily including some closure schemes. This makes it possible to “tune” the models to some extent reduce modeling biases. It is unclear how model tuning against past observations will verify in the unobserved future. We have to trust that the model tuning is not excessive. Multi-model approach (similar scenario, different models) is one approach to explore this uncertainty (i.e., the climate sensitivity of the models). This approach is adopted here.  The projections of the future climate cannot be verified, because there are no observations. Several EMIC models: CLIMBER-2 (CLIMate and BiospheRE, version 2), MPM (McGill Paleoclimate Model), LLN-2D (a two-dimensional climate model developed in Louvain-la-Neuve) and MobiDiC (an improved version of the LLN-2D) have successfully been able to simulate the last glacial inception. This does not, however, alone proof the models’ skills of correctly projecting the future events because the skill may be due to model tuning. Nevertheless, the current paradigm in climate modelling is that more confidence is laid on models that are able to simulate the past and present climate. Main results In all EMIC simulations, the onset of the next glaciation on the Northern Hemisphere strongly depends on the Earth’s orbital variations and on the atmospheric CO2 concentration. According to the simulations, there are three periods during the next 100,000 years with a potential for ice sheet formation: (1) around 10-20 kyr After Present (AP), (2) around 50-60 kyr AP, and (3) 90-100 kyr AP. These periods coincide with the Northern Hemisphere summer insolation minima. Figure ES.1 shows schematically how the EMIC simulated onset of the next glaciation depends on the solar insolation and CO2 concentration: a Northern hemisphere solar insolation minimum triggers a glacial inception, and high CO2 concentration delays the next glacial inception. In the EMIC simulations, a glacial inception did not start during the next 30,000 years if the atmospheric GHG forcing was higher than the typical pre-industrial value of 280 ppm. Therefore, as the present-day greenhouse gas concentrations are high in a glacial perspective, and are projected to stay high for thousands of years, the next glacial inception seems unlikely before 30 kyr AP. After 30 kyr the possibility of the onset of a glaciation increases being highest during the Northern Hemisphere insolation minima at 50–60 kyr AP and 90–100 kyr AP. Sustained high atmospheric greenhouse gas concentration might even further delay the onset of the next glacial.

8

prolonged high emissions

Interglacial

Stadial

Interglacial

Interglacial

Intergl.

June Solar Insolation at 65 o N (Wm-2)

Stadial Interstadial

Stadial

Stadial

Interstadial/ Integl.

Interstadial

Stadial

Interstadial/ Integl.

Interstadial

Stadial

Interstadial/ Integl.

CO2

no effect of the anthropogenic emissions

550 500 450 0

20

40 60 80 Time After the Present (1,000 years)

100

120

Figure ES.1. A schematic figure of the role of the solar insolation and the atmospheric CO2 concentration in the next glacial inception. According to simulations, sustained high greenhouse gas emissions might delay the next glacial inception post 120,000 years after the present. As for simulations with low greenhouse gas emissions, the onset of the next glaciation is determined by the solar insolation, and could occur around 50,000-60,000 or 90,000-100,000 years after the present.

Discussion and conclusions The Earth’s climate system is a non-linear coupled dynamic system. Its evolution is driven by external forcing factors, and it has a capacity for rich internal variability, too. All governing processes are not yet fully understood. Earth system modelling is thus necessarily incomplete. Also, the forcing factors are not sufficiently well known over extended periods relevant for past and future glaciations. Human influence on the planet is highly uncertain. Thus, prediction of future glaciations is excluded. We can, however, project scenarios of the future forcing factors into plausible future climate conditions. We understand the sources of uncertainty and their relative importance. Therefore, although the Earth system models cannot provide us with a complete description of the future evolution of the Earth’s climate conditions, they can significantly limit the set of possible future states. One key finding here is that a glacial inception during the next 100,000 years is possible, but unlikely before 30 kyr AP. For investigating the consequences of future glaciations in detail, we recommend use of paleo-climatological reconstructions and simulations of the past climate, too. This is because of the somewhat qualitative nature of the projections of the future. However, the proxy data does not alone unfold the future climate. The model simulations are the best available estimates of the future climate on a time-scale of 100,000 years. Climate of the next 10,000 years will be studied in more detail in the following report, currently in preparation.

9

1

INTRODUCTION

Posiva Oy is the nuclear waste management organization in Finland responsible for research into the final disposal of spent nuclear fuel and for the construction, operation and eventual backfilling and closure of the final disposal facility. In 2001 the Finnish Parliament ratified the Government‘s favourable Decision in Principle (DiP) on Posiva’s application to locate the repository at Olkiluoto, where the spent fuel from Finnish nuclear power reactors is planned to be disposed of in a storage facility, to be constructed at a depth of 400 m in the crystalline bedrock. Planning this storage requires careful consideration of many aspects. This study is prepared to support the formulation of scenarios in the Safety Case: it attempts to formulate possible future climate evolutions on a regional scale for Olkiluoto. The aim of this study is to estimate the duration, extent and time frame of cold periods, as well as their probabilities and possible extremes, in different climate scenarios for the next 120 kyr. The climate data will then be available for use as input in downstream computing of the evolution of permafrost, surface hydrology, biosphere and deep groundwater conditions for Olkiluoto. Earth System Models of Intermediate Complexity (EMICs) are used to simulate the future climate on a time-scale of 120 kyr. EMICs are able to simulate a large set of climate processes and feedbacks (Claussen et al. 2002), comparable with that of General Circulation Models. Due to their coarse resolution, EMICs have a fast computational speed. Simulations with EMICs have shown that the models have strong sensitivity to the atmospheric carbon dioxide (CO2) concentration. The higher the atmospheric CO2 concentration in the simulations, the warmer the climate will be and the later the next glaciation (Berger and Loutre 2002; Cochelin et al. 2006; Archer and Ganopolski 2005). According to the simulations of Loutre and Berger (2000), no glacial episode is expected to begin within the next 50 kyr even under a natural atmospheric CO2 regime. Simulations with different EMICs by Loutre and Berger (2000), Archer and Ganopolski (2005), Texier et al. (2003) and Cochelin et al. (2006) have shown that sustained high atmospheric greenhouse gas concentrations may further delay the onset of the next glacial period. There are many uncertainties related to future atmospheric CO2 concentration, such as the amount of the future anthropogenic emissions and the ability of the climate system to take up this extra carbon. Further, ice core records show that over the last 650 kyr the CO2 concentration has decreased during glacial periods about 45–95 ppm (EPICA community members 2004), the quantitative and mechanistic explanation for these variations is still unknown. As the uncertainties related to the future atmospheric CO2 concentrations are large, we use scenarios. For formulating future climate scenarios for Olkiluoto, we use simulations performed with the CLIMBER-2 EMIC model (Petoukhov et al. 2000; Ganopolski et al. 2001) coupled with the SICOPOLIS ice sheet model (Greve 1997; Calov et al. 2005a). The simulations were performed by the Potsdam Institute for Climate Impact Research. Three simulations are explored in detail. The first is a simulation of the last glacial cycle

10

(Ganopolski et al. 2010), simulating the period from 126 kyr before present (BP) until the present with greenhouse gas concentrations derived from the Vostok ice core. The other two simulations cover the future 120 kyr. As a fully coupled climate - ice sheet carbon cycle model for this time-scale is not yet available, the future climate scenarios are with constant CO2 concentration. The utilized atmospheric CO2 concentrations in these simulations into the future (Ganopolski et al. unpublished) are 280 ppm, which is typical of natural interglacial conditions, and 400 ppm, a concentration that is easily reached within a few decades if anthropogenic carbon dioxide emissions continue. For our purposes the data of the CLIMBER-2 model simulations have too coarse spatial resolution, as there is only one grid point over Fennoscandia. Therefore, we downscale the large-scale CLIMBER-2 climate data with a regression model, a Generalized Additive Model (GAM, Wood 2006, Vrac et al. 2007, Martin et al. 2010a and 2010b). For fitting the regression equations we utilize observations of the present climate and results of short time-slice simulations (Kjellström et al. 2009) performed by the Swedish Meteorological and Hydrological Institute with the regional climate model RCA3. This report is organized as follows. In Chapter 2 we summarize the background knowledge of the Earth’s climate system, climate forcings and the last glaciation in Fennoscandia, and introduce projections of the future climate. In Chapter 3 we introduce our research material and fitting of the regression models for downscaling the large-scale CLIMBER-2 simulations to the regional scale. In Chapter 4 we display results of the CLIMBER-2 simulation of the last glacial cycle. In Chapter 5 we show the results of the two future simulations with CLIMBER-2: the climate of the next 120 kyr in a scenario in which the atmospheric CO2 concentration is a constant i) 280 ppm and ii) 400 ppm. In Chapter 6 we discuss our findings, compare them to other studies and propose suggestions on their use.

11

2

CLIMATE AND CLIMATE CHANGE

2.1

The climate system

The Earth’s climate system consists of the atmosphere, the hydrosphere, the cryosphere, the geosphere and the biosphere. Climate is the ‘average weather’ over a certain period of time, usually 30 years; however, it could also be defined over a longer period of time (e.g., millions of years). The climate system is continuously evolving under the influence of external factors (climate forcings) and of its own internal dynamics. Internal climate forcings include interactions and feedback effects between the climate components, e.g., atmosphere-biosphere interaction and the ice-albedo feedback effect. External climate forcings include anthropogenic changes in the composition of the Earth’s atmosphere, as well as natural phenomena such as solar variations and volcanic eruptions. The solar radiation and its variations power the Earth’s climate system. The radiation balance of the Earth can be altered by: 1) changing the incoming solar radiation (e.g., by changes in the Earth’s orbit around the sun); 2) changing the fraction of solar radiation that is reflected (called the ‘albedo’; e.g., by changes in snow cover, clouds, atmospheric particles or vegetation); and 3) altering the long-wave radiation from the Earth back into space (e.g., by changing greenhouse gas concentrations or cloud cover). The climate system responds to such changes directly or indirectly through a variety of feedback mechanisms. The amount of incoming solar radiation at the top of the atmosphere averaged over the entire planet is 342 Wm-2 on an annual mean basis (see Figure 1). About 30 % of the incoming solar radiation that reaches the top of the atmosphere is reflected back into space. More than two-thirds of this reflection is due to clouds and aerosols, i.e., small particles in the atmosphere. The remaining one-third is reflected by the Earth’s surface: especially snow, ice and deserts. Changes in these reflective components of the climate system affect the climate. For example, major volcanic eruptions have a dramatic effect on the climate by ejecting material into the stratosphere far above the highest clouds (even as high as 30 km). It takes about a year or two before these aerosol particles fall into the troposphere and are deposited to the surface by precipitation. During this time the aerosols affect the reflectivity of the atmosphere and can cause a drop in the global mean surface temperature.

12

107

Reflected Solar Radiation 107 Wm-2

235

Incoming Solar Radiation 342 Wm-2

342

Refelected by Clouds, Aerosol and Atmospheric gases 77 Wm -2

Emitted by Atmosphere

165

Emitted by Clouds

30

Absorbed by Atmosphere 24

168 Wm -2 Absorbed by the Surface

24 Thermals

40 Atmospheric Window Greenhouse Gases

Latent 78 Heat 350

Reflected by Surface 30

Outgoing Longwave Radiation 235 Wm -2

390 78 Surface EvapoRadiation transpiration

40

324 Back Radiation

324 Absorbed by Surface

Figure 1. Estimate of the Earth’s annual and global mean energy balance. Modified from Kiehl & Trenberth (1997). Units are Wm-2. That part of the solar energy that is not reflected back to space is absorbed by the Earth’s surface and the atmosphere. To balance the incoming energy, the surface of the Earth radiates energy back into the atmosphere in the form of long-wave thermal radiation. Some of this long-wave radiation passes right through the atmosphere. However, a major part of it is absorbed and re-emitted in all directions by atmospheric greenhouse gases. This phenomenon, the natural greenhouse effect, warms the Earth’s surface by about 33 °C with the result that the global mean temperature is about 14 °C. The most important natural greenhouse gases are water vapour and carbon dioxide (CO2). Clouds too emit some long-wave radiation, thus participating in the greenhouse effect. However, this effect is more than compensated for by the solar radiation reflected by the clouds, and on average clouds tend to have a cooling effect. Human activities result in changes in the radiative forcing of the climate system (see Figure 2). The anthropogenic emissions of greenhouse gases CO2, methane (CH4), nitrous oxide (N2O), halocarbons intensify the greenhouse effect and warm the climate. The total radiative forcing of these long-lived greenhouse gases is estimated to be about 2.6 Wm-2 from which the contribution of CO2 is estimated to be about 1.6 Wm-2 (see Figure). Climate change is already observed, for example the global mean temperature has increased by ca. 0.8 °C during the last 150 years (IPCC 2007). In addition to greenhouse gases, human activities result in aerosol particle emissions. The aerosol particles have both cooling and warming effects with a net cooling effect. This cooling effect has partly compensated the warming effect of the increased amount of greenhouse gases; however, the net antropogenic radiative effect is estimated to be about 1.6 Wm-2.

13

Figure 2. Summary of the principal components of the radiative forcing of climate change. The values represent the forcings in 2005 relative to the start of the industrial era (about 1750). Positive forcings lead to warming of climate and negative forcings lead to a cooling. The thin black line attached to each coloured bar represents the range of uncertainty for the respective value. Modified from Forster et al. (2007).

Because of the Earth’s almost spherical form, most of the incoming solar energy is received in the tropics. Atmospheric and oceanic circulations transport the energy from the equatorial areas to higher latitudes. When water is evaporated from the sea and land surface, the water vapour absorbs energy, latent heat, which is released as this water vapour condenses into clouds (see Figure 1). In the tropics, this energy released from the latent heat is the primary driver of the atmospheric circulation. The ocean circulation is driven by the surface winds and through changes in the ocean’s surface temperature and salinity through precipitation and evaporation. 2.2 Past climate changes The global climate is changing continuously. The causes and the time-scales of the climate changes vary (see Table 1). Changes in the various components of the climate system, such as the size of the ice sheets, the type and distribution of vegetation or the temperature of the atmosphere or the ocean, all influence the large-scale circulation features of the atmosphere and oceans. The slowest climate changes take hundreds of millions of years, while the fastest happen in decades. An important factor altering the

14

global climate during the Earth’s past 4.5 billion years has been the continental drifts, which alter the Earth’s land form and land-ocean distribution, also affecting the composition of the atmosphere. The continental drifts likewise affect the ocean currents. The drifts are considered to be the fundamental reason for the glacial times that have occurred on the time-scales of hundreds of millions of years. Table 1. Time-scales of the phenomena causing climate change. Source: Harvey (2000). Driving phenomenon Continental drifts, uplift of mountains Changes in Earth’s orbit Changes in greenhouse gas concentrations (e.g., metamorphosis of sea-floor sediments, weathering, oceanic biological pump ) Changes in solar activity Internal climate variability Individual volcanic eruptions

Time-scale (years) 10,000,000 – 100,000,000 10,000 – 100,000 100 – 100,000,000

10 – 1,000,000,000 10 – 1,000 1–4

There are many feedback mechanisms in the climate system that can either amplify (give ‘positive feedback’) or diminish (give ‘negative feedback’) the effects of climate forcing. For example, as increasing concentrations of greenhouse gases warm the Earth’s climate, snow and ice begin to melt. This melting reveals darker land and water surfaces that have previously been beneath the snow and ice, and these darker surfaces absorb more of the solar radiation, causing more warming, which causes more melting, and so on, in a self-reinforcing cycle. This feedback loop, known as the ‘ice-albedo feedback’, amplifies the initial warming caused by increasing concentrations of greenhouse gases. Sufficient instrumental records of past climate cover only the last 150 years. Information of the past climates prior to instrumental records is achieved from proxy data. The principal data sources for paleoclimatic reconstructions, as listed by Brandley (1999, p. 5), are glaciological (ice cores), geological (marine and terrestrial sediments), biological (tree rings, pollen) and historical (written and phonological records) proxies. Uncertainties related to the paleoclimatological data are, e.g., timing uncertainties, sediment disturbances, and uncertain relationships between the climate variables and the proxies. 2.2.1

Glacial-interglacial variability and dynamics

The proxy data from ice cores and ocean sediments document that, during the last 650 kyr, the climate has varied from glacial to interglacial conditions with a strong, approximately 100 kyr cyclicity (EPICA community members 2004). Within the glaciations, the climate varied between cold periods, stadials, and relatively warm periods, interstadials, that lasted several hundreds or thousands of years. During the cold stadials, ice sheets usually grew and during the warmer interstadials ice sheets usually shrank. In Figure 3, the interglacials of the last 140 kyr have been marked with grey shading on a timeline along with the marine isotope stages (MIS). MISs are warm

15

and cool periods in the past, derived from oxygen isotope data in the deep ocean core samples. Quaternary studies provide evidence that past glacial-interglacial variations have been largely driven by the Earth’s orbital changes (e.g., Crucifix et al. 2006). Three components of orbital variation affect the temporal and geographical distribution of solar insolation: precession at periods of 19 and 23 kyr, obliquity at periods of about 40 kyr and eccentricity at periods of 100 and 400 kyr (Berger 1978; Jansen et al. 2007, see Figure 4). The Milankovitch theory proposes that glaciations are triggered by minima in summer insolation in northern high latitudes, enabling winter snowfall to persist all year and therefore to accumulate, generating the Northern Hemisphere glacial ice sheets. For example, the onset of the last glaciation, about 117 kyr BP (Figure 3b and Figure 3e), corresponds to an insolation minimum in Northern Hemisphere high latitudes (Figure 3a). Climatic interactions (feedbacks) on a global scale have also been involved in the processes of glacial inception and deglaciation. These interactions are, e.g., the responses of the carbon cycle (atmospheric CO2 concentration (Figure 3c), the hydrological cycle (ice-albedo feedback, ocean current changes) and the terrestrial biosphere (albedo, CO2 concentration) (Crucifix et al. 2006). The CO2 concentration in Figure 3c seems to go hand in hand with the temperature proxy in Figure 3b. These CO2 variations have amplified the past glacial cycles by altering the greenhouse effect. It is estimated, that about half of the global cooling during the last glaciations resulted from the decrease in the atmospheric CO2 concentration. In Chapter 2.3 we represent relevant features of the carbon cycle and past atmospheric CO2 concentration variations. Climate changes are related to large spatial variations. For example, during the LGM, the global mean temperature is estimated to have been 4 to 7 ºC lower than the preindustrial climate (Jansen et al. 2007). However, the local cooling during the LGM is estimated to have been 7 to 11 ºC in Central Antarctica (Stenni et al. 2001; MassonDelmotte et al. 2006), about 21 ºC in Greenland (Dahl-Jensen et al. 1998) and only 0 to 3.5 ºC in the Tropical Indian sea (Barrows & Juggins 2005; Rosell-Mele et al. 2004).

560

a)

540 520 500 480 460 440 420

4 0

b)

-120

-100

-80

-60

-40

-20

-4 -8

'cooler' -12

300

c) 260

Global Sea level (m)

220 20 0 -20 -40 -60 -80 -100 -120 -140

CO 2 (ppm)

d 18 O(% o )

'warmer'

Solar insolation 60 oN (W m-2 )

16

180

d)

e)

3.2

'LESS ICE'

4.0 4.4

d18O(%o)

3.6 Proxy for Global Ice Volume

4.8 'MORE ICE'

5.2

140

120

SaalianEemian

6

100

80

60

Early Weichselian

5e 5d 5c 5b 5a

40

Middle Weichselian

4

3

20

0 Time (kyr before 2005)

Late HoloWeichs. cene

2

1

Geochronology Marine isotope stages and substages (MIS)

Figure 3. a) June Solar insolation at 60° N (Berger & Loutre 1991), b) measured anomaly in the 18O concentration from the NGRIP (North Greenland Ice Core Project) ice core, a proxy for local temperature, c) CO2 concentration from the Vostok ice core, (Petit et al. 1999), d) a reconstruction of the global sea level by Waelbroeck et al. (2002) based on deep ocean 18O and temperature records, e) Glacials and interglacials during the last 140 kyr along with the marine isotope stages (MIS) and a proxy for global ice volume. The grey shading indicates interglacials. Modified from Jansen et al. (2007).

17

Figure 4. Schematic of the Earth’s orbital changes (Milankovitch cycles) that drive the ice age cycles. ‘T’ denotes changes in the tilt (or obliquity) of the Earth’s axis, ‘E’ denotes changes in the eccentricity of the orbit (due to variations in the minor axis of the ellipse), and ‘P’ denotes precession, that is, changes in the direction of the axis tilt at a given point of the orbit. Source: Jansen et al. (2007). 2.2.2

Past sea level variations

Changes in continental ice sheets affect directly the sea level. The water to form the ice sheets originates from the oceans. Thus, as the global ice volume increases, the global sea level decreases and vice versa (Figure 3d and Figure 3e). During the coldest stage of the last glaciation, about 20 kyr BP, large water masses were stored in the ice sheets of the Northern Hemisphere, and the global sea level was about 120 m lower than at present (Waelbroeck et al. 2002, Figure 3d). During deglaciation, the ice sheets melt and the water returns to the oceans, increasing the sea level again. Past sea level changes have responded to temperature changes. About 3 million years ago, global temperatures were higher than at the present time, and according to reconstructions, the global sea level was 20–30 m higher than present levels. Even higher temperatures 40 million years ago were associated with a sea level 60–70 m higher than the present-day one. During the last interglacial, the Eemian (about 130–117 kyr BP), the global sea level is estimated to have been 4 to 6 m higher than at present, and the summer temperatures were 3–6 °C warmer (Anderson et al. 2006; Sime et al. 2009). The contribution to this higher sea level from the melting of the Greenland ice sheet is estimated to be about 2 m (Cuffey & Marshall 2000). The mass loss of an ice sheet also affects the Earth’s gravitational field and thereby regional sea levels. For example, the loss of the Greenland ice sheet reduces the gravitational pull in the North Atlantic, hence lowering sea levels in that region but enhancing sea level rises in other regions such as the Pacific Ocean and Southern Hemisphere. Likewise, a mass loss in the West Antarctic ice sheet will lead to an even stronger sea level rise on European and North American coasts compared to the global mean (Mitrovica et al. 2001). Ice sheets affect the local sea level by depressing the land masses beneath them even by hundreds of metres. As the ice sheet retreats, the load on the lithosphere and asthenosphere is reduced, and they rebound back towards their equilibrium levels.

18

These rebound movements are slow: for example, the uplift caused by the termination of the last glaciation is still continuing in the Baltic Sea area. Descent of the sea bed occurs when melting waters from the continental ice sheets return in the ocean. The additional weight of the returning water isostatically depresses the oceanic crust. This leads to a corresponding rise in the continental crust so that the isostatic balance remains between oceans and continents. The calculations of (Kivioja 1967) showed that if an amount melting water corresponding to one metre thick layer on the oceans, was transferred to the ocean, the oceans would deepen on average by 0.25 metres and at the same time the continents would rise on average by 0.6 metres. Thus, the average sea level would rise by only 0.15 metres. However, these isostatic adjustments take millennia and therefore only an equally slow change in sea level will be damped by isostatic adjustment. If the melting of continental ice sheets is more rapid, the earth’s crust is not able follow the change with the same speed, and the sea level could rise on the cost areas a lot more than calculated above. Recent global warming of about 0.8 °C has increased the global sea level by about 0.15–0.20 m during the last century (Church & White 2006). Of this, mountain glaciers and ice caps account for about 0.05 m and oceanic thermal expansion for another 0.05 m. Possible sources for the remaining 0.05–0.10 m are the large ice sheets of Greenland and Antarctica. Measurements have shown that the Greenland Ice Sheet and the West Antarctic Ice Sheet have been losing mass with at an increasing rate during the last 10– 15 years (Velicogna 2009). The East Antarctic ice sheet is, according to present-day knowledge, approximately in balance. 2.2.3

The last glaciation in Fennoscandia and in Olkiluoto

Evidence of the impact of glacial advances in Fennoscandia has been obtained from the two most recent glaciations, the MIS 6 during the Saalian Stage and the Weichselian (MIS 5d–2). During the MIS 6 (which began about 200 kyr BP), an ice sheet covered the entire Fennoscandia and large parts of the North Eurasia. Over Olkiluoto (Figure 5) the ice sheet was in maximum about 3 km thick (Eronen & Lehtinen 1996). The Saalian glacial was terminated by the Eemian interglacial (MIS 5e) about 130 kyr BP. During the Eemian period the climate warmed rapidly, the Saale ice sheet retreated and Finland became ice-free. However, due to isostatic depression of the crust, large parts of Western and Southern Finland, including Olkiluoto, were submerged in the saline Eemian Sea (see Figure 6). During the Eemian climatic optimum, air temperatures at Olkiluoto were 4 to 5 °C higher than the present-day ones (Eronen & Lehtinen 1996) and the global sea level is estimated to have been 4 to 6 m higher (e.g., Sime et al. 2009).

19

70 68 66 64 62

Olkiluoto

60 58 56 12 15 18 21 24 27 30 33 36 39 Figure 5. The location of Olkiluoto on the western coast of Finland.

Figure 6. Schematic figure of the Eemian Sea about 130 kyr BP, modified from Funder et al. (2002). The Eemian interglacial ended by a rapid cooling of climate about 117 kyr BP and the Weichselian glaciation started. During the MIS 5d and MIS 5b stadials Northern Fennoscandia became covered by ice. Most of Southern Finland, however, remained ice-free during the Early Weichselian (Svendsen et al. 2004 and references therein), with cold tundra conditions prevailing. The MIS 5d and MIS 5b stadials were separated by a warmer MIS 5c interstadial, and the MIS 5b stadial was followed by a warmer MIS 5a interstadial. At the beginning of the MIS 4 stadial (ca. 70–60 kyr BP) the climate cooled and the Fennoscandian ice sheet spread over most of Fennoscandia. After the MIS 4 stadial, the

20

glacier retreated at least from Southern Finland. During the warmer MIS 3, the climate varied from interstadial to stadial conditions, and ice-free conditions with tundra type vegetation prevailed in Southern and Central Finland for several thousands of years (Ukkonen et al. 1999; Lunkka et al. 2001; Lunkka et al. 2008) with annual mean temperatures 2–6 °C lower than at present (Arppe & Karhu 2006). During the coldest stage of the last glaciation (LGM), the MIS 2 stadial at about 20–18 kyr BP, the Fennoscandian ice sheet grew to its maximum during the Weichselian (see Figure 7). Lunkka et al. (2001) suggested that the Fennoscandian ice sheet grew from Southern Finland after 25 kyr BP in only 7 kyr to the LGM position some 1,000 km southeast in the northwestern Russian Plain. According to the ICE-5G model by Peltier (2004), the ice sheet height, during the LGM, in the Finnish west coast, was about 2 km above present day sea level. During the LGM, the Olkiluoto area is estimated to have been depressed by about 600 m due to glacial loading (Eronen et al. 1995) based on sea level displacement curve fitted to observations in the Olkiluoto – Lake Säkylän Pyhäjärvi area. The simulated annual mean surface air temperatures above the ice sheet over Olkiluoto during the LGM were of the order of -20 °C (Siegert et al. 1999). After the glacial maximum, at about 18 kyr BP, the climate warmed and the glaciers started to melt rapidly. The ice sheet did not retreat at a constant speed. Lunkka et al. (2001) reported that the ice sheet retreated from its maximum position in North-Western Russia to the Bothnian Bay in 8 kyr. A temporal retreat speed of 60–80 m/year was reported (from Lake Kubenskoye to the northern parts of Lake Ladoga over 5 kyr). During the Younger Dryas stadial (about 12.7–11.5 kyr BP) the retreat of the glaciers stopped temporarily for about 1 kyr. After the insolation maximum of the Northern Hemisphere 11.5 kyr BP, the climate warmed so much that the Earth entered the current interglacial, the Holocene.

Figure 7. The maximum ice sheet extent during the Pleistocene (the epoch from 2.588 million to 11.5 kyr BP) and at the last glaciation (Data source: Ehlers & Gibbard 2004).

21

The Baltic Sea experienced rapid and extreme changes as the Fennoscandian ice sheet retreated (see Figure 8). At about 12.6–10.3 kyr BP a freshwater lake, the Baltic Ice Lake (Figure 8a), was gradually formed in the Baltic Sea basin. As the Fennoscandian ice sheet retreated from Central Sweden at about 10.3 kyr BP, straits opened from the Baltic Sea basin to the ocean. This started the brackish water Yoldia Sea stage (Figure 8 b). Due to isostatic land uplift the straits closed up about 9.5 kyr BP and the Yoldia Sea turned into the fresh water Ancylus Lake (Figure 8c). The Fennoscandian ice sheet retreated from the Olkiluoto area at the end of the Yoldia Sea stage, about 9.5 kyr BP, but the area was depressed from the weight of the ice sheet, and remained submerged for about 6 kyr. The Ancylus Lake was transformed into the Litorina Sea (Figure 8d) at about 7.5 kyr BP when the eustatically-rising ocean broke through the Danish Straits (Björck 1995). After that the Baltic Sea turned little by little into a brackish water body. The isostatic adjustment continued, and the Olkiluoto area emerged from the sea 2.5–3 kyr BP (Eronen & Lehtinen 1996).

Figure 8. Schematic sequence of the development of the Baltic Sea. a) Baltic Ice Lake, b) Yoldia Sea, c) Ancylus Lake modified from Björck (1995) and d) Litorina Sea modified from Eronen (1990).

Temperature variations have also occurred at Olkiluoto also during the last 3,000 years. For example, Northern Hemisphere experienced an on average warmer climate in the ‘Medieval warm period’ between AD 1000 and 1200. The medieval warm period is considered to have been heterogeneous in terms of spatial and temporal extent of the warming. It is estimated that the Northern Hemisphere mean temperatures were 0.1 ºC to 0.2 ºC below the 1961 to 1990 mean (Jansen et al. 2007). A colder than average period prevailed in Europe during the ‘Little Ice Age’, the period between AD 1600 and 1850. During the last 150 years the global mean temperature has increased ca. 0.8 °C (IPCC 2007) and over the same period the annual mean temperature of Finland has risen ca. 1 °C.

22

2.3

Atmospheric CO2 concentration

During the industrialized era (the last 250 years), humankind has increased the amount of the atmospheric CO2 primarily by the combustion of fossil fuels and deforestation, but also by cement production and other land-use changes. Due to human activities, the atmospheric CO2 concentration has risen from its pre-industrial values of 280 ppm to about 388 ppm (Tans 2010) during the last 250 years. While human activities affect the climate change in many direct and indirect ways, the CO2 emissions are considered the single largest anthropogenic factor contributing to climate change. The atmospheric CO2 concentration is a key factor causing uncertainty in the future climate scenarios. In this chapter we present the known key factors affecting the atmospheric CO2 concentration. 2.3.1

The natural carbon cycle

Carbon is continuously flowing between the oceans, the atmosphere and land reservoirs. In Figure 9 the natural or unperturbed carbon exchanges (estimated to be those prior to 1750) are shown in black and the anthropogenic in red. It is estimated that prior to the industrialized era (before the year 1750) there was about 597 Gton carbon in the atmosphere (see Figure 9). In the year 2000 the amount of carbon in the atmosphere was about 780 Gton. About 99.5 % of this carbon is in the form of carbon dioxide molecules and the rest in methane molecules. Carbon is stored in the terrestrial biosphere in various chemical compositions to approximately the same amount as in the atmosphere. In soil and detritus the amount of carbon is somewhat higher than in the atmosphere. In the oceans there is about 15 times more carbon than in the atmosphere, vegetation and soil altogether. However, most of the carbon stored in the ocean is deep below the surface, in layers that are not directly in contact with the atmosphere. In the surface ocean layers, which are directly in contact with the atmosphere, there is roughly the same amount of carbon as in the atmosphere. The marine biota stores a relatively small amount, about 3 Gton of carbon. Nonetheless, the marine biota controls the amount of atmospheric carbon dioxide very effectively. Terrestrial vegetation converts atmospheric carbon into plant biomass by photosynthesis at the rate of about 120 Gton C per year. Plant, soil and animal respiration (including decomposition of dead biomass) returns carbon to the atmosphere as CO2 under oxidizing conditions, or as CH4 under anaerobic conditions. About 0.2 Gton C per year is removed from the atmosphere by the weathering of carbonate and silicate rocks. A small amount (about 0.8 Gton C per year) of carbon is transported from the land to the oceans via rivers, either dissolved or as suspended particles (e.g., Richey 2004). Carbon dioxide is continuously exchanged between the atmosphere and the ocean. The direction of the exchange is determined by the carbon dioxide partial pressure difference between the atmosphere and the ocean surface layer. In the atmosphere the partial pressure of CO2 is approximately the same everywhere, but in the surface layers of the ocean it varies. In ocean areas where the carbon dioxide partial pressure is higher than that in the atmosphere, carbon dioxide is released into the atmosphere to the extent of about 90 Gton C per year. Approximately the same amount of carbon is dissolved into the ocean in areas where the difference in the partial pressure is in the opposite sense.

23

Only a small amount of the CO2 dissolved in the ocean remains there as CO2. The rest reacts with water to form bicarbonate (HCO3–) and carbonate (CO32–) ions. Carbon dioxide, HCO3– and CO32– are collectively known as dissolved inorganic carbon (DIC). At present only 1 % of all inorganic carbon in the oceans is carbon dioxide. Phytoplankton in the ocean surface layers take up carbon through photosynthesis. Some of that is removed from the surface layer as dead organisms and particles sinks. This ‘biological pump’ removes carbon from the surface layers to the deep ocean at a rate of about 10 Gton C per year. Because of the biological pump, the total amount of the dissolved inorganic carbon (DIC) in the ocean surface layers is clearly smaller than deeper in the ocean. Because this decreases the CO2 concentration in the surface waters, it also decreases the atmospheric CO2 concentration. It is estimated that the biological pump alone keeps the atmospheric CO2 concentration 50% lower than it would otherwise be. A small amount of carbon, about 0.2 Gton C per year, is deposited in deep ocean sediments. These calcium carbonates form limestone rocks. If this sedimentation process were to act alone, it would take up all the carbon in the oceans, atmosphere and land areas in 200,000 years. However, on a time-scale of hundreds of millions of years, the carbon stored in the limestone is eventually released back into the atmosphere through volcanic eruptions and cracks in the rock. Fossil fuels are formed on a time-scale of hundreds of millions of years from plants and ocean biota that have been buried in sediments. The heat and the pressure have transformed the remains into gaseous and liquid hydrocarbons.

Atmosphere 597 + 165 Weathering

Respiration

0.2

119.6

GPP 120

2.6

Land sink

(~780 yr. 2000) 1.6

Land use change

6.4 70.6

70

22.2

20

Vegetation, Soil and Detritus 2300 + 101 - 140

Fossil fuels 3700 - 244 Volcanoes etc.

Weathering by rivers 0.8

50

Surface urfa Ocean 900 + 18 90.2

101

39 1.6

Intermediate and Deep Ocean 37,100 + 100 Reservoir sizes in GtC Fluxes and rates in GtC per year

GPP = Gross Primary Production

Marine Biota 3 11 Geological Reservoirs

0.2

Surface sediment 150

Figure 9. The global carbon cycle for the 1990s, showing the main annual fluxes in GtC yr–1: pre-industrial ‘natural’ fluxes are shown in black and ‘anthropogenic’ fluxes in red (modified from Denman et al. 2007, figure 7.3)

24

2.3.2

Changes in the atmospheric CO2 concentration before the industrialized era

Prior to the industrialized era (i.e., before 1750), the atmospheric CO2 concentration had been stable between 260 and 280 ppm for 10,000 years (Figure 10a). Conversely, further in the past, the atmospheric CO2 concentration has varied vastly. Ice core records show that during the last 650,000 years the atmospheric CO2 concentration has varied in the range of 180 ppm to 300 ppm over the glacial-interglacial cycles (Petit et al. 1999, EPICA community members 2004, Figure 3c and Figure 10b). During the interglacials, the atmospheric CO2 concentration was in the range of the pre-industrial values, while during cold glacial times the concentration was even a third lower. The variations are believed to be related to changes in the ocean conditions, but the quantitative and mechanistic explanation is still unknown. One possibility is that, during glacial periods, the marine biota in the ocean surface layer was more active than at present, and the ocean biological pump therefore worked more effectively sequestrating CO2 from the atmosphere. The estimates of the carbon dioxide concentration prior to 650 kyr BP are based on indirect methods, and are therefore quite inaccurate. However, it seems that over the last hundreds of millions of years the carbon dioxide concentration was most of the time clearly higher than at present (Royer 2006). From time to time the concentration seems to have been even ten times higher than at present. This explains to a certain extent why the global climate has been for long periods much warmer than at present. The mechanisms and the range of the atmospheric CO2 concentration variations on a time-scale of a million years are known only roughly. It seems that both changes in volcanic activity and changes in the weathering of silicate minerals have affected carbon release and uptake. During times of high volcanic activity and/or low weathering of silicate minerals, high concentrations of CO2 have built up in the atmosphere. On the other hand, during times of low volcanic activity and/or high weathering of silicate minerals the atmospheric CO2 concentrations have decreased. The decreasing trend of the atmospheric CO2 of the past tens of millions of years has been explained by the uplift of the Himalayas (driven by the ongoing collision between the Indian subcontinent and Asia), which has increased weathering.

25

20

Carbon Dioxide (ppm)

400 360 320

15

Time kyr BP 10 5

0

Dome C (Monnin et al. 2004) Law Dome (MacFarling Meure et al. 2008) Southpole (Siegenthaler et al. 2005b) Kohnen Station (Siegenthaler et al. 2005a) Mauna Loa (Keeling & Whorf 2005) NOAA/CMDL Global Spline

280 240 200

CO 2 (ppm)

300

b

260 220

180 600

500

400

300

200

100

0

Time kyr BP

Figure 10. Atmospheric CO2 concentration during a) the last 20 kyr reconstructed from Antarctic and Greenland ice and firn data (symbols) and direct atmospheric measurements (red and magenta lines and b)the last 650 kyr reconstructed from Antarctic ice core data (EPICA community members 2004). Modified from Jansen et al. (2007) figures 6.3 and 6.4. 2.3.3

Human interference in the carbon cycle

The additional CO2 due to human activities leads to a perturbed global carbon cycle. In Figure 9 the ‘anthropogenic’ fluxes are shown in red. At present about half of the anthropogenic CO2 is taken up by the terrestrial ecosystems and the oceans (Sabine et al. 2004). However, the dissolution of the CO2 at the ocean surface slows as the surface waters equilibrate with the atmosphere, and the uptake depends on the rate of carbon transport to the deep ocean. As the oceans take up CO2, they become more acid, and in consequence CaCO3 is eventually released from deep sediments. Regarding future climate projections, the following are essential questions: how much additional carbon will human kind release into the atmosphere, and how much and how fast will the oceans and the terrestrial biosphere be able to take up anthropogenic carbon dioxide in the future? Mankind has already released about 300 Giga tonnes of carbon (Gton C) into the atmosphere and the atmospheric CO2 concentration has increased from pre-industrial values of 280 ppm to 388 ppm (Tans 2010). The potential for future release from all known conventional fossil fuel reserves is estimated to be at least 5,000 Gton C (Rogner 1997).

26

The SRES (Special Report on Emission Scenarios) scenarios are emission scenarios developed by Nakicenovic et al. (2000); these are used as the basis for the IPCC (2007) climate projections. The CO2 emissions and cumulative concentrations of the A2, A1B and B1 scenarios are illustrated in Figure 11. In the B1 scenario, the world’s economy will aim at sustainable development, and the growth of the world’s population will level off. A2 is a consumer society scenario in which energy is mainly produced by fossil fuels (leading to high emissions) and the world’s population will grow rapidly. The A1B is a scenario lying in between A2 and B1. The estimates of the atmospheric CO2 concentration reached in the year 2100 ranges in the B1 and A2 scenarios from 550 ppm to 850 ppm, respectively.

Figure 11. a) CO2 emissions and b) cumulative atmospheric CO2 concentrations of the SRES scenarios A2, A1B and B1.

When estimating the lifetime of the atmospheric CO2, a particular event from the past, the Palaeocene-Eocene Thermal Maximum (PETM, Kennett & Stott 1991, Pagani et al. 2006), has been highlighted. The PETM appears to be analogous to the potential global warming climate event in the future. Approximately 55 million years ago, the global mean temperature rose by several degrees Celsius in just 1,000 to 10,000 years, and it took at least 100,000 years for the climate to recover to its pre-PETM state. The carbon isotope concentrations in the marine and continental records show that a large amount of carbon must have been released into the atmosphere and ocean. The source of this carbon is not certain, but possible sources include methane from the decomposition of the clathrates of the sea floor (Katz et al. 1999), CO2 from volcanic activity, or oxidation of sediments rich in organic matter. The estimated magnitude of the carbon release is of the order of 1,000 to 2,000 Gton C (Dickens et al. 1997), similar to the magnitude of that projected to be released during the coming century by anthropogenic emissions (Nakicenovic et al. 2000). On a multi-millennial time-scale, it is possible that due to global warming large amounts of carbon could be released from, e.g., the ocean floor methane clathrates (Fyke & Weaver 2006).

27

Eby et al. (2009) simulated the carbon cycle on a multi-millennial time-scale with a fully-coupled Earth System Climate Model (including carbon uptake by the terrestrial biosphere, the changing ocean circulation, weathering feedback, and sediments). The model was able to reproduce the change in the atmospheric CO2 concentration during the past 200 years. In simulations of the future they found that the lifetime of the atmospheric CO2 depended on the total amount of the emissions. With emissions ranging from 1280 to 5120 Gton C, after 10,000 years 17–30 % of the emissions still remained in the atmosphere and the atmospheric CO2 concentration was in the order of 400 ppm for the 1280 Gton C emissions, and around 1000 ppm for the 5120 Gton C emission. Likewise, modelling experiments reported by Archer et al. (2009) show that with emissions ranging from 1000 to 5000 Gton C, 20–35 % of emissions will remain in the atmosphere after 200–2,000 years. On time-scales of 3,000 to 7,000 years, the remaining atmospheric fraction of the CO2 will be further drawn down by neutralization of CaCO3. 2.4

Future climate scenarios

The human induced climate change is projected with climate models. In these models the effects of changes in the forcing conditions – e.g. enhanced greenhouse gas concentrations – to the climate system are simulated. There are large uncertainties in the climate change projections, uncertainties like natural variation of climate, the uncertainties in the future emissions and model’s simplicities. When projecting the feedback effects of the ecosystems, the uncertainty is caused also by the uncertainty of human action and nature’s ability to adapt to the changing environment. Present state-of-the-art comprehensive atmospheric General Circulation Models (GCMs) coupled with modules simulating the marine, land biospheres and the sea ice are major tools for the study of past, present and future climates. The resolution of such global GCMs is usually 100–300 km (IPCC, 2007). The relatively high resolution and complex calculations make the GCMs suitable for modelling the climate on a 100– 1,000 years’ time-scale and several time-slice runs of the distant past climate have likewise been made (Otto-Bliesner et al. 2006; Kjellström et al. 2010; Strandberg et al. 2010). However, due to their high computing costs, GCMs are computationally far too demanding to simulate full glacial cycles, i.e., with a time-scale of 100,000 years. Simulations of glacial cycles started with rather simple energy-balance models coupled to simplified ice-sheet models (Pollard 1982; Deblonde et al. 1992; Gallée et al. 1991). In these experiments the simulated ice sheets were forced by variations in the Earth’s orbital parameters and these ice sheets experienced large variations at all major orbital frequencies (related to precessional angle, obliquity and eccentricity). Berger et al. (1999) showed that simulation of a glacial cycle with variations in both the orbital forcing and atmospheric CO2 concentration improved the agreement between the simulated and reconstructed glacial-interglacial variations, suggesting that the climate – carbon cycle feedback plays an important role in shaping glacial cycles. The Earth system Models of Intermediate Complexity (EMICs, Claussen et al. 2002) are models with properties lying between the simple energy-balance models and complex GCMs. EMICs incorporate substantially more physical processes than the

28

simple models, but are far less computationally demanding than coupled GCMs because of their low spatial resolution and simplified governing equations. This makes EMICs suitable for climate simulations on a time-scale of 100,000 years. 2.4.1

Future climate scenarios on a time-scale of centuries

Temperature and precipitation Projected changes of temperature and precipitation for the 21st century in Southern Finland from Jylhä et al. (2009) are illustrated in Figure 12. The projections are averages of results from 19 GCMs for the SRES scenarios A2, A1B and B1. SRES scenarios are emission scenarios developed by Nakicenovic et al. (2000) and used as a basis for the IPCC (2007) climate projections. The CO2 emissions of the A2, A1B and B1 scenarios were illustrated in Figure 11. During the ongoing century temperature and precipitation are projected to increase in Southern Finland, the increase being greatest in winter. The rate of projected climate change depends on the greenhouse gas emissions and on the climate model (Jylhä et al. 2009). Until 2040 the temperature and the precipitation increase in the Southern Finland approximately at the same rate in all scenarios. At the end of the ongoing century the differences between the scenarios grow (Figure 12). The variation of the projections of temperature and precipitation in different models’ is seen in the large range of the grey bars indicating the 90 % probability range.

5

b) 25 A2 A1B B1

4 3

A2 A1B

2

B1

1 0 2000

2020

2040

Year

2060

2080

2100

Relative change in precipitation (%)

Mean temperature change (o C)

a) 6

20

A2 A1B B1

15 10 A2 A1B

5

B1

0 2000

2020

2040

Year

2060

2080

2100

Figure 12. Projected changes of annual b) mean temperature and c) precipitation for the 21st century relative to the period 1971–2000 in southern Finland. The projections are averages of results from 19 GCMs (Jylhä et al. 2009) for the emission scenarios A2, A1B and B1. The grey bars at right indicate the best estimate (solid line within each bar) and the 90 % probability range in the ensemble of simulations for the period 2070– 2099.

29

Sea level projections The average sea level in the Finnish coastline is affected mainly by three factors: the post-glacial land uplift, changes in the total water amount of the Baltic Sea (which is altered by currents through the Denmark Strait), and the rise of global sea level. The calculated effect of the land uplift in Olkiluoto was 5.5 mm/year in the 20th century (Johansson et al. 2004). The present-day Greenland ice sheet has the potential to raise the average global sea level by 7 m. It is estimated that a rise of 3.1 ± 0.8 °C in global mean temperature could lead to the total melting of the Greenland ice sheet (Gregory & Huybrechts 2006). The West Antarctic ice sheet (WAIS) contains enough ice to increase the global sea level by 5 m. However, it has been estimated by Bamber et al. (2009) that a total collapse of the WAIS would lead to a global mean sea level rise of only 3.3 m. Over the last 750 kyr, WAIS has collapsed at least once, most likely during a particularly long interglacial about 400 kyr BP (Scherer et al. 1998). Paleoclimatic evidence (Naish et al. 2009) and simulations (Pollard & Deconto 2009) suggest that an abrupt discharge of the WAIS has occurred at global mean temperatures 1–2 °C above the present. The East Antarctic Ice Sheet could raise the average global sea level by 50 m; however, the ice sheet is, according to present-day knowledge approximately in balance, and its abrupt discharge is believed extremely unlikely. Atlantic thermohaline circulation During the last ten years it has been especially examined how the fresh water influx from the ice sheet affects the Atlantic thermohaline circulation. The weakening of the Atlantic thermohaline circulation during the current century is according to model simulations very likely. None of the models, however, show a total shutoff of the circulation in the next 100 years, although according to some simulations it might happen later in time (Mikolajewicz et al. 2007). Experts estimate that the shutdown of the Atlantic thermohaline circulation might be possible if the global mean temperature rose by 3–5 ºC above present (Lenton et al. 2008). After passing the threshold the shutdown would occur gradually in approximately 100 years. In the north Atlantic area the weakening of the Atlantic thermohaline circulation seems to slow down the warming of the climate but will not stop it (Mikolajewicz et al. 2007). 2.4.2

Future climate scenarios on a time-scale of 100,000 years

The temporal variations of the Earth’s orbital parameters are known with a high degree of certainty. Berger & Loutre (1991) have calculated the past and future changes in the orbital variations. For the next 100 kyr the eccentricity is going to be low (the Earth’s orbit around the sun will be almost a circle), which dampens the effect of precession, and the changes in the incoming solar radiation at high latitudes will be smaller than during the termination of the Eemian interglacial, see Figure 13. A similar situation of low eccentricity occurred about 400 kyr BP (Berger et al. 2003), when the interglacial may have lasted even 45 kyr (Loutre & Berger 2002). The EMIC models and model systems listed in Table 2 are able to successfully simulate the present-day climate and the most recent glacial inception, around 120–115 kyr BP

30

(see Table 2 for references). A summary of simulations into the future performed with these EMICs is represented in Table 3. According to the simulations of Loutre & Berger (2000), a glacial episode is not expected to begin within the next 50 kyr, even under a natural atmospheric CO2 regime. It is likely that sustained high atmospheric greenhouse gas concentrations will further delay the onset of the next glacial period (Loutre & Berger 2000; Archer & Ganopolski 2005).

Solar Insolation o 60 N June (Wm -2)

Time (kyr) -650 -600 -550 -500 -450 -400 -350 -300 -250 -200 -150 -100

-50

0

50

100

540 520 500 480 460 440

Elster glacial Holstein Belvédère Interglacial Interglacial

Saalian glacial Oostermeer Interglacial

Weichselian glacial Eemian Interglacial

Holocene Interglacial

Figure 13. June Solar insolation at 60 °N (green) (Berger & Loutre 1991). Variations of deuterium (įD; black), a proxy for local temperature, and the atmospheric concentrations of the greenhouse gases CO2 (red), CH4 (blue) derived from air trapped within ice cores from Antarctica and from recent atmospheric measurements (Petit et al. 1999, Indermühle et al. 2000, EPICA community members 2004, Spahni et al. 2005, Siegenthaler et al. 2005a,b). The shading indicates the last interglacial warm periods. Interglacial periods also existed prior to 450 ka, but these were apparently colder than the typical interglacials of the latest Quaternary. Modified from Jansen et al. (2007).

31

Table 2. EMIC models. Model name

Model type

LLN 2-D MPM

EMIC EMIC

CLIMBER

EMIC

CLIMBERSICOPOLIS

EMICIce sheet model EMIC

MobiDic CLIMBERGREMLINS

EMICIce sheet model

Present-day climate Simulation of a Glacial inception around 120–115 kyr BP Gallée et al. (1991) Berger et al. (1999) Petoukhov et al. Wang & Mysak (2002) (2005) Petoukhov et al. (2005) Calov et al. (2005a) Calov et al. (2005a)

Petoukhov et al. (2005) BIOCLIM (2003)

BIOCLIM (2003) BIOCLIM (2003)

32

Table 3. Simulations of the next glacial inception with different models and CO2 scenarios. Model (reference) LLN-2D 2000) ” ” ” ” ” ” ” ”

(Loutre&

CO2 Scenario Berger constant 210 ppm

Next glacial inception 1) immediate

constant 220 ppm constant 230 ppm constant 240 ppm constant 250 ppm constant 260 ppm constant 270 ppm constant 280 ppm constant 290 ppm

immediate immediate 50 kyr AP 50 kyr AP 50 kyr AP 50 kyr AP (the run ends at 50 kyr AP) >100 kyr AP

MPM (Cochelin et al. 2006) ” ” ” ” ” ”

constant 240 ppm constant 250 ppm constant 260 ppm constant 270 ppm constant 280 ppm constant 290 ppm constant 300 ppm

immediate immediate immediate immediate 50 kyr AP 50 kyr AP >100 kyr AP

CLIMBER-SICOPOLIS (Archer & Ganopolski 2005) ”

fossil fuel release 300 GtC (Archer 2005) fossil fuel release 1000 GtC (Archer 2005) fossil fuel release 5000 GtC (Archer 2005)

50 kyr AP

CLIMBER-SICOPOLIS (Ganopolski & al. unpublished) ”

constant 280 ppm

immediate

constant 400 ppm

>120 kyr AP

CLIMBER-GREMLINS (BIOCLIM 2003)

over the next 50 kyr CO2 concentration above 290 ppm, then decrease regularly towards 190 ppm at 100 kyr AP fossil fuel release 3160 GtC (Archer 2005) fossil fuel release 5160 GtC (Archer 2005)

50 kyr AP (with a cooler period 10–20 kyr AP)

over the next 50 kyr CO2 concentration above 290 ppm, then decrease regularly towards 190 ppm at 100 kyr AP fossil fuel release 3160 GtC (Archer 2005) fossil fuel release 5160 GtC (Archer 2005)

100 kyr AP (with cooler periods 10–20 kyr AP and 50–60 kyr AP)



“ “ MoBiDiC (BIOCLIM 2003)

1)

100 kyr AP >500 kyr AP

Northern Hemisphere mostly ice-free during the next 200 kyr Northern Hemisphere mostly ice-free during the next 200 kyr

Northern Hemisphere mostly ice-free during the next 200 kyr Northern Hemisphere mostly ice-free during the next 200 kyr years

here immediate onset of the next glaciation means that glaciers start to build up on the Northern Hemisphere during the next 10,000 years

33

Employing the MPM model, Cochelin et al. (2006) showed, that glacial inception could be immediate1) with pre-industrial levels of CO2 (270 ppm). Higher CO2 concentration levels (280–290 ppm) delay the glacial inception until about 50 kyr from now. This result was also supported by the simulations of Berger & Loutre (2002). Even higher CO2 concentrations (300 ppm) might push the next glacial inception beyond the next 100 kyr (Cochelin et al. 2006). In the simulations of the glacial onset by Archer & Ganopolski (2005) with the CLIMBER-2 model, the insolation value to trigger glacial onset depended strongly on the atmospheric CO2 concentration, higher CO2 concentrations requiring a deeper minimum in insolation to trigger the glaciation. According to their simulations, under a natural CO2 forcing scenario the next glaciation would start after 50 kyr. However, they noted that a slight change in the insolation trigger could easily tip the simulation into the onset of a glaciation immediately rather than in 50 kyr. Archer & Ganopolski (2005) also found that a carbon release of 300 Gton (close to antropogenic emissions up to now) could prevent glaciation for the next 50 kyr. In a simulation with a carbon release of 1,000 Gton (approximately the amount that would be released by humankind by the end of the 21st century in the SRES A2 scenario) glaciation was prevented for the next 130 kyr. With a carbon release of 5,000 Gton C (from fossil fuels or methane hydrate deposits) glaciation would not occur within 500 kyr. In the BIOCLIM-project (BIOCLIM 2004, Texier et al. 2003), most of the simulations omitting anthropogenic greenhouse gas emissions conclude that: (i) the climate is likely to experience a long-lasting (~50 kyr) interglacial; (ii) the next glacial maximum is expected to be at its most intense at around 100 kyr After Present (AP), with a likely interstadial at ~60 kyr AP; and (iii) after 100 kyr AP continental ice will melt rapidly, leading to an ice volume minimum 20 kyr later. In simulations with higher atmospheric CO2 concentration scenarios the amplitude and the timing of future climatic changes depended on the CO2 scenario and on the initial conditions related to the assumed present-day ice volume.

1)

here immediate onset of the next glaciation means that glaciers start to build up on the Northern Hemisphere during the next 10,000 years

34

35

3

MATERIALS AND METHODS FOR THE 120,000 YEAR TIME-SCALE CLIMATE SCENARIOS FOR OLKILUOTO

According to earlier model simulations, the future atmospheric CO2 concentration will have a crucial effect on the climate evolution and the onset of the next glaciation (Table 3). The more there is going to be CO2 in the atmosphere, the later the onset of the next glaciation. As there are a lot of unknowns and uncertainties about the future anthropogenic emissions and the global carbon cycle, formulating realistic future scenarios is difficult or almost impossible. Qualitatively, the simulations into the future reveal time periods potential for certain climatic states, e.g., interglacial or glacial periods. However, quantitative information about the climatic states is achieved only from the past. Therefore, although the climate is not going to repeat itself exactly, for estimating possible extreme climatic states, the last glaciation provides the best quantitative information about the evolution of the climate system during a glacial cycle. Details of the global model simulations are summarized in Chapter 3.1. The statistical downscaling methods for downscaling the coarse scale global climate data are presented in Chapter 3.2. and the actual results in Chapter 4. 3.1

Global model simulations with CLIMBER-2

The CLIMBER-2 EMIC consists of six Earth system components: atmosphere, ocean, sea ice, land surface, terrestrial vegetation and ice sheets (Petoukhov et al. 2000). The model has a low spatial resolution (see Figure 14), the latitudinal resolution being 10°. The atmospheric module is a 2.5-dimensional statistical-dynamical model with a longitudinal resolution of roughly 51°. Despite the low horizontal resolution, the atmospheric module has many features in common with more sophisticated models (GCMs). The ocean model is a 2-dimensional zonally-averaged 3-basin oceanic module (Atlantic, Indian and Pacific). The vegetation model is a 2-layer soil moisture module, VECODE (Brovkin et al. 1997). A comparison of CLIMBER-2 model results with present-day climate data by Petoukhov et al. (2000) showed that the model successfully describes the seasonal variability of a large set of characteristics of the climate system, including radiative balance, temperature, precipitation, ocean circulation and the cryosphere. Sensitivity experiments by Ganopolski et al. (2001) showed that the CLIMBER-2 model is able to simulate the climate response to changes in different types of forcing and boundary conditions (such as freshwater flux into the Northern Atlantic, atmospheric CO2 concentration, solar insolation and vegetation cover) in reasonable agreement with the results of GCMs. In an intercomparison of eight EMICs by Petoukhov et al. (2005), the CLIMBER-2 equilibrium and transient responses to a doubling of the atmospheric CO2 concentration were within the range of corresponding GCM simulations participating in the atmosphere-slab ocean model intercomparison project and the Coupled Model Intercomparison Project, phase 2 (CMIP2). The CLIMBER-2 model showed similar temperature and precipitation changes having magnitudes comparable to those found in the GCMs.

36

90

LATITUDE

60 30 ATLANTIC

0

PACIFIC

PACIFIC INDIAN

-30 -60 -90

-120

-60 0 60 LONGITUDE

120

180

Figure 14. Representation of the Earth’s geography in the CLIMBER-2 model. Dashed lines show the atmospheric grid, solid lines separate ocean basins (modified from Petoukhov et al. 2000). For simulating glacial climates, the CLIMBER-2 model has been coupled with the high resolution 3-dimensional thermomechanical ice sheet model SICOPOLIS (Greve 1997, Figure 15). The resolution of the ice sheet model is 1.5° x 0.75° and the model domain extends on the Northern Hemisphere from 21° N to 85.5° N. The ice sheet model simulates the extent and thickness, velocity, temperature, water content and age for the ice sheet. In addition, the ice sheet model calculates the isostatic displacement and the basal temperature. The climate and ice sheet components are coupled bi-directionally using a physically-based energy and mass balance interface (SEMI) as described in detail by Calov et al. (2005a). The SEMI model has the same horizontal resolution as the ice sheet model and provides the latter with ice sheets’ annual mass balance and temperature. Isostatic adjustments of the lithosphere due to changing glacial load are calculated by a local lithosphere relaxing asthenosphere model (Le Meur & Huybrechts 1996), with a relaxation time of 3,000 years. Geothermal heat flux data, for the lower boundary of the ice sheet model, is interpolated from the data by Pollack et al. (1993). In areas of missing data, the following values are applied: 42 mW m-2 for Precambrian shields, 60 mW m-2 for Paleozoic orogenic areas, and 100 mW m-2 for Mesozoic/Cenozoic orogenic areas (Lee 1970). In ice sheet modelling, the geothermal heat flux values used as input have significant effect on the simulated basal temperature, melt water and ice flow velocities (Greve & Hutter 1995, Näslund et al. 2005). The sensitivity of the ice sheet model to geothermal heat flux was reported by Greve (2005). Calov et al. (2005a) showed that the CLIMBER-2-SICOPOLIS model system is capable of simulating the glacial–interglacial variations of the physical climate system. They simulated the last glacial inception with prescribed solar insolation (Berger 1978) and atmospheric CO2 concentrations from the Vostok ice core data (Barnola et al. 1987). In the ice core data the atmospheric CO2 concentration drops rapidly during the last glacial inception from ca. 275 ppm (115 kyr BP) to ca. 235 ppm (105 kyr BP). In the simulation, the last glacial inception was a strongly non-linear process that occurred when the solar insolation in northern high latitudes fell below certain threshold. In the

37

model sensitivity studies Calov et al. (2005b) found that the used atmospheric CO2 concentration affected the ice sheet growth rate. In the simulation of the last glacial inception with a constant atmospheric CO2 concentration of 280 ppm, a typical interglacial value, the glacial inception appeared, but the areal extent of the Northern Hemisphere ice sheets was only two thirds of the ice sheet area simulated with the CO2 concentration values from the Vostok ice core data. In the simulation of the present-day Greenland ice sheet with CLIMBER-2-SICOPOLIS, the model overestimated the area of the ice sheet by 15 % and the volume by 30 % due to an underestimation of ablation (Calov et al. 2005a). POTSDAM Statistical-Dynamical Atmosphere Model T, P, SWR, LWR, WND, CLD

Land Surface Model

SEMI Surface Energy and Mass Balance Interface

iELV, SL, iFRC, iMLT

iCAL

DDST D U S T E R

Dust Model

iMBL iTSUR

SICOPOLIS

Ocean Model

RDST

Ice Sheet Model

iVOL, iMLT, iMASK, iSLD

Abbrs. CLD DDST iFRC iCal iELV iMASK iMBL iMLT iSLD iTSUR iVol LWR P RDST SL SWR T WND

Physical meaning Total cloud fraction Deposition rate of dust Fraction of land covered by ice Ice calving rate into the ocean Surface elevation above sea level Ice sheet mask Ice sheet surface mass balance Ice sheet surface melt rate Ice sheet sliding velocity Ice sheet surface annual temperature Total ice volume Downward long-wave radiation at the surface Total precipitation rate Radiative forcing of aeolian dust Sea level Downward short-wave radiation at the surface Surface air temperature Module of wind speed

Figure 15. Flow diagram of the CLIMBER-2-SICOPOLIS model system (modified from Ganopolski et al. 2010, Figure 1).

Archer & Ganopolski (2005) simulated the onset of the next glaciation in different CO2 scenarios with the CLIMBER-2-SICOPOLIS model system. Ganopolski et al. (2010) simulated the full last glacial cycle with the CLIMBER-2-SICOPOLIS model system. Their simulations were forced by variations in the Earth's orbital parameters calculated following Berger (1978, Figure 13), and atmospheric greenhouse gas concentrations derived from the Vostok ice core. This simulation is described more in detail in Chapter 4. As a continuation to the simulations of Ganopolski et al. (2010), Ganopolski et al. (unpublished) used the CLIMBER-2 SICOPOLIS model system to perform simulations 120 kyr into the future applying two CO2 concentration values (see Table 4). The first simulation employed a constant atmospheric CO2 concentration of 280 ppm, a typical interglacial value. In the second simulation the corresponding concentration was 400 ppm, a value that will be reached before 2020 due to anthropogenic emissions. The time step used in CLIMBER-2 was one day and in SICOPOLIS one year. The output data were stored every 1000 model years as monthly means of the output of CLIMBER-2 and as annual means of the output of SICOPOLIS. The simulations are described in more in detail in Chapter 4.

38

Table 4. CLIMBER-SICOPOLIS simulations used in this study. Simulation name Last glacial cycle Future 280 ppm Future 400 ppm 3.2

Simulation time from 126 kyr BP until the present from the present until 120 kyr AP from the present until 120 kyr AP

CO2 concentration Vostok ice core data constant 280 ppm constant 400 ppm

Statistical downscaling – Generalized Additive models

For our purposes, the data of the CLIMBER-2 EMIC simulations have too low spatial resolution. In climate science, several downscaling methods have been used. A common method is to downscale the results of a global model dynamically with a regional model. Regional models usually have sophisticated atmosphere and biosphere modules at a resolution of ~50 km and are therefore computationally demanding. Hence downscaling of this kind is possible only for short time-slices of 50–100 years, not for the entire 120,000 year period. A computationally less demanding method is statistical downscaling. In statistical downscaling, the statistical relationships between the observed small-scale variables (derived from observations) and larger-scale variables (e.g., from a global model) are derived using, for example, regression analysis. These derived statistical relations are then applied to downscale the large-scale variables of climate simulations to a smaller scale. Vrac et al. (2007) introduced a statistical downscaling method for paleoclimatological purposes, based on a GAM-type regression model (Generalized Additive Model, Wood 2006). They fitted a GAM-type regression model by finding the statistical relationships between the CRU-observed climate (1961–1990) and the low-resolution CLIMBER-2 EMIC model simulation of present-day climate. They then used this regression model to downscale the present-day and last glacial maximum climates simulated by CLIMBER2 over Western Europe. Martin et al. (2010a and 2010b) developed these GAMs further, using them to downscale the future climate 100 years AP. All these GAMs, however, were fitted only with the present-day climate, and were used on the assumption that the statistical relations between the large and the small scales remain unchanged for the last glacial maximum and future climate. Here we build a GAM that is able to downscale the large-scale variables of the CLIMBER-2 EMIC over Fennoscandia in very different climatic conditions. For this purpose we fit a GAM not only to the present day climate but also to the results of RCA3 regional model simulations (Kjellström et al. 2009) of the last glacial maximum climate (with an extensive ice sheet over Fennoscandia, about 21 kyr BP) and the climate during a cool stadial within Marine Isotope Stage 3, MIS 3 (with a small ice sheet over Fennoscandia, about 44 kyr BP), see Chapter 3.2.1. With the obtained GAM we are able to downscale the monthly mean temperatures and total precipitation produced by the CLIMBER-2 model simulations for the whole of the last glacial cycle and for 120 kyr into the future in the 280 ppm CO2 scenario. The obtained GAM is also used for downscaling the monthly total precipitation of the 120 kyr into the future with CO2 concentration of 400 ppm. For downscaling the temperature of the 400 ppm future scenario we use a GAM which is fitted to present-day observed climate only. In

39

Chapters 3.2.1-3.2.3 we represent data used for fitting the GAM. In Chapter 3.2.4 we summarize the fitting and evaluation of the GAMs. 3.2.1

Time-slice simulations with the regional model RCA3

Kjellström et al. (2009) used the regional climate model RCA3 (Kjellström et al. 2005; Samuelsson et al. 2010) for downscaling CCSM3 global model simulations (Collins et al. 2006; Otto-Bliesner et al. 2006; Kiehl et al. 2006) in three climatic periods: i) warm case: a possible future period in a climate warmer than today. The future case is characterised by high greenhouse gas concentrations in the atmosphere and a complete loss of the Greenland ice sheet. ii) glacial case: the Last Glacial Maximum (LGM) at 21 kyr BP, with an extensive ice sheet covering large parts of northern Europe (Strandberg et al. 2010). iii) permafrost case: a stadial within Marine Isotope Stage 3 (MIS 3) at 44 kyr BP, representing a cold period with a relatively small ice sheet covering parts of Fennoscandia (Kjellström et al. 2010). The setup of the CCSM3 and RCA3 simulations are listed in Table 5. A simulation of the recent past climate (AD 1961–1990) was used as a reference. Previous studies have shown that RCA3 is capable of simulating the climate in Europe in a realistic way (Kjellström et al. 2005; Samuelsson et al. 2010). This applies both for annual and seasonal mean conditions as well as for extremes and the daily variability of, for instance, temperature. Here RCA3 was run over Europe with a horizontal grid spacing of 0.44º (approximately 50 km). In the RCA3 model, the forcing from the greenhouse gases and aerosols are described as CO2-equivalents. As there was no information available about the past or future changes in the intensity of the solar insolation, the solar constant was set to its present day value 1365 Wm-2 in every CCSM3 and RCA3 simulation. In the simulations with the CCSM3 global model, the orbital year was calculated following Berger (1978). RCA3 does not include a routine for calculating astronomical forcing and its variation over time. Sensitivity studies by Kjellstöm et al. (2009) showed that the RCA3 had low sensitivity to the insolation variations, due to the fact that the regional model is to a very strong degree governed by the large-scale climate features of the global model. Instead, other locally stronger forcing factors, like the ice sheet extent, play a much more important role for the regional climate as simulated by RCMs. Based on this, the RCA3 simulations were all performed with astronomic conditions reflecting the recent past (AD 1990). For the warm case climate simulation, two vegetation setups were used. The first one was present-day data by Bonan et al. (2002), Bonan & Levis (2006) (RP in Table 5). The second setup used simulated vegetation from a future climate scenario representative of the year 2300 (Scholze et al. 2006, GHG in Table 5).

40

In the RCA3 glacial, permafrost and recent past climate simulations the recent past vegetation was applied in the ice-free regions as a first guess; the resulting simulated climate was then used in the dynamic vegetation model LPJ-GUESS (Smith et al. 2001) to produce new vegetation consistent with the simulated climate. Thereafter the RCA3 run was repeated with the new vegetation determined by the LPJ-GUESS.

Table 5. Forcing conditions in CCSM3 for three cases. (BP, PI and RP stand for before present, pre-industrial and recent past (AD 1961–2000), respectively). Additional sensitivity experiments undertaken in CCSM3 are marked in italics. Square brackets indicate when different forcing conditions have been used in RCA3. (ICE-5G data: Peltier 2004, CLIMBER2: Calov et al. 2005a)

Insolation Orbital year CO2 CH4 N2O Ozone Sulphate Dust, sea salt Ice sheets Land-sea distribution Sea level (compared to RP) Topography, bathymetry Vegetation

Warm 1365 Wm-2 RP 750 ppmv [841 ppmv] RP (1714 ppbv) [-] RP (311 ppbv) [-] PI [-] PI [-] PI [-] RP (excl. Greenland ice sheet) RP RP [+7 metres] RP (excl. Greenland ice sheet) RP / GHG

Glacial (LGM) 1365 Wm-2 21 ka BP [RP] 185 ppmv [168 ppmv] 350 ppbv [-] 200 ppbv [-] PI [-] PI [-] PI / PI x 3 [-] ICE-5G ICE-5G -120 m ICE-5G, RP

Permafrost 1365 Wm-2 44 ka BP [RP] 200 ppmv [187 ppmv] 420 ppbv [-] 225 ppbv [-] PI [-] PI [-] PI [-] SKB(2006), CLIMBER2, ICE-5G ICE-5G [SKB 2006], RP -120 m [-70 m] ICE-5G [SKB, 2006]

RP / LGM

RP

The seasonal cycles of temperature and precipitation at the Olkiluoto site, as simulated by the RCA3 regional climate model in four climatic situations, are depicted in Figure 16. For more details about the RCA3 simulation results, see Appendix 1.

41

Temperature (Celsius)

Precipitation (mm/month)

a) Warm

b) Recent Past

c) Permafrost

d) Glacial

Figure 16. Simulated seasonal cycles of temperature (°C) and precipitation (mm/month) at the grid box closest to the Olkiluoto site (red line). The spatial variability in the 3x3-grids is displayed with the dashed lines representing ±1 standard deviation calculated from the 9 grid boxes, and the grey area representing their absolute maximum and minimum. The green line for temperature and precipitation is the observed seasonal cycle from the CRU data set in the period 1961–1990. In the warm case, the uncertainty range is defined by ±1 standard deviation of the data calculated from the 9 surrounding grid boxes using three additional simulations for the 21st century with RCA3. Figure source: Kjellström et al. (2009).

42

Simulated Recent Past climate The forcing conditions of the recent past climate simulation were representative of the year 1990 in the global model CCSM3, as described in Collins et al. (2006) and in Kjellström et al. (2005) for the regional model RCA3. The RCA3 equivalent CO2 concentration was 333 ppm. In the recent past climate simulation (Figure 16b) the seasonal cycle in temperature in the area of Olkiluoto is weaker than the in observational data of the Climate Research Unit (green contour). The maximum precipitation at Olkiluoto is simulated to be in autumn, two months later than in the observations. Nonetheless, in some of the surrounding grid points, the maximum precipitation coincided with the observed. However, in all grid points the precipitation is overestimated during the winter half of the year by approximately 10-20 mm/month. Simulated Warm climate In the warm case simulation the global model CCSM3 was forced with atmospheric CO2 concentration of 750 ppm as a continuation of the simulation by Kiehl et al. (2006). The CO2 concentration value 750 ppm was chosen to represent a concentration, which could be reached even when not using all the ‘conventional’ fossil fuels, i.e., oil, coal and gas (Lenton et al. 2006). As a comparison, in the A1B and A2 SRES-scenarios, the CO2 concentrations are 700 ppm and 850 ppm, respectively, at the end of the 21st century. The warm case value (750 ppm) lies in between them. In RCA3, the utilized CO2 equivalent was 841 ppm, calculated after IPCC (2001). In the warm case simulation the Greenland ice sheet was assumed to have completely melted and there were no ice sheets on the Northern Hemisphere. The Antarctic ice sheet was assumed to remain identical to that in present day conditions. In CCSM3, the topography of Greenland was lowered due to ice sheet loss, otherwise the present day topography was used. In RCA3, however, isostatic adjustment in the Baltic Sea area was included allowing for a further rebound of the crust to a level corresponding to what is expected 3 kyrs into the future. In the global model CCSM3 the 7 m sea-level rise was neglected as having only marginal effects on the coarse land-sea distribution. In the regional model simulation the 7 m sea-level rise due to the melting of Greenland Ice sheet was included (Appendix 1, figure 2). In the CCSM3 simulation the global annual mean temperature was 17.2 ºC (16.9 ºC for present day vegetation), this is 2.4 ºC (2.1 ºC) higher than in the simulation of recent past climate. In the RCA3 simulations the warming was strongest in Northern Europe, where the annual mean temperature increased by 3–5 ºC compared to the recent past climate simulation. In the Olkiluoto area the annual mean temperature for the warm case (Figure 16a) was about 4 °C higher than in the recent past simulation (Figure 16b). These estimates are in consensus with the results of Jylhä et al. (2009) (depicted in Chapter 2.4.1 in Figure 12), who projected, that by the end of this century, the annual mean temperature of southern Finland will increase by 2.5–5.5 ºC, in the A1B scenario, and by 3–6.5 ºC in the A2 scenario, relative to the period 1971–2000. As in Jylhä et al. (2009), the future warming in the RCA3 warm case simulation was stronger in winter than in summer. In the warm case simulation, the westerlies over the North Atlantic strengthened, particularly in winter, leading to a winter precipitation increase in

43

Northern Europe and decrease in Southern Europe. In the RCA3 future warm case climate simulation the monthly mean total precipitation was in the area of Olkiluoto around the year higher than in the recent past simulation, on average 20 % higher. As a comparison, Jylhä et al. (2009) projected, that by the end of this century, the annual total precipitation of southern Finland will increase on average by 10 %, in the A1B scenario, and by 20 % in the A2 scenario, relative to the period 1971–2000. As in Jylhä et al. (2009) the precipitation increased in the RCA3 warm case simulation more in winter than in summer. Simulated Last Glacial Maximum climate In the simulated glacial case, the global model CCSM3 was forced with an atmospheric CO2 concentration of 185 ppm (Otto-Bliesner et al. 2006). In RCA3, the utilized CO2 equivalent was 168 ppm, calculated after IPCC (2001). The setup of the ice sheets in the glacial case simulation was based on the ICE-5G data (Peltier 2004) in both CCSM3 and RCA3. The ice sheets were treated as mountains covered with ice; hence the dynamics of the ice sheets were not simulated. In the CCSM3 simulation, present day bathymetry was used with the coastline changed according to a 120 m lowering (Lambeck 2002). In the RCA3 simulation the coastlines from ICE-5G were utilized (Appendix 1, Figure 2). In the CCSM3 glacial case simulation the global annual mean temperature was 7.9 ºC, this is 6.9 ºC lower than in the simulation of recent past climate. In the RCA3 glacial case simulation the annual mean temperatures, in the ice-free parts of Europe, were 5– 10 ºC lower compared to recent past climate simulation. Over the ice sheet the temperature was up to 40 ºC colder in winter (Appendix 1, figure 3). A large portion of this cooling is due to the increased surface elevation over the ice sheet. There was a significantly larger inter-annual variation in surface air temperature in the simulated glacial climate than in the recent past climate. In the vicinity of Olkiluoto (Figure 16d) the monthly mean temperatures were year round clearly below 0 ºC: in July around -5 ºC and in January around -40 ºC. In the global simulation with the CCSM3 model, the Atlantic storm track shifted southwards due to changes in the upper troposphere flow caused by the Rocky Mountains and the Laurentide ice sheet. In the RCA3 simulation this caused more precipitation over the Iberian Peninsula, Italy and Southern Alps. In the RCA3 simulations, the winters in Fennoscandia were dry (monthly precipitation less than 30 mm), except for the south western part of the ice sheet. During the LGM summer RCA3 produced relatively large amounts of precipitation in the vicinity of the west and south slopes of the Fennoscandian ice sheet due to forced convection (Appendix 1, Figure 4). In the area of Olkiluoto the winters were drier compared to recent past climate but summers were notably wetter due to convective precipitation caused by the heterogeneous terrain of the ice sheet. Simulated Permafrost climate In the permafrost case simulation the global model CCSM3 was forced with atmospheric CO2 concentration of 200 ppm (Kjellström et al. 2010). In RCA3, the utilized CO2 equivalent was 187 ppm, calculated after IPCC (2001). The Fennoscandian ice sheet data of the SKB (2006) project was used. The Laurentide ice sheet extent was from a simulation with CLIMBER-2 (Calov et al. 2005a) with an adjustment being a complete removal of the ice sheet in Alaska (Näslund et al. 2008). ICE-5G data by

44

Peltier (2004) was used for the extent of the Antarctic ice sheet. In the global model simulation the topography was a combination taken from the ice sheets and LGM topography by a lowering of sea-level by 120 m. The 120 m sea level lowering is an overestimation for MIS 3 as it is estimated to have been only ca. 70 m (Lambeck 2002, Siddall et al. 2008). At the coarse scale of global model CCSM3, this difference had only modest impact on the position of the coastline. In the regional model these differences had significant impact and therefore a sea level -70 m was used in the RCA3 simulation. In the permafrost case simulation the global mean temperature was about 9.2 ºC, i.e., about 5.4 ºC lower than in the recent past climate simulation but about 1.3 ºC higher than in the glacial case simulation. In the vicinity of Olkiluoto (Figure 16c) the annual mean temperature was about -6 °C. The seasonal cycle of temperature was stronger than in the recent past climate simulation with monthly mean temperature in July around 10 ºC and in January around -25 ºC. A noteworthy change in the large-scale atmospheric circulation was the amplification of the topographic wave over the Rocky Mountains and Laurentide ice sheet compared to the recent past climate simulation. However the amplification was only half compared to the glacial case. This led to a southward shift of the Atlantic storm track and decreased precipitation in Northern Fennoscandia and increased precipitation in Southern Fennoscandia compared to the recent past climate simulation (Appendix 1, figure 6). In the area of Olkiluoto the precipitation was lower for most of the year compared to recent past climate simulation. According to the model, during summer, the length of the completely snow-free season was three months. Likewise, during winter, the length of a season with a more or less constant snow cover was at least three months. This kind of cold, dry and partially snow-free conditions are favourable for permafrost growth. For our downscaling with the GAM-type regression model; we used monthly mean temperature and precipitation values of the RCA3 simulations described above. 3.2.2

Climate Research Unit observational data

For our present-day climate, we used the Climate Research Unit (CRU) high-resolution climate data, version 2.1. (Mitchell & Jones 2005). For land areas the original resolution of the monthly mean surface temperature and total precipitation data was 0.5°x0.5°. To cover areas where CRU high-resolution data were not available, i.e., sea areas, we used the Jones et al. (1999) 5°x5° temperature data. Similarly, where high-resolution precipitation data were not available, we used the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP, Xie & Arkin, 1997) derived from the years 1979 to 2000, with the original data resolution of 2.5°x2.5°. 3.2.3

NOAA terrain data

Elevation data were downloaded from the webpage of the National Geophysical Data Center, (http://www.ngdc.noaa.gov/mgg/gdas/gd_designagrid.html) NOAA Satellite and Information Service, and interpolated bi-linearly to a 1.5°x0.75° resolution for the Fennoscandian area (59° N… 70° N and 3° E … 35° E).

45

3.2.4

Fitting the statistical model

Our statistical downscaling consists in using statistical relationships between the CLIMBER global model’s output (Y, the predictand) and the regional-scale climate variable for the corresponding time period (variables X1,...,Xp, the predictors). These statistical relationships need to be calibrated upon an observed or modelled climate. For the fitting of the GAM all the data must be represented at the same spatial resolution. We use a resolution of 1.5°x0.75° over the Fennoscandian area (59° N… 70° N and 3° E … 35° E) as illustrated in Figure 17. The CLIMBER-2, the RCA3 and the CRU data were interpolated bi-linearly onto this resolution, in which the SICOPOLIS data were already given. The statistical relationships are calibrated by stepwise screening of the data, using the data for one month at a time. For present-day climate predictors we use CRU observed monthly mean temperature and precipitation data (see Chapter 3.2.2) and for present day topography we utilize NOAA terrain data (see Chapter 3.2.3). For climate and topography predictors of 44 kyr BP and 21 kyr BP we utilize temperature, precipitation and elevation data of the RCA3 model simulations described in Kjellström et al. (2009). We calculated the direction and the angle of the steepest slope from the elevation data sets for each time period.

Figure 17. The downscaling area and resolution for the GAM. The land grid cell nearest to Olkiluoto is marked green. The statistical model GAM is a non-linear multi-regression method. We model the statistical expectation (E) of the predictand (Y) using a sum of univariate smooth functions of the p predictors (X1,...,Xp), such that E (Y | X1} Xp)

p

¦ f (X )  H j

j

j 1

where the potentially non-linear smooth functions fj have a non-parametric form. It is assumed in this model, that the residuals İ follow a normal distribution. Our aim is to downscale the mean temperature and the total precipitation for each month. Therefore, the predictand Y will be either the monthly mean surface air temperature or the logtransformed monthly mean total precipitation of CLIMBER (shown for temperature in Figure 18). The total precipitation is better modelled by log-normal than by normal

46

distribution (Cheng & Qi 2002). In the fitting of the GAM, we have on the predictor side (the right-hand side of the equation) information for each grid point on the elevation, the shortest distance to the nearest glacier, the latitude and the longitude of each grid point, and the direction and angle of the steepest slope as shown for temperature in Figure 18. The GAM is fitted to the three different climate periods at the same time. We assume that if the fitted GAM is valid in all these three different climate situations (interglacial climate (present-day), glacial climate (21 kyr BP) and permafrost climate (44 kyr BP), then this GAM can be used for downscaling the CLIMBER simulations throughout the full glacial cycle, i.e., the simulations of the last glacial cycle and the future 280 ppm simulation. The climate of the future 400 ppm simulation is crucially different from the 280 ppm simulation, the former being mostly as warm as today or even warmer. Therefore, for the temperature of the 400 ppm scenario, we use a GAM which was fitted by finding the statistical relations of the CLIMBER present-day and CRU observational data only. In Appendix 2 we show that the fitted GAMs are able to explain more than 94 % (45 %) of the surface air temperature (precipitation) spatial variance (see Appendix 2 Table A2.1 (Table A2.2)), and to reproduce approximately the observed and RCA3-simulated monthly mean temperatures and total precipitation (see Appendix 2, Figures A2.1 – A2.6). In Figure 19 we depict the utilisation of the GAM for temperature of the Last Glacial Cycle simulation of CLIMBER. As input we have the bi-linearly interpolated CLIMBER temperature data and elevation and ice sheet data of the SICOPOLIS model. The statistical relationships achieved from fitting of the GAM are now used to produce downscaled temperature and log-precipitation of the last glacial cycle CLIMBER data. As output we get downscaled temperature data for the last glacial cycle. We compared the direct coarse output of the CLIMBER model (Last Glacial Cycle run) and the downscaled GAM output to Holocene proxy data. The comparison shows that the downscaling with the GAM improves the course output of CLIMBER model simulations of annual mean temperature (see Appendix 2 Figures A2.8 – A2.9.) In the next chapters, we use these fitted GAMs to downscale the whole CLIMBER-2 simulation of the last glacial cycle (Chapter 4) and four simulations 100–120 kyr into the future (Chapter 5).

47

Figure 18. Flow diagram of the fitting of a GAM for downscaling monthly mean surface temperature by finding statistical relationships between the global model output (predictors) and the regional-scale climate variables (predictands). Abbreviations are explained in Table 6.

Figure 19. Flow diagram of the utilization of the GAM for downscaling large scale global model output into regional scale. Table 6. Abbreviations of data fluxes shown in Figure 18 and Figure 19. Abbreviations T el st ic CRU NG RC CLI SIC GAM lat lon

Surface Air temperature elevation of the surface (land, ice sheet or sea) direction of the steepest slope distance to the ice sheet margin (or nearest ice cap) data of the Climate Research Unit data of the National Geophysical Data Center model output or setup of the RCA3 model model output of the CLIMBER model model output of the SICOPOLIS model statistical model latitude longitude

48

49

4

LAST GLACIAL CYCLE SIMULATIONS

Ganopolski et al. (2010) simulated the last glacial cycle with the CLIMBER-2SICOPOLIS model system discussed in Chapter 3. The simulation was forced by variations in the Earth's orbital parameters calculated following Berger (1978, see Figure 20a), and by atmospheric greenhouse gas concentrations (CO2, CH4 and N2O) derived from the Vostok ice core. Since the radiative scheme of the CLIMBER-2 model includes only CO2, the radiative effect of the CH4 and N2O was incorporated via equivalent CO2 concentration. The equivalent CO2 concentration was determined as the CO2 concentration which had the same radiative forcing as the combined radiative forcing of the greenhouse gases CO2, CH4 and N2O. The additional (compared to present day) radiative forcing of the atmospheric dust Rd was parameterized by the relation: Rd = RdLGM V/VLGM (Schneider et al. 2006). RdLGM is the radiative forcing of dust at the LGM. V is the simulated Northern Hemisphere ice volume and VLGM is an approximate estimate for the Northern Hemisphere ice volume during the LGM (VLGM = 100 metres of sea-level equivalent). 4.1

Simulated ice sheet and climate evolution in the baseline experiment

The CLIMBER-2-SICOPOLIS model system contains numerous so-called “tunable” parameters. Ganopolski et al. (2010) performed about a hundred model simulations with different combinations of the tunable model parameters. Their findings about the sensitivity of the model system on the different parameter settings are summarized in Chapter 4.2. Of all the simulations, Ganopolski et al. (2010) selected a so-called Baseline Experiment which was the simulation with the best fit compared to paleodata: Northern Hemisphere ice sheet volume reconstruction during the last glacial cycle and ice sheet extent during the LGM. The parameter settings of the Baseline Experiment were also used in the future simulations A (Chapter 5). The major features of this CLIMBER-2-SICOPOLIS last glacial cycle simulation are depicted in Figure 20, Figure 21 and Figure 22. The global sea level change (Figure 20e) is calculated from the Northern Hemisphere ice volume change; it does not take into account the ice volume changes of the Southern Hemisphere, sea level variations due to changes in the gravitational field of the Earth, isostatic displacement due to ice sheet advance or retreat, or oceanic thermal expansion. Compared to proxy data the CLIMBER-2-SICOPOLIS model system is able to reproduce all the major ice volume and temperature changes of the last glacial cycle (amplitude and timing). The last glacial cycle simulation starts from the Eemian interglacial conditions at 126 kyr BP, with a climate warmer than at present (Figure 20b) in both the Northern Hemisphere and Fennoscandia, and with very little ice in Northern Europe except for some ice caps on the Scandinavian mountains. In the Northern Hemisphere, the total volume of the ice sheets is lower than at present, raising the mean global sea level on average by 4 m.

540 500 460

b

Global Fennoscandia

c

Global Fennoscandia

15 10 5 0 -5 -10 -15 -20 -25 -30

Temperature (Celsius)

a

0 5 10 15 20 25 30 35 40 45 50

Ice Volume (106 km3)

420

1000 800 600 400 200

Global Sea Level (m) relative to present day

Precipitation (mm/year)

June Solar Insolation at 65o N (Wm-2 )

50

d 0 -20 -40 -60 -80

Northern Hemisphere Greenland North America North Eurasia Fennoscandia (a part of North Eurasia)

e

130

110

90

Eemian

Early Weichselian

5e

5d 5c 5b 5a

70

50

30

Middle Weichselian

4

3

10

0 Time (kyr) Before Present

Late HoloWeichs. cene Geochronology

2

1

Marine isotope stages and substages (MIS)

Figure 20. CLIMBER-2-SICOPOLIS simulation of the last glacial cycle: a) June solar insolation at 65° N, b) global and Fennoscandian annual mean temperature, c) global and Fennoscandian annual mean total precipitation, d) Northern Hemisphere ice volume evolution and e) Northern Hemisphere ice volume expressed in units of sea level change.

51

a) About 90 kyr BP

b) About 60 kyr BP

?

c) LGM, about 20 kyr BP Figure 21. Simulated and reconstructed ice sheet extension over Northern Eurasia a) about 90 kyr BP, b) about 60 kyr BP and c) during the LGM about 20 kyr BP. The magenta contours follow the reconstructions by Svendsen et al. (2004). The yellowbrown areas represent land, the blue areas sea, and the white areas the ice sheet (thickness in metres) as simulated by the SICOPOLIS model in the last glacial cycle simulation by Ganopolski et al. (2010).

Temperature (Celsius)

52

10 5

a)

observed present day climate

0

Olkiluoto grid point grid south of Olkiluoto grid north of Olkiluoto

-5 -10 -15 -20 -25

Ice sheet height (m) Bedrock height (m)

Precipitation (mm/year)

-30

b) 700 600 500 400 300 200 100 0

2000 1500 1000 500 0 -500 -1000

c)

130

110

90

Eemian

Early Weichselian

5e

5d 5c 5b 5a

70

30

50 Middle Weichselian

4

3

10

0 Time (kyr before present)

Late HoloWeichs. cene Geochronology

2

1

Marine isotope stages and substages (MIS)

Figure 22. a) Annual mean surface temperature around Olkiluoto. The contours are presented for times with no ice sheet present and are downscaled with the GAM from the CLIMBER-2 last glacial cycle simulation. The temperatures marked with triangles represent periods with an ice sheet present at the corresponding altitude; the values are obtained from the SEMI model calculations. b) The GAM downscaled annual mean total precipitation around Olkiluoto (downscaled with the GAM from the CLIMBER-2 last glacial cycle simulation), c) ice sheet height (contour) and bedrock level (+) around Olkiluoto in the last glacial cycle CLIMBER-2-SICOPOLIS simulation.

The simulated temperature changes (Figure 20b) follow the insolation changes (Figure 20a). In Fennoscandia, the temperature changes are locally greater than those globally. The simulated annual mean temperatures decrease at the end of the Eemian and during the early Weichselian MIS 5d stadial. The Fennoscandian ice sheet starts to develop in Northern Europe from the Scandinavian mountains, being at its maximum during the MIS 5d stadial at about 112 kyr BP. The ice sheet formation leads to a 40 m drop in the

53

average global sea level. The climate warms during the MIS 5c interstadial and the ice sheet retreats, being at its smallest at about 100 kyr BP. This leads to a 40 m sea level rise. During the MIS 5b stadial the global climate cools again, and the Fennoscandian ice sheet grows, being largest at about 90 kyr BP. In Figure 21a the ice sheet maximum extent during MIS 5b is illustrated in the CLIMBER-2-SICOPOLIS simulation and compared with the paleoclimatic reconstruction (Svendsen et al. 2004 and references therein) in Northern Eurasia. In this reconstruction, the Fennoscandian ice sheet extends more to the northeast than in the simulation. In the simulation the ice sheet extends to Central and Southern Finland, but in the reconstruction it does not. In the simulation, the position of the ice sheet margin is near Olkiluoto, while in the grid cell 80 km south of Olkiluoto, there is no longer an ice sheet. This is crucial to the temperature and precipitation evolution, which are depicted in Figure 22. In the Olkiluoto grid cell with an ice sheet of 1 km thick, the mean annual air temperature on the ice sheet is between 10 and -15 °C. In the ice-free grid cell south of Olkiluoto, periglacial conditions prevail during the MIS 5b stadial with a mean annual air temperature between -5 and 0 °C, for several thousand years. In the simulations of Siegert et al. (2001), the ice sheet likewise reaches parts of Southern Finland during MIS 5b. During the next interstadial, MIS 5a, the climate warms up again, and the Fennoscandian ice sheet retreats, reaching a minimum at about 80 kyr BP. After MIS 5a, during the MIS 4 stadial, the global climate starts to cool again and the ice sheets grow, both in the paleoclimatological reconstructions and in the CLIMBER-2 simulation. At Olkiluoto the ice sheet reaches a thickness of about 2.5 km, with ice sheet surface annual temperatures lying between -15 and -30 °C (Figure 22). The ice sheet is at its largest in Fennoscandia at about 60 kyr BP, as illustrated in Figure 21b. The simulation shows less ice over Fennoscandia compared to that in the reconstruction and again, the model does not simulate the ice sheet on the Barents Sea that is present in the reconstruction. After MIS 4, at the beginning of MIS 3, the global climate starts to warm up again, and large parts of Fennoscandia become ice-free for several thousand years in both the simulation and the reconstructions. It has been suggested by Ukkonen et al. (1999) and Lunkka et al. (2001) that this ice-free period could even have lasted in southern Fennoscandia until 25 kyr BP. In the simulation, however, after 50 kyr BP the Fennoscandian ice sheet already starts to grow again and covers most of Fennoscandia by about 40 kyr BP, leaving only small parts of Southeast Finland and Southern Sweden ice-free, until these areas too become covered by ice at 28 kyr BP (Southeast Finland) and 20 kyr BP (Southern Sweden). In both, the reconstructions and simulations, the Fennoscandian ice sheet reaches its absolute maximum during the last glacial cycle at about 20 kyr BP. In Figure 21c the Last Glacial Maximum is illustrated; the simulated spatial extent of the ice sheet is in reasonable agreement with the paleoclimate reconstructions with some underestimation of the extent of the southern margin of the ice sheet. In the CLIMBER-SICOPOLIS simulation, the ice sheet thickness at Olkiluoto reaches 2.5 km, which is well in the range of the ICE-5G model reconstruction (Peltier 2004) in which the ice sheet height in

54

the Finnish west coast during the LGM was about 2 km above present day sea level. In the CLIMBER-SICOPOLIS simulation, the area of Olkiluoto is depressed due to glacial loading to a maximum of about 615 metres, which likewise agrees well with the estimated depression of 600 m by Eronen et al. (1995) based on sea level displacement curve fitted to observations in the Olkiluoto – Lake Säkylä Pyhäjärvi area. In the simulation, the global climate starts to warm after the last glacial maximum, and the Fennoscandian ice sheet retreats. The whole of Fennoscandia is almost totally icefree by about 7 kyr BP. This is in reasonable agreement with palaeoclimate reconstructions. During the Holocene climate optimum of 6 kyr BP, the simulated climate is warmer and the ice caps are smaller than at present. In summary, in the simulations with the CLIMBER-2-SICOPOLIS model system Olkiluoto experienced 4-5 glaciations during the last glacial cycle. The glaciations took place during the cold MIS 5d, MIS 5b, MIS 4 stadials and between late MIS 3 and MIS 2 (Figure 22). During the warmer MIS 5c, MIS 5a interstadials and MIS 3 the ice retreated from the Olkiluoto area. The glaciation ended when the current interglacial started about 11,500 years ago. 4.2

Sensitivity studies as a basis for the model parameter settings

Many parameterisations of the surface energy and mass balance interface (SEMI) seemed to have a considerable effect on the simulated glacial cycle (figure 11b in Ganopolski et al. 2010). In the experiment where the so-called “elevation-desert effect” on precipitation (Eq. 7 in Calov et al. 2005a) was disabled, i.e., the precipitation was simply interpolated from the coarse resolution CLIMBER grid, the Northern Hemisphere ice volume grew even 40 % higher than in the Baseline Experiment. Likewise, the parameterisation of the slope effect on precipitation (Eq. 6 in Calov et al. 2005a) had considerable effect on the ice volume development. Switching off the slope effect decreased the precipitation amounts over the marginal parts of the ice sheets resulting in smaller ice sheets. In the Baseline Experiment a sub-grid correction for the North American surface temperature was implemented. Due to the coarse resolution of the atmospheric grid in CLIMBER-2, North America is represented in one grid point. This hinders the representation of North America’s zonal temperature gradient. With a subgrid temperature correction a zonal temperature gradient was implemented in the North American continent. The magnitude of the temperature correction was 3 ºC. Disabling the temperature correction had a considerable reducing effect on the simulated ice sheet evolution. The enhancement factor of the ice sheet model, which affects the deformation rate of the ice, showed a rather weak sensitivity in the sensitivity experiments (figure 11a in Ganopolski et al. 2010). Likewise, doubling of the bedrock relaxation time-scale seemed to have a small effect on the Northern Hemisphere ice sheet development. The bottom sliding parameter of the ice sheet model, which determines the bottom sliding over the areas where the ice sheet base is temperate, seemed to have a rather strong sensitivity. In an experiment where all continental grid points were treated as rocks, the basal sliding was low resulting in thicker ice sheets. In an experiment, where all

55

continental grid points were treated as terrestrial sediments, the ice sheets were more mobile and thinner. When switching off the effect of glaciogenic dust, the simulated ice sheet volume was larger during the whole glacial cycle. 4.3

Scenario with ice-free conditions at Olkiluoto during the MIS 3

Radiocarbon dates from mammoths indicate that during the MIS 3, large areas of Southern and Central Finland were ice-free for at least 10 kyr (Ukkonen et al. 1999). Arppe & Karhu (2006) estimated from the isotopic composition of mammoth skeletal remains that the mean annual temperatures during this period were 2–6 °C lower compared to present-day values (present-day values at Olkiluoto around +4 °C). In the CLIMBER-SICOPOLIS last glacial cycle simulation the Fennoscandian ice sheet advanced already between 50 kyr BP and 40 kyr BP covering most of Fennoscandia, leaving only small parts of Southeast Finland and Southern Sweden ice-free, until these areas too became covered by ice at 28 kyr BP (Southeast Finland) and 20 kyr BP (Southern Sweden). We investigated how the GAM-downscaled temperature and precipitation for Olkiluoto would differ, if the ice sheet would not cover Olkiluoto between 55 kyr BP and 30 kyr BP. We used the GAMs to downscale surface temperature and precipitation from the CLIMBER temperature and precipitation for Olkiluoto assuming the Fennoscandian ice sheet was small during the time period from 55 kyr BP until 30 kyr BP. The downscaling was otherwise the same as in Chapter 4.1 except for the Fennoscandian ice sheet, which was kept constant and had the same size as in the RCA3 permafrost case simulation (Appendix 1, Figure A1.1, right) and the topography was likewise as in the RCA3 permafrost case simulation (Appendix 1, Figure A1.2, right). The results are shown in Figure 23. As comparison we depict also the downscaled temperature for Olkiluoto with the ice sheet simulated by CLIMBER-SICOPOLIS (same as in Figure 22 green contour). The GAM-downscaled annual average temperature in the Olkiluoto area is between -4 and -10 °C in the ice free period, which is lower than the estimate of Arppe & Karhu (2006) but well in the range of the RCA3 permafrost case simulation (annual mean temperature circa -7 °C). During this kind of ice free and cold time period, conditions would be favourable for widespread permafrost development.

Precipitation (mm/year)

Temperature (Celsius)

56

10 5 0 -5 -10 -15 -20 -25 -30

Olkiluoto with ice sheet Olkiluoto no ice sheet during MIS 3

a)

600 500 400 300 200 100 0

b)

130 110 90 70 50 30 10 0 Time kyr Be fore Pre se nt

Figure 23. Simulated a) annual mean surface temperature: at 2m height (line), on the ice sheet on the corresponding height (triangle) and b) total precipitation at Olkiluoto with the ice sheet simulated by SICOPOLIS model (green) and ice-free for the period 55 kyr BP until 30 kyr BP (magenta). 4.4

Simulated ice sheet retreat

In the CLIMBER-SICOPOLIS simulation of the Last Glacial Cycle, the ice sheet retreated from Olkiluoto between 13 kyr BP and 12 kyr BP as depicted in Figure 24 and Figure 25. The average retreat speed defined from the ice sheet data was 200 m/year northward and the angle of the 1000 m ice sheet thickness isoline was 0.25º (N)/ 1º (E). During the ice sheet retreat the grid cell nearest to Olkiluoto remained below global sea level due to glacial loading.

57

Figure 24. Ice sheet thickness and angle of the 1000 m thickness isoline at 13 kyr BP (right) and at 12 kyr BP (left) in the CLIMBER-SICOPOLIS simulation. The numbers are the ice sheet thickness of each grid point and the 1000 m, 1500 m, 2000 m and 2500 m isolines are interpolated from the ice sheet data.

Figure 25. The ice sheet height (violet), basal height (yellow) and the global mean sea level (blue) during ice sheet retreat at the 22.5º E longitude in the CLIMBERSICOPOLIS simulation.

58

59

5

SIMULATIONS OF THE FUTURE USING CLIMBER-2

As a continuation to the simulations of Ganopolski et al. (2010), Ganopolski et al. (unpublished) made simulations with the CLIMBER-2-SICOPOLIS model system into the future. These simulations were made with two different model setups: Simulation A and Simulation B. In Simulation A, the tunable parameters of the model system were exactly the same as in the simulation of the last glacial cycle (results of which were presented in Chapter 4). Simulation A was run for 120 kyr into the future applying two CO2 concentration scenarios: constant atmospheric CO2 concentrations of 280 ppm and 400 ppm. In Simulation B there were small differences in the parameter settings in the mass balance scheme (SEMI model) compared to the Simulation A. Simulation B was run 100 kyr into the future applying the same two CO2 concentration scenarios as for simulation A, i.e., constant atmospheric CO2 concentrations of 280 ppm and 400 ppm. The solar insolation changes at 65° N during the next 120 kyr are depicted in Figure 26a. The global and Fennoscandian annual mean temperatures, total precipitation and Northern Hemisphere ice volume of the CLIMBER-2-SICOPOLIS model simulations into the future with constant CO2 concentrations of 280 ppm and 400 ppm are given in Figure 26b-d. The sea level change (Figure 26e) is calculated from the Northern Hemisphere ice volume change; it does not take into account potential sea level changes occurring due to ice volume changes of the Southern Hemisphere, sea level variations due to changes in the gravitational field of the Earth, isostatic displacement due to ice sheet advance or retreat, or oceanic thermal expansion. In the future scenario with a constant 280 ppm atmospheric CO2 concentration, the global mean sea level height will be lowered by the Northern Hemisphere ice sheet development by about 20 m during the stadials 60 kyr AP and 100 kyr AP. In the future scenario of 400 ppm CO2 concentration, the sea level drops by 1–2 m during the solar insolation minima around 60 kyr AP and 100 kyr AP due to modest ice sheet development on the Northern Hemisphere. The differences between simulations A and B distinctly reveal the sensitivity of the model system to its parameter settings, in this case the mass balance scheme.

June Solar Insolation at 65 oN (Wm -2)

60

540 500 460 420

a 20

15

40

60

80

100

120

Figures b and c

Temperature (Celsius)

10

Global 400ppm_A Global 400ppm_B

5 0

Global 280ppm_A Global 280ppm_B

-5

b Fennoscandia 400ppm_A Fennoscandia 400ppm_B Fennoscandia 280ppm_A Fennoscandia 280ppm_B

Precipitation (mm/year)

1000 800 600 400 200

c

Global Sea Level (m) relative to present day

NH Ice Volume (10 6 km 3 )

0

Figures d and e 400ppm_A 400ppm_B 280ppm_A 280ppm_B

5 10 15

d 5 0 -5 -10 -15 -20

e

0

20

40

60

80

100

120

Time (kyr) After Present

Figure 26. CLIMBER-2-SICOPOLIS future simulations for the next 130 kyr in two atmospheric CO2 concentration scenarios: a) June solar insolation at 65° N, b) global and Fennoscandian annual mean temperature, c) global and Fennoscandian annual mean total precipitation, d) Northern Hemisphere ice volume evolution and e) Northern Hemisphere ice volume expressed in units of sea level change.“A” stands for the Simulation A and “B” for Simulation B.

61

5.1

Simulation into the future with a constant CO2 concentration of 280 ppm

In the CLIMBER-2-SICOPOLIS simulation 120 kyr into the future with a constant CO2 concentration of 280 ppm (a typical natural interglacial value for CO2 concentration) Northern Eurasia cools during the next 10 kyr, as is seen in the annual mean temperature of Fennoscandia in Figure 26b. As the climate cools the precipitation decreases (Figure 26c). The cooling of the climate is not very evident in the global mean temperature. This cooling, however, leads to ice sheet growth in both Fennoscandia and North America, which can be seen in the ice volume curves in Figure 26d. In Northern Europe the ice sheet development starts from the Scandinavian mountains and spreads from there, covering large areas of Fennoscandia by 20 kyr AP. The local changes of mean temperature, precipitation, ice sheet, and basal height and temperature at Olkiluoto are depicted in Figure 27. During the first stadial around 20 kyr AP, the grid point for Olkiluoto is close to the ice sheet margin. In simulation A, the ice sheet does not quite reach Olkiluoto during the first stadial. Nonetheless, the ice sheet reaches the grid point 80 km north of Olkiluoto. In simulation B the ice sheet reaches Olkiluoto 13 kyr AP (Figure 27), and the ice sheet grows to become 1.2 km thick during the period 13–17 kyr AP. At the grid point 80 km south of Olkiluoto there is no ice sheet (in simulation B), and the annual mean temperature is there slightly below 0 degrees Celsius. About 20 kyr AP, the climate in Fennoscandia starts to warm again due to increasing solar insolation, and the ice sheet starts to retreat. In simulation B the Olkiluoto grid point is free from ice at 27 kyr AP; the ice-free period lasts from 27 to 53 kyr AP with annual mean temperatures near present-day values (+4 °C), but with slightly higher summer temperatures. This ice-free period in Fennoscandia with mean temperatures comparable with those of the present-day climate lasts from about 32 kyr AP until 47 kyr AP. After 43 kyr AP the solar insolation in the Northern Hemisphere decreases again, and the climate in Fennoscandia becomes cooler. The Fennoscandian ice sheet starts to grow, reaching a maximum extent at about 60 kyr AP and covering most of Fennoscandia at that time. During this second glaciation, too, the ice sheet margin is near Olkiluoto. In simulation A the ice sheet reaches Olkiluoto at about 50 kyr AP (simulation B: 54 kyr AP) and reaches a maximum thickness of 1.8 km (simulation B: 1 km). The solar insolation increases again between 55 kyr AP and 67 kyr AP leading to ice sheet retreat. In simulation A the ice sheet retreats from Olkiluoto at 60 kyr AP (simulation B: 61 kyr AP). Fennoscandia becomes ice-free at about 67 kyr AP, i.e., around the time of the summer insolation maximum in the Northern Hemisphere. An ice-free period with annual mean temperatures comparable to or slightly warmer than at present lasts from about 67 kyr AP to 82 kyr AP. At Olkiluoto the climate is slightly warmer than at present, especially during summers. After the following insolation maximum at 88 kyr AP, the climate of Fennoscandia cools and the Fennoscandian ice sheet starts to advance; by 100 kyr AP the ice sheet

62

covers most of Fennoscandia, being at its thickest, about 3 km, above the Bothnian Bay. The ice sheet reaches Olkiluoto at 90–95 kyr AP. Around 100 kyr AP the thickness of the ice sheet there is 1.5 km. During this stadial, the grid point south of Olkiluoto is also covered with ice. In simulation A the ice retreats from Olkiluoto at about 107 kyr AP and the climate warms up to interglacial conditions. Hence, in the simulation of the future with a CO2 concentration of 280 ppm using CLIMBER-2-SICOPOLIS, the Olkiluoto grid point experiences three stadials during the next 120 kyr. It is important to note that during the two first stadials Olkiluoto might be ice-free or near the margin of the ice sheet for several thousand years, even for as long as 10 kyr.

Figure 27. a) June solar insolation at 65° N, b) Annual surface temperature around Olkiluoto. The contours are drawn for times with no ice sheet present. The values are downscaled with the GAM from the CLIMBER-2 future simulation with a 280 ppm CO2 concentration. The temperatures marked with triangles represent times with an ice sheet present, the values being inferred from the SEMI model, c) The GAM-downscaled mean annual total precipitation around Olkiluoto (downscaled from the CLIMBER-2 future scenario with a 280 ppm CO2 concentration), d) ice sheet thicknesses around Olkiluoto in the CLIMBER-2-SICOPOLIS simulation.

63

5.2

Simulation into the future with a constant CO2 concentration of 400 ppm

In the simulation of the future with a constant high CO2 concentration of 400 ppm, the global climate is remarkably warmer than in the 280 ppm simulation (Figure 26b) and the ice sheet development is modest (Figure 26d). During the insolation minimum at about 17 kyr AP, which in the 280 ppm scenario caused a large ice sheet over Fennoscandia, in the 400 ppm simulations there is only some ice on the Scandinavian mountains. The next insolation minimum at about 54 kyr AP produces a small ice sheet from the Scandinavian mountains to Lapland and parts of Northern Ostrobothnia. This ice sheet retreats and Fennoscandia is almost ice-free in a climate slightly warmer than the present-day observed climate for the time period 65 kyr AP to 95 kyr AP. After 95 kyr AP a minor ice sheet starts to develop over the Scandinavian mountains, reaching Lapland around 100 kyr AP. During the next 120 kyr the ice sheets do not reach Olkiluoto in the 400 ppm scenario. The mean temperatures at Olkiluoto are most of the time warmer than at present (+4 °C) (Figure 28). An exception is the insolation minimum at about 50 kyr AP, where the mean temperatures are about the same as at present (in the 280 ppm scenario, an extensive ice sheet development took place during this time). The monthly and annual total precipitation values are higher than in the present climate, but in general the variations in the annual precipitation rates during the next 120 kyr are small.

64

Figure 28. a) June solar insolation at 65° N, b) annual surface temperature around Olkiluoto downscaled with the GAM from the CLIMBER-2 future scenario with a 400 ppm CO2 concentration. c) The GAM downscaled mean annual total precipitation around Olkiluoto (downscaled from the CLIMBER-2 future scenario with a 400 ppm CO2 concentration).

65

6

DISCUSSION AND CONCLUSIONS

The aim of the present work is to formulate plausible future climate states at Olkiluoto, Finland on a regional scale, and especially to explore the possibility of extreme climatic conditions. For this purpose we used simulations performed with an EMIC model, CLIMBER-2 (Petoukhov 2000; Ganopolski 2001), coupled to the SICOPOLIS ice sheet model (Greve 1997; Calov 2005a). Three scenarios were explored in detail. The coldest of the scenarios was a simulation of the last glacial cycle, extending from 126 kyr BP until today, with the radiative forcing of atmospheric CO2, CH4 and N2O concentrations derived from the Vostok ice core data (Ganopolski et al. 2010). The two other scenarios were simulations for the future. The colder one was a simulation of 120 kyr into the future with a constant CO2 concentration of 280 ppm. In the milder simulation the corresponding CO2 concentration was 400 ppm (Potsdam Institute for Climate Impact Research, Ganopolski et al. unpublished). The large-scale output of the CLIMBER-2 simulations was downscaled with a GAM-type regression model. We found that it was possible to choose the parameters in GAM so that the model performed well in three extremely different climate states over Fennoscandia. We were able to downscale monthly temperature and precipitation values for the full previous glacial cycle simulation as well as for the two future climate scenarios. In the simulation of the last glacial cycle, large climatic variations occurred at Olkiluoto: a mild interglacial climate, a cold periglacial climate, a glacial climate with an ice sheet over the Olkiluoto area, and Olkiluoto submerged below sea level due to glacial load. The timings of the ice advances and retreats agreed with the paleoclimatic reconstructions on Northern Hemispheric scale, but there were differences between the simulated and reconstructed ice sheet extent over Northern Europe. Sensitivity studies showed that the simulated ice sheet evolution was sensitive to the model system’s surface energy and mass-balance interface and dust module. These sensitivities lead to uncertainties on the local extent of the simulated ice sheet. Therefore, in the simulations of the CLIMBER-SICOPOLIS model system the uncertainties about the location of the ice sheet margin are large, and in case an area is near the margin, it might as well be ice free as beneath the ice sheet. In the CLIMBER-2-SICOPOLIS simulations into the future with a constant 280 ppm CO2 concentration, Northern Europe experienced three stadials with ice sheet advances during the next 120 kyr. The first stadial started immediately during the next millennia with the coldest period occurring at 15-25 kyr AP. The second stadial occurred around 50–60 kyr AP. The third stadial, at around 90–100 kyr AP, was colder than the two previous ones, and during this the ice sheet over Olkiluoto was, at its thickest, about 1.7 km thick. Between the stadials, i.e., during the interstadials, the temperature conditions at Olkiluoto were near those of the present day. In the CLIMBER-2-SICOPOLIS simulation 120 kyr into the future with a constant 400 ppm CO2 concentration, a mild interglacial climate prevailed throughout the simulation and no ice sheet existed in the vicinity of Olkiluoto. A summary of the future simulations performed with CLIMBER-2 and other EMICs was presented in Chapter 2.2 in Table 3. All EMIC models (CLIMBER-2, MPM and

66

LLN-2D, MoBiDiC) brought out the feature that the onset of the next glaciation on the Northern Hemisphere is strongly dependent on the Earth’s orbital variations and on the atmospheric CO2 concentration. The CO2 concentration threshold required for the immediate1 onset of the next glaciation varied from model to model. The CLIMBER-2SICOPOLIS model system produced the onset of the next glaciation in Northern Europe and North America immediately with a CO2 concentration of 280 ppm. Cochelin et al. (2006) showed, using the MPM model, that glacial inception could be immediate with pre-industrial levels of CO2 (240–270 ppm). Higher CO2 concentration levels (280–290 ppm) delayed the glacial inception to about 50 kyr from now. With the MPM model, even higher CO2 concentrations (>300 ppm) pushed the next glacial inception beyond of the next 100 kyr period. The LLN-2D NH EMIC simulated an immediate onset of the next glaciation on the Northern Hemisphere only with CO2 concentrations of 230 ppm or lower (Loutre & Berger 2000). The modelling results of the EMICs clearly reveal the sensitivity of the models to the atmospheric CO2 concentration. The future atmospheric CO2 concentration will have a crucial effect on the global climate. Present-day greenhouse gas concentrations are high in a glacial perspective, and will stay high for thousands of years (Archer 2005, Eby et al. 2009). With present-day greenhouse gas projections, the next glacial inception seems unlikely before 30 kyr AP. Hence, for the next 30 kyr the Earth’s climate will probably remain in an interglacial state. After 30 kyr the possibility of the onset of a glaciation increases, being highest during the Northern Hemisphere insolation minima at 50–60 kyr AP and 90–100 kyr AP. Nonetheless, sustained high atmospheric greenhouse gas concentrations might even further delay the onset of the next glacial period (Loutre & Berger 2000, Archer & Ganopolski 2005). It remains impossible to predict future climate development exactly. Nonetheless, scenarios can be constructed. Kjellström et al. (2009) defined possible extreme climatic situations on the time-scale of 100 kyr. These extreme climatic situations were: a glacial climate, a periglacial climate and a climate warmer than at present. With the EMIC simulations 120 kyr into the future, we now estimate the potential time periods (or time windows) for these extreme climatic periods to occur over Fennoscandia and Olkiluoto. The simulations for the future show that insolation minima at 10–20 kyr AP, 50–60 kyr AP and 90–100 kyr AP hold a potential for cooling the climate in the Northern Hemisphere. The onset of a glaciation at 10–20 kyr AP is unlikely for the reasons mentioned earlier in this chapter (high atmospheric greenhouse gas concentrations and low eccentricity). However, the later two stadials hold a higher potential for the onset of a glaciation. In the CLIMBER-2-SICOPOLIS 280 ppm scenario simulation, the CO2 level does not decrease during glaciation, and hence the climate remains warmer during the stadials than in a realistic glaciation. We therefore suggest that for investigating the effects of a glaciation, data of the CLIMBER-2-SICOPOLIS last glacial cycle simulation should be used. For investigating the effects of a glaciation over the Olkiluoto region, the MIS 4 stadial ca. 70–60 kyr BP in the simulation is a glaciation lasting about 25 kyr, with the 1

here immediate onset of the next glaciation means that glaciers start to build up on the Northern Hemisphere during the next 10,000 years

67

ice sheet thickness reaching 2.5 km in that period. The area of Olkiluoto is depressed about 600 m. The simulated ice sheet retreated in about 8 kyr, which is comparable to the ice sheet retreat speed during the end of the last glaciation. A periglacial climate could occur at Olkiluoto between the stadials, in the period about 70–80 kyr AP. For investigating a periglacial climate, data for the MIS 3 (from 55kyr BP until 30 kyr BP), with a small ice sheet over Fennoscandia, is recommended to be used. Periglacial climate could occur at Olkiluoto during the stadials 50–60 kyr AP or 90–100 kyr AP also, if there was glaciation in Fennoscandia, but the ice sheet would not reach Olkiluoto. The Olkiluoto area could be near an ice sheet margin during a stadial as well as before and after a glaciation. For investigating climate near an ice sheet margin the data of the MIS 5b stadial about 93-85 kyr BP is recommended. It is evident that, due to the ongoing global warming, the climate of the next centuries will be warmer and wetter than at present. For investigating a climate warmer than the present, the data of the 400 ppm CLIMBER-2 simulation is recommended. On the other hand, for investigating the effects of an extreme warm climate, we recommend that the data of the RCA3 regional model simulation with a CO2 concentration of 750 ppm be used. The delivered data is listed in Appendix 3. It should be remembered that unpredictable events like super-volcanoes and meteorites are not considered in these scenarios. The present scenarios are based on the current best knowledge. The next step to improve glacial-scale (100 kyr) climate scenarios is the development of a fully coupled sophisticated climate-ice sheet-carbon cycle model.

68

69

7

ACKNOWLEDGEMENTS

We owe our gratitude to Andrey Ganopolski of the Potsdam Institute for Climate Impact Research for providing the climate simulations of CLIMBER-2-SICOPOLIS. Erik Kjellström and Gustav Strandberg of the Rossby Centre, Swedish Meteorological and Hydrological Institute, are acknowledged for the regional climate simulations with RCA3 model. We acknowledge Heikki Seppä from the Department of Geosciences and Geography, University of Helsinki for providing temperature and precipitation reconstructions for Northen Europe. We acknowledge the Climate Research Unit for the 1961-1990 high-resolution temperature and precipitation data. CMAP Precipitation data were provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, USA, from their Web site at http://www.esrl.noaa.gov/psd/ Elevation data were downloaded from the webpage of the National Geophysical Data Center, NOAA Satellite and Information Service (http://www.ngdc.noaa.gov/mgg/gdas/gd_designagrid.html). We acknowledge the contributors to the R project. The GAM fitting was done through the R software package “mgcv” downloaded from website of the R project for statistical computing (http://www.r-project.org/) We acknowledge Kimmo Ruosteenoja from the Finnish Meteorological Institute, Heini Laine and Nuria Marcos from Saanio & Riekkola Oy, Ari Ikonen, Anne Lehtinen and Aimo Hautojärvi from Posiva Oy, and Erik Kjellström from the Swedish Meteorological and Hydrological Institute for providing valuable comments on the manuscript, and Robin King for language-checking.

70

71

REFERENCES Anderson, P., Bermike, O., Bigelow, N., Brigham-Grette, J., Duvall, M., Edwards, M. E., Frechette, B., Funder, S., Johnsen, S., Knies, J., Koerner,R., Lozhkin, A., Marshall, S., Matthiessen, J., Macdonald, G., Miller, G., Montoya, M., Muhs, D., Otto-Bliesner, B., Overpeck, J., Reeh, N., Sejrup, H., Spielhagen, R., Turner, C. & Velichko, A. 2006. Last Interglacial Arctic warmth confirms polar amplification of climate change. Quaternary Science Reviews, 25, 1383–1400. Archer, D. 2005. Fate of fossil fuel CO2 in geologic time. Journal of Geophysical Research, 110, C09S05, doi:10.1029/2004JC002625. Archer, D. & Ganopolski, A. 2005. A movable trigger: Fossil fuel CO2 and the onset of the next glaciation. Geochemistry, Geophysic, Geosystems. 6, doi:10.1029/2004GC000891. Archer, D. & Brovkin, V. 2008. The millennial atmospheric lifetime of anthropogenic CO2. Climatic Change, 90, 283-297. Archer, D., Eby, M., Brovkin, V., Ridgwell, A., Cao, L., Mikolajewicz, U., Caldeira, K., Matsumoto, K., Munhoven, G., Montenegro, A. & Tokos, K. 2009. Atmospheric Lifetime of Fossil Fuel Carbon Dioxide. Annual Review of Earth and Planetary Sciences, 37, 117-134. Arppe, L. & Karhu, J.A. 2006. Implications for the late Pleistocene climate in Finland and adjacent areas from the isotopic composition of mammoth skeletal remains. Palaeogeography, Palaeoecology, 231, 322-330. Bamber J.L., Riva E.M.R., Vermeersen B.L.A & LeBrocq A.M. 2009. Reassessment of the Potential Sea-Level Rise from a Collapse of the West Antarctic Ice Sheet. Science, 342, 901-903. Barnola, J.-M., Raynaud, D., Korotkevich, Y.S. & Lorius, C. 1987. Vostok ice core provides 160,000-year record of atmospheric CO2. Nature, 329, 408–414. Barrows T.T., and Juggins S. 2005. Sea-surface temperatures around the Australian margin and Indian Ocean during the Last Glacial Maximum. Quaternary Science Reviews, 24, 1017–1047. Berger, A. 1978. Long-Term variations of Caloric Insolation Resulting from the Earth’s orbital Elements. Quarternary Research, 9, 139–167. Berger, A. & Loutre, M.F. 1991. Insolation values for the climate of the last 10 million years. Quaternary Science Reviews, 10, 297–317. Berger, A. & Loutre, M.F. 2002. An exceptionally long interglacial ahead? Science, 297, 1287–1288.

72

Berger, A., Li, X.S. & Loutre, M.F., 1999. Modeling northern hemisphere ice volume over the last 3 million years, Quaternary Science Reviews, 18, 1–11. Berger A., Loutre M.F. & Crucifix M. 2003. The Earth’s Climate in the Next hundred thousand years (100 kyr). Surveys of Geophysics, 24, 117–138. BIOCLIM 2003. Deliverable D7: Continuous climate evolution scenarios over western Europe (1000km scale). http://www.andra.fr/bioclim/ BIOCLIM 2004. Deliverable D10 - 12: Development and Application of a Methodology for Taking Climate-Driven Environmental Change into Account in Performance Assessments. http://www.andra.fr/bioclim/ Björck, S. 1995. A review of the history of the Baltic Sea, 13.0–8.0 ka BP. Quaternary International, 27, 19–40. Bonan, G.B., Levis, S., Kergoat, L. & Oleson, K.W. 2002. Landscapes as patches of plant functional types: An integrating concept for climate and ecosystem models. Global Biogeochemical Cycles, 16, 1021, doi:10.1029/2000GB001360. Bonan, G.B. & Levis, S. 2006. Evaluating aspects of the Community Land and Atmosphere Models (CLM3 and CAM3) using a dynamic global vegetation model. Journal of Climate 19, 2290–2301. Bradley, R.S. 1999. Paleoclimatology. Reconstructing Climates of the Quaternary, 2nd ed. International Geophysics Series, Volume 68. xv+613 pp. San Diego, London, Boston, New York, Sydney, Tokyo, Toronto: Academic Press. Brovkin, V., Ganopolski, A. & Svirezhev, Y. 1997. A continuous climate-vegetation classification for use in climate-biosphere studies. Ecological Modelling, 101, 251–261. Calov, R., Ganopolski, A., Claussen, M., Petoukhov, V. & Greve, R. 2005a. Transient simulation of the last glacial inception. Part I: glacial inception as a bifurcation in the climate system. Climate Dynamics, 24, 545–561. Calov, R., Ganopolski, A., Claussen, M., Petoukhov, V., Claussen, M., Brovkin, V. & Greve, R. 2005b. Transient simulation of the last glacial inception. Part II: sensitivity and feedback analysis. Climate Dynamics, 24, 563–576. Cheng, M & Qi, Y. 2002. Frontal rain-fall rate distribution and some conclusions on the threshold method. Journal of Applied Meteorology, 41, 1128-1139. Church, J.A. & White, N.J. 2006. A 20th century acceleration in global sea-level rise. Geophysical Research Letters, 33, L01602. Claussen, M., Mysak, L., Weaver, A., Crucifix, M., Fichefet, T., Loutre, M.F., Weber, S., Alcamo, J., Alexeev, V., Berger, A., Calov, R., Ganopolski, A., Goosse, H., Lohmann, G., Lunkeit, F., Mokhov, I., Petoukhov, V., Stone, P. & Wang, Z. 2002.

73

Earth system Models of Intermediate Complexity: Closing the gap in the spectrum of climate system models, Climate Dynamics, 18, 579–586. Cochelin, A.-S.B., Mysak, L.A. & Wang, Z. 2006. Simulation of long-term future climate changes with the green McGill paleoclimate model: The next glacial inception. Climatic Change, 79, 381–401, DOI 10.1007/s10584-006-9099-1 Collins, W.D., Bitz, C.M., Blackmon, M.L., Bonan, G.B., Bretherton, C.S., Carton, J.A., Chang, P., Doney, S.C., Hack, J.J., Henderson, T.B., Kiehl, J.T., Large, W.G., McKenna, D.S., Santer, B.D. & Smith, R.D. 2006.: The Community Climate System Model (CCSM3), Journal of Climate, 19, 2122–2143. Crucifix, M., Loutre, M.F. & Berger, A. 2006. The climate response to the astronomical forcing. Space reviews, 125, 213–226. Cuffey, K.M. & Marshall, S.J. 2000. Substantial contribution to sea-level rise during the last interglacial from the Greenland ice sheet. Nature, 404, 591–594. doi:10.1038/35007053. Dahl-Jensen D., Mosegaard K., Gundestrup N., Clow G. D., Johnsen S. J., Hansen A. W. & Balling N. 1998. Past temperature directly from the Greenland Ice Sheet. Science, 282, 268–271. Deblonde, G., Peltier, W.R. & Hyde, W.T. 1992. Simulations of continental ice sheet growth over the last glacial-interglacial cycle: experiments with one level seasonal energy balance model including seasonal ice albedo feedback. Global Planetary Change, 6, 37–55. Denman, K.L., Brasseur, G., Chidthaisong, A., Ciais, P., Cox, P.M., Dickinson R.E., Hauglustaine, D., Heinze, C., Holland, E., Jacob, D., Lohmann, U., Ramachandran, S., da Silva Dias, P.L., Wofsy, S.C. & Zhang, X. 2007. Couplings between Changes in the Climate System and Biogeochemistry. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor & H.L. Miller (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Dickens, G.R., Castillo, M.M. & Walkeret J.C.G., 1997. A blast of gas in the latest Paleocene: Simulating first order effects of massive dissociation of oceanic methane hydrate. Geology, 25(3), 259–262. Eby, M., Zickfeld, K., Montenegro, A., Archer, D., Missner, K.J. & Weaver, A.J. 2009. Lifetime of Anthropogenic Climate Change: Millennial Time Sale of Potential CO2 and Surface Temperature Perturbations. Journal of Climate, 22, 2501–2511. Ehlers, J. & Gibbard, P.L. (Eds.) 2004. Quaternary glaciations – extent and chronology. Part I: Europe. Elsevier, 475 pp. EPICA community members 2004. Eight glacial cycles from an Antarctic ice core. Nature, 429(6992), 623–628.

74

Eronen, M., 1990. Itämeren kehitys. In Alalammi, P. (ed). Atlas of Finland, Folio 123– 126: Geology, 15–18. National Board of Survey & Geographical Society of Finland, Helsinki. Eronen, M., Glückert, G., van de Plassche, O., van der Plicht, J. & Rantala, P. 1995. Land uplift in the Olkiluoto-Pyhäjärvi area, south-western Finland, during the last 8000 years. Report YJT-95-17. Nuclear Waste Commission of Finnish Power Companies, Helsinki. 26 p. Eronen, M. & Lehtinen, K. 1996. Description of the Quaternary geological history of Romuvaara, Kivetty and Olkiluoto sites (in Finnish with an English abstract). Posiva Oy, Eurajoki, Finland. Working Report PATU-96-74, 37 p. Funder, S., Demidov, I. & Yelovicheva, Y. 2002. Hydrography and mollusc faunas of the Baltic and the White Sea-North Sea seaway in the Eemian. Palaeogeography, Palaeoclimatology, Palaeoecology, 184, 275–304. Fyke, J.G. & Weaver, A.J. 2006: The effect of potential future climate change on the marine methane hydrate stability zone. Journal of Climate, 19, 5903–5916. Gallée, H., van Ypersele, J.P., Fichefet, T., Tricot, C. & Berger, A. 1991. Simulations of the last glacial cycle by a coupled, sectorially averaged climate ice-sheet model. 1. The climate model, Journal of Geophysical reseach, 96, 13139–13161. Ganopolski, A., Petoukhov, V., Rahmstorf, S., Brovkin, V., Claussen, M., Eliseev, A. & Kubatzki, C. 2001. CLIMBER-2: a climate system model of intermediate complexity. Part II: model sensitivity. Climate Dynamics, 17, 735–751. Ganopolski, A., Calov, R. & Claussen, M. 2010. Simulation of the last glacial cycle with a coupled climate ice-sheet model of intermediate complexity. Climate of the Past, 6, 229–244. Gregory, J. M. & Huybrechts, P. 2006. Ice-sheet contributions to future sea-level change. Philosophical Trasactions of the Royal Society A, 364, 1709–1731. Greve, R. & Hutter, K. 1995. Polythermal three-dimensional modelling of the Greenland ice sheet with varied geothermal heat flux. Annals of Glaciology, 21, 8–12. Greve, R. 1997. Application of a polythermal three-dimensional ice sheet model to the Greenland ice sheet: Response to steady-state and transient climate scenarios. Journal of Climate, 10, 901–918. Greve, R. 2005. Relation of measured basal temperatures and the spatial distribution of the geothermal heat flux for the Greenland ice sheet. Annals of Glaciology 42, 424–432. Harvey, L.D.D. 2000. Global Warming: The hard science. Pearson Education Limited, Harlow, 336 pp. ISBN 0582-381167-3.

75

Indermühle A., Monnin E., Stauffer B., Stocker T. F. & Wahlen M. 2000. Atmospheric CO2 concentration from 60 to 20 kyr BP from the Taylor Dome ice core, Antarctica. Geophysical Research Letters, 27(5), 735–738. IPCC, 2007. Summary for Policymakers. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M. & Miller, H.L. (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Jansen, E., Overpeck, J., Briffa, K.R., Duplessy, J.-C., Joos, F., Masson-Delmotte, V., Olago, D., Otto-Bliesner, B., Peltier, W.R., Rahmstorf, S., Ramesh, R., Raynaud, D., Rind, D., Solomina, O., Villalba R. & Zhang, D. 2007: Palaeoclimate. In: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., Qin, D., Manning, M., Chen, Z., Marquis, M., Averyt, K.B., Tignor, M. & Miller, H.L. (eds.)]. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA. Johansson M. M., Kahma K. K., Boman H. & Launiainen J. 2004. Scenarios for sea level on the Finnish coast. Boreal Environment Research, 9, 153–166. Jones P. D., New M., Parker D. E., Martin S. & Rigor I. G. 1999. Surface air temperature and its changes over the past 150 years. Reviews of Geophysics 37(2), 173199. Jylhä, K., Ruosteenoja, K., Räisänen, J., Venäläinen, A., Tuomenvirta, H., Ruokolainen, L., Saku, S. & Seitola, T. 2009. Arvioita Suomen muuttuvasta ilmastosta sopeutumistutkimuksia varten. ACCLIM-hankkeen raportti 2009. (The changing climate in Finland: estimates for adaptation studies. ACCLIM project report 2009.) Ilmatieteen laitos, Raportteja 2009, 4, 102. (In Finnish, abstract, extended abstract and captions for figures and tables also in English) Katz, M.E., Pak, D.K., Dickens, G.R., & Miller, K.G. 1999. The source and fate of massive carbon input during the Latest Paleocene Thermal maximum. Science, 286, 1531–1533. Keeling, C.D., and T.P. Whorf, 2005: Atmospheric CO2 records from sites in the SiO air sampling network. In: Trends: A Compendium of Data on Global Change. Carbon Dioxide Information Analysis Center, Oak Ridge National Laboratory, U.S. Department of Energy, Oak Ridge, TN. Kennett, J.P. & Stott, L.D. 1991. Abrupt deep-sea warming, palaeoceanographic changes and benthic extinctions at the end of the Palaeocene. Nature 353, 225–229. Kiehl, J. T. & Trenberth, K. E. 1997. Earth's Annual Global Mean Energy Budget. Bull. Amer. Meteor. Soc., 78, 197–208.

76

Kiehl, J.T., Shields, C. A., Hack, J.J. & Collins, W.D. 2006. The Climate Sensitivity of the Community Climate System Model Version 3 (CCSM3) Journal of Climate, 19, 2584–2596. Kivioja, L.A. 1967. Effects of mass transfers between land-supported ice caps and oceans on the shape of the earth and on the observed mean sea level. Bulletin Géodésique. 85, 281–288. Kjellström, E., Bärring, L., Gollvik, S., Hansson, U., Jones, C., Samuelsson, P., Rummukainen, M., Ullerstig, A., Willén U. & Wyser, K., 2005. A 140-year simulation of European climate with the new version of the Rossby Centre regional atmospheric climate model (RCA3). Reports Meteorology and Climatology, 108, SMHI, SE-60176 Norrköping, Sweden, 54 pp. Kjellström, E., Strandberg, G., Brandefelt, J., Näslund, J.-O., Smith, B. & Wohlfarth, B., 2009. Climate conditions in Sweden in a 100,000 year time perspective. Swedish Nuclear Fuel and Waste Management Co. (SKB) Stockholm, Sweden. SKB Technical Report TR-09-04. Kjellström, E., Brandefelt, J., Näslund, J.O., Smith, B., Strandberg, G., Voelker, A. H. L. & Wohlfarth, B. 2010. Simulated climate conditions in Fennoscandia during a MIS 3 stadial. Boreas 10. DOI 10.1111/j.1502-3885.2010.00143.x Lambeck, K., Esat, T.M. & Potter, E.K. 2002. Links between climate and sea levels for the past three million years, Nature, 419, 199–206. Lee, W.H.K. 1970. On the global variations of the terrestrial heat flow. Physics of the Earth and Planetary Interiors, 2, 332–341. Le Meur, E. & Huybrechts, P. 1996. A comparison of different ways of dealing with isostasy: examples from modelling the Antarctic ice sheet during the last glacial cycle. Annals of Glaciology, 23, 309–317. Lenton T., Wilkinson P., Edwards N, Marsh R., Price J, Ridgwell A, Shepherd J., Cox S.J., The GENIE team, 2006. Millennial timescale carbon cycle and climate change in an efficient Earth system model. Climate Dynamics 26: 687–711 Lenton T. M., H. Held, E. Kriegler, J. W. Hall, W. Lucht, S. Rahmstorf and H. J. Schellnhuber, 2008. Tipping Elements in the Earth System, Proceedings of the National Academy of Sciences USA, 105, 6, 1786–1793. Loutre, M.F. & Berger, A. 2000. Future Climate Changes: Are we entering an exceptionally long Interglacial? Climatic Change, 46, 61–90. Loutre M.F. & Berger A.L. 2002. Marine Isotope stage 11 as an analogue for the present interglacial. Global and Planetary Change, 36, 209–217. Lunkka, J. P., Saarnisto, M., Gey, V., Demidov, I. & Kiselova, V. 2001. Extent and age of the Last Glacial Maximum in the south-eastern sector of the Scandinavian Ice Sheet. Global and Planetary Change, 31, 407–425.

77

Lunkka, J. P., Murray, A. & Korpela, K. 2008. Weichselian sediment succession at Ruunaa, Finland, indicating a Mid-Weichselian ice-free interval in eastern Fennoscandia. Boreas, 37, 234–244. MacFarling Meure, C., et al., 2006: The Law Dome CO2, CH4 and N2O ice core records extended to 2000 years BP. Geophys. Res. Lett., 33, L14810, doi:10.1029/2006GL026152. Martin, A., Vrac, M., Paillard, D., Dumas, C. & Kageyama, M. 2010a. Statisticaldynamical downscaling for an Earth Model of Intermediate Complexity. Part I: Methodology and calibrations. Climate Dynamics (submitted) Martin, A., Vrac, M., Paillard, D., Dumas, C. & Kageyama, M. 2010b. Statisticaldynamical downscaling for an Earth Model of Intermediate Complexity. Part II: Application to past and future climates. Climate Dynamics (submitted) Masson-Delmotte V., Kageyama M., Braconnot P., Charbit S., Krinner G., Ritz C., Guilyardi E., Jouzel J., Abe-Ouchi A., Crucifix M., Gladstone R. M., Hewitt C. D., Kitoh A., LeGrande A. N., Marti O., Merkel U., Motoi T., Ohgaito R., Otto-Bliesner B., Peltier W. R., Ross I., Valdes P. J., Vettoretti G., Weber S. L., Wolk F. & YU Y. 2006. Past and future polar amplification of climate change: climate model intercomparisons and ice-core constraints. Climate Dynamics, 26 (5), 513–529. Mikolajewicz U, Gröger M, Maier-Reimer E, Schurgers G, Vizcaíno M & Winguth A. 2007. Long-term effects of anthropogenic CO2 emissions simulated with a complex earth system model. Climate Dynamics, 28:599–633. doi:10.1007/s00382-006-0204-y. Mitchell, T.D. & Jones, P.D. 2005. An improved method of constructing a database of monthly climate observations and associated high-resolution grids. International Journal of Climatology 25, 693-712 (doi:10.1002/joc.1181). Mitrovica, J X., Tamisiea, M. E., Davis, J. L. & Milne, G. A. 2001. Recent mass balance of polar ice sheets inferred from patterns of global sea-level change. Nature, 409, 1026–1029. Monnin E., Steigb E. J., Siegenthalera U., Kawamuraa K., Schwandera J., Stauffera B., Stockera T. F., Morsec D. L., Barnolad J.-M., Bellierd B., Raynaudd D. & Fischere H. 2004. Evidence for substantial accumulation rate variability in Antarctica during the Holocene, through synchronization of CO2 in the Taylor Dome, Dome C and DML ice cores. Earth and Planetary Science Letters, 224(1–2), 45–54. Naish, T., Powell, R., Levy, R., Wilson, G., Scherer, R., Talarico, F., Krissek, L., Niessen, F., Pompilio, M., Wilson, T., Carter, L., Deconto, R., Huybers, P., McKay, R., Pollard, D., Ross, J., Winter, D., Barrett, P., Browne, G., Cody, R., Cowan, E., Crampton, J., Dunbar, G., Dunbar, N., Florindo, F., Gebhardt, C., Graham, I., Hannah, M., Hansaraj, D., Harwood, D., Helling, D., Henrys, S., Hinnov, L., Kuhn, G., Kyle, P., Läufer, A., Maffioli, P., Magens, D., Mandernack, K., McIntosh, W., Millan, C., Morin, R., Ohneiser, C., Paulsen, T., Persico, D., Raine, I., Reed, J., Riesselman, C., Sagnotti, L., Schmitt, D., Sjunneskog, C., Strong, P., Taviani, M., Vogel, S., Wilch, T. &

78

Williams, T. 2009. Obliquity-paced Pliocene West Antarctic ice sheet oscillations. Nature, 458, 322–328. Nakicenovic, N. & Swart, R. (eds.). 2000. Special Report on Emission Scenarios. A special report of Working group III of the Intergovernmental Panel on Climate Change. Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 599 pp. Näslund, J.-O., Jansson, P., Fastook, J., Johnson, J., & Andersson, L., 2005. Detailed spatially distributed geothermal heat flow data for modeling of basal temperatures and melt water production beneath the Fennoscandian ice sheet. Annals of Glaciology, 40, 95–101. Näslund, J.-O., Wohlfarth, B., Alexanderson, H., Helmens, K., Hättestrand, M., Jansson, P., Kleman, J., Lundqvist, J., Brandefelt, J., Houmark-Nielsen, M., Kjellström, E., Strandberg, G., Knudsen, K.L., Krog Larsen, N., Ukkonen, P. & Mangerud, J. 2008. Fennoscandian paleo-environment and ice sheet dynamics during Marine Isotope Stage (MIS) 3. Report of a workshop held September 20-21, 2007 in Stockholm, Sweden. Report R-08-79, Swedish Nuclear Fuel and Waste Management, SE-10124 Stockholm, Sweden, 52 pp. Otto-Bliesner, B.L., Brady, E.C., Clauzet, G., Tomas, R., Levis, S. & Kothavala, Z., 2006. Last glacial maximum and Holocene climate in CCSM3. Journal of Climate 19: 2526–2544. Pagani, M., Caldeira, K., Archer, D. & Zachos, J.C. 2006. An ancient carbon mystery. Science, 314, 1556–1557. Peltier, W.R. 2004. Global glacial isostacy and the surface of the ice-age circulation. Earth: The ICE-5G (VM2) model and GRACE. Annual Review of Earth and Planetary Sciences, 32, 111–149. Petit, J.R., Jouzel, J., Raynaud, D., Barkov, N. I., Barnola, J. M., Basile, I., Bender, M., Chappellaz, J., Davis, M., Delaygue, G., Delmotte, M., Kotlyakov, V. M., Legrand, M., Lipenkov, V.Y., Lorius, C., Pepin, L., Ritz, C., Saltzman, E. & Stievenard, M. 1999. Climate and atmospheric history of the past 420,000 years from the Vostok ice core, Antarctica. Nature, 399, 429–436. Petoukhov, V., Ganopolski A., Brovkin V., Claussen M., Eliseev A., Kubatzki & Rahmstorf S. 2000, CLIMBER-2: A climate system model of intermediate complexity. Part I: Model description and performance for present climate. Climate Dynamics, 16, 1–17. Petoukhov, V., Claussen M., Berger A., Crucifix M., Eby M., Eliseev A.V., Fichefet T., Ganopolski A., Goosse H., Kamonkovich I., Mokhov I.I., Montoya M., Mysak L.A., Sokolov A., Stone P., Wang, Z. & Weaver, A.J. 2005. EMIC Intercomparison Project (EMIP-CO2): comparative analysis of EMIC simulations of climate, and of equilibrium and transient responses to atmospheric CO2 doubling. Climate Dynamics, 25,363–385.

79

Pollack, H.N., Hurter, S.J. & Johnson, J.R. 1993. Heat flow from the Earth’s interior: analysis of the global data set. Reviews of Geophysics, 31, 267–280. Pollard, D. 1982. A simple ice-sheet model yields realistic 100 kyr glacial cycles, Nature, 296, 334–338. Pollard, D. & Deconto, R. M. 2009. Modelling West Antarctic ice sheet growth and collapse through the past five million years. Nature, 458, 329–332. Richey, J.E. 2004. Pathways of atmospheric CO2 through fluvial systems. In: The Global Carbon Cycle: Integrating Humans, Climate, and the Natural World [Field, C. & Raupach, M. (eds.)]. SCOPE 62, Island Press, Washington, DC, pp. 329–340. Rogner, H.H. 1997. An assessment of world hydrocarbon resources. Annual Review of Energy and the Environment, 22, 217-262. Rosell-Melé A., Bard E., Emeis K-C., Grieger B., Hewitt C., Müller P. J. and Schneider R. R.2004: Sea surface temperature anomalies in the oceans at the LGM estimated from the alkenone-UK’37 index: comparison with GCMs. Geophyical Research Letters, 31, L03208, doi:10.1029/2003GL018151. Royer, D. 2006. CO2-forced climate thresholds during the Phanerozoic. Geochimica et Cosmochimica Acta, 70(23), 5665–5675. Sabine, C.L., Heimann, M., Artaxo, P., Bakker, D., Chen, C.-T.A., Field, C.B., Gruber, N., Le Quéré, C., Prinn, R.G., Richey, J.E., Lankao, P.R., Sathaye, J. & Valentini, R. 2004. Current status and past trends of the global carbon cycle. In: The Global Carbon Cycle: integrating Humans, Climate and the Natural World [Field, C. & M. Raupach (eds)]. SCOPE 62, Island Press, Washington, DC, pp. 17–44. Samuelsson, P., Jones, C., Willén, U., Ullerstig, A., Gollvik, S., Hansson, U., Kjellström, E., Nikulin, G. & Wyser, K. 2010. The Rossby Centre Regional Climate Model RCA3: Model description and performance. Tellus, published online. DOI: 10.1111/j.1600-0870.2010.00478.x Scherer, R. P., Aldahan, A., Tulaczyk, S., Possnert, G., Engelhardt, H. & Kamb, B. 1998. Pleistocene Collapse of the West Antarctic Ice Sheet. Science, 281, 82–85. Schneider von Deimling, T., Held, H., Ganopolski, A. & Rahmstorf, S. 2006. Climate sensitivity estimated from ensemble simulations of glacial climate. Climate Dynamics, 27, 149–163. Sholze, M., Knorr, W., Arnell, N.W. & Prentice, I.C., 2006. A climate-change risk analysis for world ecosystems. Proceedings of the National Academy of Sciences of the United States of America, 103: 13116-13120. Siddall, M., Rohling, E.J., Thompson, W.G. & Waelbroeck C. 2008. MIS 3 Sea-level fluctuations: data synthesis and new outlook, Reviews of Geophysics, 46, RG4003, doi:10.1029/2007RG000226

80

Siegenthaler, U., et al., 2005a: Stable carbon cycle-climate relationship during the late Pleistocene. Science, 310(5752), 1313–1317. Siegenthaler, U., et al., 2005b: Supporting evidence from the EPICA Dronning Maud Land ice core for atmospheric CO2 changes during the past millennium. Tellus, 57B(1), 51–57. Siegert, M.J., Dowdeswell, J.A., Melles, M., 1999. Late Weichselian glaciation of the Russian High Arctic. Quaternary Research, 53, 273–285. Siegert, M.J., Dowdeswell, J.A., Hald, M. & Svendsen, J.I., 2001. Modelling the Eurasian Ice Sheet through a full (Weichselian) glacial cycle. Global and Planetary Change, 31, 367-385. Sime, L. C., Wolff, E. W., Oliver, K. I. C. & Tindall, J. C. 2009. Evidence for warmer interglacials in East Antarctic ice cores. Nature, 462, 342–345. SKB 2006. Climate and climate related issues for the safety assessment SR-Can. SKB TR-06-23, Swedish Nuclear waste management company, 186 pp. Smith, B., Prentice, I.C. & Sykes, M.T. 2001. Representation of vegetation dynamics in modelling of terrestrial ecosystems: comparing two contrasting approaches within European climate space. Global Ecology and Biogeography 10: 621-637. Spahni R., Chappellaz J., Stocker T. F., Loulergue L., Hausammann G., Kawamura K., Flückiger J., Schwander J., Raynaud D., Masson-Delmotte V. & Jouzel J. 2005: Atmospheric methane and nitrous oxide of the late Pleistocene from Antarctic ice cores. Science, 310(5752), 1317–1321. Stenni B., Masson-Delmotte V., Johnsen S., Jouzel J., Longinelli A., Monnin E., Röthlisberger R. & Selmo E. 2001. An oceanic cold reversal during the last deglaciation. Science, 293, 2074–2077. Strandberg, G., Brandefelt, J., Kjellström, E. & Smith, B. 2010. High resolution regional simulation of Last Glacial Maximum climate in Europe. Tellus. published online. DOI: 10.1111/j.1600-0870.2010.00485.x Svendsen, J. I., Alexanderson, H., Astakhov, V. I., Demidov, I., Dowdeswell, J. A., Funder, S., Gataullin, V., Henriksen, M., Hjort, C., Houmark-Nielsen, M., Hubberten, H. W., Ingólfsson, Ó., Jakobsson, M., Kjær, K. H., Larsen, E., Lokrantz, H., Lunkka, J. P., Lyså, A., Mangerud, M., Matiouchkov, A., Murray, A., Möller, P., Niessen, F., Nikolskaya, O., Polyak, L., Saarnisto, M., Siegert, C., Siegert, M. J., Spielhagen, R. & Stein, R. 2004. Late Quaternary ice sheet history of northern Eurasia. Quaternary Science Reviews, 23, 1229-1271. Tans, P. 2010. NOAA/ESRL (www.esrl.noaa.gov/gmd/ccgg/trends). Texier, D., Degnan, P., Loutre, M.F., Paillard, D. & Thorne, M. 2003. Modelling Sequential BIOsphere Systems under CLIMate Change for Radioactive Waste disposal.

81

Project BIOCLIM. Proceedings of the international high-level waste management conference (30.3-2.4.2003, Las Vegas, U.S.A). Thompson, W.G. & Goldstein, S.L. 2006. A radiometric calibration of the SPECMAP timescale, supplementary data, Quaternary Sci. Rev.., 25, 3207-3215, doi: 10.1016/j.quascirev.2006.02.007. Ukkonen, P., Lunkka, J.P., Junger, H. & Donner, J. 1999. New radiocarbon dates from Finnish mammoths indicating large ice-free areas in Fennoscandia during the Middle Weichselian. Journal of Quarternary Science 14, 711-714. Velicogna, I. 2009. Increasing rates of ice mass loss from the Greenland and Antarctic ice sheets revealed by GRACE. Geophysical Research Letters, 36, L19503. Vrac, M., Marbaix, P., Paillard, D. & Naveau, P., 2007. Non-linear statistical downscaling of present and LGM precipitation and temperature over Europe. Climate of the Past, 3, 669-682. Waelbroeck, C., Labeyrie, L., Michel E., Duplessy, J.C., McManus, J.F., Lambeck, K., Balbon, E. & Labracherie, M. 2002. Sea-level and deep water temperature changes derived from benthonic foraminifera isotopic records. Quaternary Science Reviews, 21, 295-305. Wang, Z. & Mysak, L. 2002. Simulation of the last glacial inception and rapid ice sheet growth in the McGill Paleoclimate Model. Geophysical Research Letters, 29, 2102, doi:10.1029/2002GL015120 Wood, S.N., 2006. Generalized additive Models: An Introduction with R. CRC/Chapman & Hall. Xie, P. & Arkin, P.A. 1997: Global precipitation: A 17-year monthly analysis based on gauge observations, satellite estimates, and numerical model outputs. Bull. Amer. Meteor. Soc., 78, 2539–2558.

82

83

APPENDIX 1

RCA3 MODEL SETUP AND RESULTS

Kjellström et al. (2009) used the regional climate model RCA3 (Kjellström et al. 2005, Samuelsson et al. 2010) for downscaling CCSM3 global model simulations (Collins et al. 2006; Otto-Bliesner et al. 2006; Kiehl et al. 2006) in three climatic periods: i) warm case: a possible future period in a climate warmer than today. The future case is characterised by high greenhouse gas concentrations in the atmosphere and a complete loss of the Greenland ice sheet. ii) glacial case: the Last Glacial Maximum (LGM) at 21 kyr BP, with an extensive ice sheet covering large parts of northern Europe (Strandberg et al. 2010). iii) permafrost case: a stadial within Marine Isotope Stage 3 (MIS 3) at 44 kyr BP, representing a cold period with a relatively small ice sheet covering parts of Fennoscandia (Kjellström et al. 2010). The land and ice extent of the glacial and permafrost case simulations are depicted in Figure A1.1. The distribution of land areas and elevation in the warm, glacial and permafrost cases are depicted in Figure A2.2. Results of the simulated glacial and permafrost case temperature and precipitation fields over Europe are depicted in Figures A1.3-6.

Figure A1.1. Land (green) and ice extent (blue) in RCA3 in Europe in the glacial case (LGM) (left) and the permafrost case (MIS3) (right). Grid boxes with a land fraction lower than 20 % are not filled. Figure source Kjellström et al. 2009.

84

3000 2500 2000 1500 1000 500 0

Figure A1.2. The distribution of land areas and elevation (m) in RCA3 for the warm (left), glacial (middle) and permafrost (right) cases. Grid boxes with a land fraction lower than 20 % are not filled. Figure source Kjellström et al. 2009.

July

January 42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

Annual mean 42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

Figure A1.3. Mean temperatures of the warmest and coldest month and annual mean for the RCA3 LGM simulation. Unit: degrees Celsius. Figure source Kjellström et al. 2009. July

January 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

Annual mean 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

Figure A1.4. Mean precipitation of the warmest month, coldest month and annual mean in the RCA3 LGM simulation. Unit: mm/month. Figure source Kjellström et al. 2009.

420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

85

July

January 42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

Annual mean 42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

42 36 30 24 18 12 6 0 −6 −12 −18 −24 −30 −36 −42

Figure A1.5. Mean temperatures of the warmest and coldest month and annual mean for the RCA3 permafrost simulation. Unit: degrees Celsius. Figure source Kjellström et al. 2009. July

January 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

Annual mean 420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

420 390 360 330 300 270 240 210 180 150 120 90 60 30 0

Figure A1.6. Mean precipitation of the warmest month, coldest month and annual mean in the RCA3 permafrost simulation. Unit: mm/month. Figure source Kjellström et al. 2009.

86

87

APPENDIX 2

EVALUATION OF THE STATISTICAL MODEL

Statistical diagnoses For the evaluation of the GAMs we calculated following statistical quantities, which are shown for the GAMs of each month in Tables A2.1 and A2.3: the percentage of Explained Deviance:

%ED 100 *

(Ym  Yo ) 2 (Yo  Yo ) 2

,

(1)

where Y0 stands for the CRU/RCA3 and Ym for the GAM-modeled surface air temperature or log-precipitation. General Cross Validation score:

GCV

N (Ym  Ym ) 2 ( N  Jd ) 2

(2)

with Ȗ = 1.4 and d stands for the effective number of degrees of freedom and N for the number of observations. The GCV score is an estimate of the expected mean square predictor error and has the same unit as Y. The spatial correlation between CRU/RCA3 and GAM surface air temperature and log-precipitation is calculated for each month by Pearsson correlation:

Cor

Explained variance:

ev

 (Ym   Ym !) ˜ (Y0   Y0 !) !  (Ym   Ym !) 2 ! ˜  (Y0   Y0 !) 2 !

 (Ym   Y0 !) 2 !  (Y0   Y0 !) 2 !

(3)

(4)

Cor and ev inform about the GAM’s skill to model spatial variability of the predictand, i.e., monthly mean temperature or precipitation. The residual mean square error (RMSE) describes the prediction error of each GAM and is calculated by: RMSE

 (Y0  Ym ) 2 !  

The largest RMSE for the precipitation GAMs, P1-P12, in Table A2.3, were in October: 22 mm/month, and the smallest in February: 11 mm/month. The largest RMSE for the temperature GAMs, T1a-T12a in Table A2.1 (used for downscaling the temperature of the CLIMBER Last Glacial cycle simulation and the 120 kyr simulation into the future with CO2 concentration of 280 ppm), were in September: 1.7 ºC; and the smallest in April: 0.6 ºC. The largest RMSE for the temperature GAMs, T1b-T12b in Table A2.1 (used for downscaling the temperature of the 120 kyr simulation into the future with CO2 concentration of 400 ppm), were in January: 0.6 ºC; and the smallest in September: 0.3 ºC.

88

Table A2.1. Summary of configurations and skill scores of monthly surface air temperature GAM calibrations. The GAMs T1a-T12a are used for downscaling the temperature of the CLIMBER Last Glacial cycle simulation and the 120 kyr simulation into the future with CO2 concentration of 280 ppm. The GAMs T1b-T12b are used for downscaling the temperature of the 120 kyr simulation into the future with CO2 concentration of 400 ppm. The hyphen “-” means “same as above”. Abbreviations are explained in Table A2.2. GAM predictand,Y

%ED ev

T1a

99

0.99 3.5

RMSE (ºC) 0.99 1.7

99 99.2 99.4 94.6

0.98 0.99 0.99 0.94

3.0 1.4 0.5 2.8

0.99 0.99 0.99 0.97

1.6 1.1 0.6 1.6

95.2 95.9 96.1 96.6 99.3

0.95 0.96 0.96 0.96 0.99

2.8 2.9 2.8 3.2 1.4

0.98 0.98 0.98 0.98 0.99

1.6 1.6 1.6 1.7 1.1

T11a T12a

predictors, Month X1…Xp TCRU, TRC(44 TCLI, el ,lat, January kyr BP), TRC(21 lon, st, sl kyr BP) February March April TCLI, el, lat, May lon, ic June July August September TCLI, el, lat, October lon, st, sl November December

99.1 99

0.99 2.6 0.99 3.2

0.99 1.4 0.99 1.6

T1b T2b T3b T4b T5b T6b T7b T8b T9b T10b T11b T12b

TCRU -

100 100 100 100 100 100 100 100 100 100 100 100

0.98 0.98 0.98 0.98 0.97 0.98 0.97 0.97 0.98 0.98 0.98 0.98

0.99 0.99 0.99 0.99 0.99 0.98 0.99 0.99 0.99 0.99 0.99 0.99

T2a T3a T4a T5a T6a T7a T8a T9a T10a

TCLI,el,lat,lon -

January February March April May June July August September October November December

GCV Cor

0.5 0.4 0.2 0.2 0.2 0.3 0.3 0.3 0.1 0.2 0.3 0.4

0.6 0.5 0.4 0.4 0.4 0.5 0.5 0.3 0.3 0.3 0.5 0.6

89

Table A2.2. Abbreviations for Tables A2.1 and A2.3. Abbreviations T P RC CRU CLI el lat lon st sl %ED ev GCV

Monthly mean surface air temperature Monthly mean total precipitation model output or setup of the RCA3 model data of the Climate Research Unit model output of the CLIMBER model elevation of the grid point latitude of the grid point longitude of the grid point direction of the steepest slope of the grid point angle of the steepest slope of the grid point percentage of the explained deviance the explained variance the General Cross Validation Score

Table A2.3. Summary of configurations and skill scores of monthly log precipitation GAM calibrations. Values in parenthesis are for precipitation, other for log precipitation. The hyphen “-” means “same as above”. Abbreviations are explained in Table A2.2. GAM

predictand, Y

P1

log(PCRU, PRC(44 kyr BP), PRC(21 kyr BP)) -

P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12

predic- Month tors, X1…Xp PCLI, el, January lat, lon, st

%ED

ev

GCV

Cor

83.6

0.82 (0.65)

0.13

0.91 (0.88)

RMSE (mm/ month) (14)

-

81.4 76.4 64.6 65.3 67.7 70.5 64 67.7 74.9 77.1 79.8

0.80 (0.62) 0.75 (0.52) 0.62 (0.45) 0.63 (0.54) 0.66 (0.57) 0.69 (0.64) 0.62 (0.59) 0.66 (0.56) 0.73 (0.60) 0.76 (0.65) 0.79 (0.59)

0.14 0.14 0.13 0.09 0.08 0.08 0.08 0.11 0.13 0.17 0.15

0.90 (0.87) 0.87 (0.83) 0.80 (0.70) 0.80 (0.77) 0.82 (0.76) 0.84 (0.79) 0.80 (0.76) 0.82 (0.81) 0.86 (0.81) 0.88 (0.84) 0.89 (0.86)

(11) (13) (14) (14) (14) (16) (18) (21) (22) (18) (18)

February March April May June July August September October November December

In Figure A2.1 we show how the fitted GAM works for the January surface temperature. The fitted GAM is able to approximately reproduce the monthly mean observed present-day temperatures and those simulated for 21 kyr BP and 44 kyr BP, with the explained deviance of 99 % and correlation 0.99 (see Table A2.1). We fitted

90

another GAM for January precipitation. The performance of this GAM is shown in Figure A2.2. The fitted GAM is able to reproduce moderately well the monthly total precipitation of observed present-day climate and simulated 21 kyr BP and 44 kyr BP climate over most of Fennoscandia with the explained deviance of 83.6 % and correlation 0.91 for log-precipitation (correlation for precipitation 0.88, see Table A2.3). Over the Scandinavian mountains, however, the GAM seems to underestimate the precipitation. The monthly mean surface temperature and total precipitation produced by the GAM fitted for July is given in Figure A2.3 and Figure A2.4. For the July monthly mean temperature GAM the explained deviance was 95.2 % and correlation 0.98, see Table A2.1. For the July monthly total log-precipitation GAM the explained deviance was 70.5 % and correlation 0.84 (correlation for precipitation 0.79, see Table A2.3).

Figure A2.1. January temperature in the global model, observations/regional model, statistical model and residual of statistical model-observations/regional model. Unit: Celsius.

91

Figure A2.2. January total precipitation in the global model (mm/month), observations/regional model (mm/month), statistical model (mm/month) and residual of statistical model-observations/regional model (%).

Figure A2.3. July temperature in the global model, observations/regional model, statistical model and residual of statistical model-observations/regional model. Unit: Celsius.

92

Figure A2.4. July total precipitation in the global model (mm/month), observations/ regional model (mm/month), statistical model (mm/month) and residual of statistical model-observations/regional model (%).

Figure A2.5 displays the performance of the monthly GAMs for surface temperature and total precipitation compared to a) the observed present-day climate, b) the RCA3modelled permafrost climate, and c) the RCA3-modelled glacial climate in the grid point nearest to Olkiluoto. The surface temperatures produced with the fitted GAMs are in good agreement with the observed and modelled surface temperatures. The error in the GAM produced annual mean temperature is in the present-day climate circa -0.1 ºC (with a maximum overestimation in September: 0.6 ºC), in the permafrost climate about 0.5 ºC (with a maximum overestimation in September: 1.7 ºC), in the glacial climate about 0.3 ºC (with a maximum overestimation in September: 1.7 ºC).

93

Temperature (Celsius)

Precipitation (mm/month)

a) present climate --- CRU(present) --- CLI(present) --- GAM(present)

b) permafrost climate --- RC(44 kyr BP) --- CLI(44 kyr BP) --- GAM(44 kyr BP)

c) glacial climate --- RC(21 kyr BP) --- CLI(21 kyr BP) --- GAM(21 kyr BP)

Figure A2.5. Comparison of the direct output of the CLIMBER model (blue), GAM downscaled (red) and the observed/RCA3modelled (green) temperature and precipitation over Olkiluoto grid in a) the present day, b) the permafrost and c) the glacial climates. Note different y-axis in the a) figure. The monthly precipitation rates are also well reproduced over Olkiluoto with our GAMs, which are able to reproduce the seasonal cycle of precipitation with largest precipitation in late summer and autumn (Figure A2.5, right side). The annual total precipitation is overestimated in the present-day climate with 11 % with maximum overestimation of about 24 % in October and February. In the permafrost climate the

94

annual precipitation is overestimated by 4 %, with maximum overestimation of circa 20 % in the November precipitation (much of this is compensated with the June underestimation of circa 20 %). In the glacial climate over Olkiluoto the annual precipitation is underestimated by 19 % with largest underestimation of circa 28 % in April precipitation. For downscaling the temperature of the 120 kyr into the future simulation with the 400 ppm CO2 concentration, we used monthly GAMs fitted to the observed temperature data of the CRU. The performance of these GAMs compared to CRU climate in the Olkiluoto region are displayed in Figure A2.6. The surface temperatures produced with the fitted GAMs are in good agreement with the observed surface temperatures. The error in the fitting for Olkiluoto grid point was 0.03 ºC for the annual mean temperature with a largest seasonal underestimation of 0.7 ºC in May temperature. The skill scores in Table A2.1 of the temperature GAMs used for the downscaling the 400 ppm simulation into the future are better than those used for the Last Glacial Cycle and 280 ppm simulations. This is expectable, since the performance range of the 400ppm GAMs cover only interglacial climate.

--- TCRU(present) --- TCLI(control) --- TGAM(present)

Figure A2.6. The performance of the monthly temperature GAMs used to downscale the temperature of the 400 ppm simulation. Comparison of the direct output of the CLIMBER model (blue), GAM downscaled (red) and the observed (green) temperature over Olkiluoto grid in the present day climate.

GAM output comparison to temperature and precipitation reconstructions We compared the predictions of the GAMs in Table A2.1 and Table A2.3 to pollenbased reconstructions available for: Toskaljavri (Seppä & Birks 2002), Tsuolbmajavri (Seppä & Birks 2001), Flarken (Seppä et al. 2005), Arapisto (Sarmaja-Korjonen & Seppä 2007), Laihalampi (Heikkilä et al. 2003) and Gilltjärnen (Antonsson et al. 2006). In figure A2.7 we show the grid points nearest to these locations. The CLIMBER model’s Fennoscandian grid represents the area of 60º N to 70º N and 9.3º W to 42.1º E. In figures A2.8 – A2.9 we show CLIMBER and GAM temperature and precipitation

95

predictions together with the reconstructions for these locations. In Figure A2.8, for the Northern locations Toskaljavri and Tsuolbmajavri, both CLIMBER and GAM predictions are mostly close to the reconstructed July mean temperature and annual precipitation. The only exception is 11 kyr BP, when the ice has recently (during the last millennia) retreated from Northern Finland and both GAM and CLIMBER predictions are too cold. In Figure A2.9 for Gilltjärnen, Laihalampi, Arapisto and Flarken, the CLIMBER prediction is clearly too cold and the the downscaling with GAMs are capable to bring the annual mean temperature predictions closer to the reconstructed temperatures.

Figure A2.7. A map with the nearest land grid points to Olkiluoto, Toskaljavri, Tsuolbmajavri, Flarken, Arapisto, Laihalampi and Gilltjärnen.

96

July mean Temperature (Celsius)

Annual Precipitation (mm)

July mean Temperature (Celsius)

Annual Precipitation (mm)

a) Toskaljavri

b)Tsuolbmajavri

Figure A2.8. Simulated July mean temperature and annual total precipitation of CLIMBER (blue) and GAM (red) and pollen-based reconstruction (green) for a) Toskaljavri and b) Tsuolbmajavri. The RMSEs (prediction errors) for the pollen-based reconstructed July mean temperature and annual mean temperature were c. 1.0 ºC and c. 340 mm, respectively (Seppä & Birks 2001; Seppä & Birks 2002).

97

Annual mean temperature (Celsius) a) Gilltjärnen

b) Laihalampi

c) Arapisto

d) Flarken

Figure A2.9. Simulated annual mean temperature of CLIMBER (blue) and GAM (red) and pollen-based reconstruction (green) for a) Gilltjärnen (RMSE for the reconstructed annual mean temperature c. 1.0 ºC (Antonsson et al. 2006)), b) Laihalampi (RMSE for the reconstructed annual mean temperature c. 0.9 ºC (Heikkilä & Seppä 2003)), c) Arapisto (RMSE for the reconstructed annual mean temperature c. 0.9 ºC (SarmajaKorjonen & Seppä 2007)), and d) Flarken (RMSE for the reconstructed annual mean temperature c. 1.0 ºC (Seppä et al. 2005)).

98

References for Appendix 2 Antonsson, K., Brooks, S.J., Seppä, H., Telford, R.J. & Birks, H.J.B. 2006. Quantitative palaeotemperature records inferred from fossil pollen and chironomid assemblages from Lake Gilltjärnen, northern central Sweden. Journal of Quaternary Science, 21, 1-11. Heikkilä, M. and Seppä, H. 2003. A 11,000 yr palaeotemperature reconstruction from the southern boreal zone in Finland. Quaternary Science Reviews, 22, 541-554. Sarmaja-Korjonen, K. & Seppä, H. 2007. Abrupt and consistent responses of aquatic and terrrestrial ecosystems to the 8200 cal yr BP cold event - a lacustrine record from Lake Arapisto, Finland. The Holocene, 17, 457-467. Seppä, H. and Birks, H.J.B. 2001. July mean temperature and annual precipitation trends during the Holocene in the Fennoscandian tree-line area: Pollen-based climate reconstructions. Holocene, 11, 527-537. Seppä, H. and Birks, H.J.B. 2002. Holocene Climate Reconstructions from the Fennoscandian Tree-Line Area Based on Pollen Data from Toskaljavri. Quaternary Research, 57, 191-199. Seppä, H., Hammarlund, D. & Antonsson, K. 2005. Low and high-frequency changes in temperature and effective humidity during the Holocene in South central Sweden: Implications for atmospheric and oceanic forcings of climate. Climate Dynamics, 25, 285-297.

99

APPENDIX 3

DELIVERABLES AND DATA

These data were delivered within the project Future climate scenarios for Olkiluoto. The report related to the data is: Pimenoff, N., Venäläinen, A. & Järvinen, H. 2011. Climate scenarios for Olkiluoto on a time-scale of 120,000 years. Posiva 2011-04. Models and Scenarios Global model: CLIMBER-2-SICOPOLIS model system Three scenarios: 1. Last glacial cycle with CO2 concentration derived from the Vostok ice core data. 2. Simulation of 120 kyr into the future with constant atmospheric CO2 concentration of 280 ppm. 3. Simulation of 120 kyr into the future with constant atmospheric CO2 concentration of 400 ppm. Regional model: RCA3 Two climate states: 1. Permafrost case: a stadial at 44 kyr BP during the Marine Isotope Stage 3. 2. Last glacial maximum case: the Last Glacial Maximum at 20 kyr BP. Statistical model to downscale the global model output into regional scale: The data of the CLIMBER-2-SICOPOLIS simulations were downscaled with a Generalized Additive Model (GAM). The GAM was built utilizing observations of present day climate (CRU) and results of the RCA3 regional model simulations. Resulting data The saved variables are listed below. The data are saved in text (.txt) for the grid nearest to Olkiluoto (Figure 17). The data are available also for the other grid points in Fennoscandia. The time step of the saved data is 1,000 years.

100

Variable

Name

Unit

Resolution

Model

Air temperature 2m (no ice sheet)

t2m

Celsius

monthly

CLIMBER/GAM

Air temperature 0m (no ice sheet)

t0m

Celsius

monthly

CLIMBER/GAM

Basal temperature

basal.T

Celsius

annual

SICOPOLIS

Homologous basal temperature

h.basal.T.

Celsius

annual

SICOPOLIS

Ice sheet surface temperature

ice.surf.T.

Celsius

annual

SEMI

Total precipitation

RR

mm/month

monthly

CLIMBER/GAM

Fraction of snowfall of total precipitation

snowfall.frac

%

monthly

CLIMBER

Vegetation

vegetation

grass/trees/ ice sheet

annual

CLIMBER

Ice sheet thickness

ice.thickness

meters

annual

SICOPOLIS

Basal height

basal.height

meters

annual

SICOPOLIS

Global mean sea level relative to the present

globalsealevel

meters

annual

SICOPOLIS

Solar insolation at 65 ºN

solar.insol.65N

Wm-2

monthly

CLIMBER

Wm-2

monthly

CLIMBER

Solar flux at the surface sol.f.surf.65N at 65ºN

101

APPENDIX 4

PUBLICATION RESULTING FROM THE PROJECT

Pimenoff, N., Venäläinen, A. & Järvinen H. in prep. Statistical downscaling of a last glacial cycle climate simulation: regional temperature and precipitation over Fennoscandia.

102

LIST OF REPORTS

POSIVA-REPORTS 2011 _______________________________________________________________________________________

POSIVA 2011-01 



An Update of the State-of-the-art Report on the Corrosion of Copper Under Expected Conditions in a Deep Geologic Repository (This report will also be printed as SKB TR-10-67) Fraser King, Integrity Corrosion Consulting Limited Cristina Lilja, Svensk Kärnbränslehantering AB Karsten Pedersen, Microbial Analytics Sweden AB Petteri Pitkänen, Marjut Vähänen, Posiva Oy ISBN 978-951-652-178-0 



POSIVA 2011-02

Olkiluoto Site Description 2011 Posiva Oy ISBN 978-951-652-179-7

POSIVA 2011-03

Effects of Bedrock Fractures on Radionuclide Transport Near a Vertical Deposition Hole for Spent Nuclear Fuel Veli-Matti Pulkkanen, VTT ISBN 978-951-652-180-3

POSIVA 2011-04

Climate Scenarios for Olkiluoto on a Time-Scale of 100,000 Years Natalia Pimenoff, Ari Venäläinen & Heikki Järvinen, Finnish Meteorological Institute ISBN 978-951-652-181-0