positional errors involved in the various independent data products. ...... direct application of remote sensing techniq
Chapter 12
Remote sensing and GIS applications in agrometeorology
This Chapter was written by Bruno Basso, Tim R. McVicar and Byong Lee, with contributions from Hojjat Yazdanpanah and H.P. Das
The set up was assisted by facilitators Giovanni Cafiero, Davide Cammarano, H.P. Das, Lorenzo De Simone, Guido D’Urso, Valentina Maddalena V.K. Sehgal, Nataraj Subash and Aymar Suleiman
The Chapter was externally reviewed by Bimal Kumar Bhattacharya and Andres Ravelo
The Chapter was coordinated by Kees Stigter
CONTENTS 12.1. Introduction to Remote Sensing Science 12.1.1. Reflective Remote Sensing 12.1.2. Thermal Remote Sensing 12.1.3. Microwave Remote Sensing 12.1.4. Earth Satellites 12.1.5. Weather Satellites 12.1.6. Derived products
12.2 Introduction to Geographic Information Systems 12.2.1 Storing geographical data 12.2.2. Raster format 12.2.3. Vector format 12.2.4. Requirements of a GIS 12.2.5. Basic components of GIS
12.3. Integration of Remote sensing and GIS 12.3.1. Data Integration 12.3.2. Data Acquisition 12.3.3. Data Processing 12.3.3. 1. Geometric Rectification 12.3.3. 2. Data Conversion 12.3.3.3. Data Analysis 12.3.3.4. Classification System 12.3.3.5. Data Generalization 12.3.3.6. Error Assessment 12.3.3.7. Sampling 12.3.3.8. Spatial Autocorrelation 12.3.3.9. Locations Accuracy 12.3.3.10. Final Product Presentation Error 12.3.3.11. Decision Making 12.3.3.12. Implementation 12.3.3.13. Theory of Vegetation Indices
12.4. Operational Agrometeorological Products Employing Remote Sensing and GIS 12.4.1. Assessment of Meteorological and Agronomic Conditions to Aid Decision on Drought using Remote Sensing and GIS
12.4.1.1. Precipitation 12.4.1.2. Solar Radiation 12.4.1.3. Agronomic Conditions 12.4.2. Operational Use of Remote Sensing and GIS for Irrigation Scheduling 12.4.3. Remote Sensing for Soil and Crop Management 12.4.3.1. Evaporation 12.4.3.2. Soil Salinity 12.4.3.3. Remote Sensing for Precision Agriculture 12.4.3.4. Crop Growth and Intercepted Radiation 12.4.3.5. Nutrient Management 12.4.3.6. Pest Management 12.4.3.7. Selection of Growth Traits 12.4.3.8. Crop Yield Estimation 12.4.4. Assessing Environmental Sensitive Areas for Desertification Risk through Remote Sensing and GIS 12.4.5. Aquaculture and Remote Sensing 12.4.6. Operational use of Remote Sensing for Identification of Fishing Zones 12.4.7 . Forest Management through Remote Sensing
12.5 Collaborations for resource sharing in GIS and Remote Sensing
12.1
Definition and role of Remote Sensing
Remote sensing is the science and art of obtaining information about an object through the analysis of data acquired by a device that is not in contact with the object (Lillesand and Keifer, 1994). Remotely sensed data can be of many forms, including variations in force distribution, acoustic wave distribution or electromagnetic energy distributions and can be obtained from a variety of platforms, including satellite, airplanes, remotely pilot vehicles, handheld radiometers or even bucket trucks. They may be gathered by different devices, including sensors, film camera, digital cameras, video recorders. Our eyes acquire data on variations in electromagnetic radiations. Instruments capable of measuring electromagnetic radiation are called sensors. Sensors can be differentiated in two main groups: Passive sensors: without their own source of radiation. They are sensitive only to radiation from a natural origin. Active sensors: which have a built in source of radiation. Examples are Radar (Radio detection and ranging) and Lidar (Light detection and ranging) systems. This can be analogue (photography) or digital (multispectral scanning, thermography, radar). The elements of a digital image are called resolution cells (during the data acquisition) or pixels (after the image creation). The implementation of remote sensing data by the user requires some knowledge about the technical capabilities of the various sensor systems. The technical capabilities of the sensor systems can be listed in three resolutions: •
Spatial resolution:
concerns the size of the resolution cell on the ground in the direction of the flight and
across. The size of the pixel determines the smallest detectable terrain feature. •
Spectral resolution: concerns the number, location in the electromagnetic spectrum and bandwidth of the specific wavelength bands or spectral bands. This resolution differs between sensors and largely determines their potential use.
•
Temporal resolution: concerns the time lapse between two successive images of the same area. This primarily determined by the platform used, and secondly by the atmospheric conditions.
Remote Sensing provides spatial coverage by measurement of reflected, emitted and backscattered radiation, across a wide range of wavebands, from the earth’s surface and surrounding atmosphere. Remote Sensing-based spatial coverage was measured by reflected and emitted electromagnetic radiation from the earth’s surface and surrounding atmosphere. Remote sensed data of the land surface is possible across a wide range of wavebands, from the ultra-violet (UV), visible (VIS), near infrared (NIR), short wave infrared (SWIR), mid-infrared (MIR), thermal infrared (TIR), and microwave (MV) regions of the electromagnetic spectrum. They are located in so called ‘atmospheric windows’ as there is a signal from the surface – there is not total absorption (or scattering) of the light due to atmospheric constituents. Each waveband provides different information about the atmosphere and land surface. Clouds, rainfall, surface temperatures, temperature and humidity profiles, solar and net radiation, and the fundamental processes of photosynthesis and evaporation can affect the reflected and emitted radiation detected by satellites. The research challenge is to develop models that can be inverted to extract relevant and reliable information from remotely sensed data, providing final users near-real-time information.
In general the possible tools of remote sensing technique can be grouped into three main categories, namely, satellite, radar, and near-to-surface instruments. The platform for remote sensing can be either fixed or moving, terrestrial or operating from different altitudes, and be either manned or unmanned. Considering the operating time, the platform can be classified as temporary, semipermanent or virtually permanent. These aspects are important in order to understand the quality and quantity of the information available to the agrometeorological service. The distance of the instruments affects directly the resolution and the precision of the information. The resolution of observation can vary from a few square meters, with a scanner mounted on a vehicle, to continental scale, using a meteorological satellite. Due to the large volumes of data generated, an increase in temporal resolution is usually at the expense of spatial and spectral resolution. Relevant to agricultural meteorology are the earth resource satellites with high spatial resolution and the meteorological satellites with high temporal resolution.
12.1.1. Reflective Remote Sensing The reflective portion of the Electromagnetic Spectrum (EMS) ranges nominally from 0.4 to 3.75 µm. Light of shorter wavelength than this is termed ultraviolet (UV). The reflective portion of the EMS can be further subdivided into the visible (0.4--0.7 µm), near-infrared (NIR) (0.7--1.1 µm), and shortwave-infrared (SWIR) (1.1--3.75 µm). Remote sensing converts an analogue photon flux to digital images, where the number of quantisation levels is a function of the number of bits used to represent the photon flux. The number of quantisation levels equals two to the power of the number of bits. That is, 7-bit data provide 128 (27) levels of quantisation, 8-bit data 256 (28), 10-bit data 1024 (210) and 12-bit data 4096 (212). The ability of remote sensing measurements to distinguish different properties of the Earth’s surface in the EMS is partly determined by the level of quantisation. Many remote sensing instruments have channels situated in the red and NIR wavelength of the spectrum.
These two reflective bands are often combined to produce vegetation indexes. The most common
linear combinations are the simple ratio (NIR/red) and normalised difference vegetation index (NDVI), which is derived from (NIR--red)/(NIR+red). Several publications (Wiegand et al., 1991; Kaufman and Tanre 1992; Thenkabail et al. 1994; Leprieur et al. 1996; Penuelas and Filella, 1998) provide comprehensive listings of vegetation indexes. There are positive correlations between NDVI and factors such as plant condition (Sellers 1985); foliage presence (including leaf area index (LAI)) (Curran et al. 1992; Tucker 1979; Nemani and Running 1989; McVicar et al. 1996a; 1996b; 1996c); and per cent foliage cover (Lu et al. 2001). Based on the relationship between LAI and NDVI, many previous studies have modelled crop yield by integrating the area under the NDVI curve (denoted
∫ NDVI ) for all or part of the growing season (Rasmussen 1998a; Rasmussen 1998b;
McVicar et al. 2002; Honghui et al. 1999). Some researchers have used regression-based models developed from NDVI data acquired at specific times during the growing season (Maselli et al. 1992; Smith et al. 1995). The amount of green leaf is one determinant of the signal strength in the reflective portion of the EMS. Several
other important factors that affect the acquired value, including both soil colour and sun-target-sensor geometry. The following discussion draws heavily on Roderick et al. (2000), whose work elegantly illustrated the effect of soil colour and quantisation on overall reflectance, and showed how change can be distinguished using remotely sensed measurements. The authors assumed a Lambertian surface (one that has no angular dependency), with no shade present and two soils, one dark (5% albedo1) and the other bright (30% albedo), with green grass (10% albedo) in the red portion of the EMS. On a dark soil, a 20% increase in green grass results in an overall 1% increase in albedo; on a bright soil it results in a 4% decrease. Hence, the albedo of the elements that make up the scene and their relative proportions are important to the overall albedo. For dark soils sensed with a 7-bit instrument, a 15% change in cover is needed to change the recorded digital number by 1. This reduces to 0.5% for a 12-bit system. At the same level of quantisation, a larger relative difference between end-members require a smaller changes in per cent grass cover to detect change. This is because there is a greater difference in the albedo of the end-members when looking at bright soils (20% difference), than when considering the dark soil (5% difference) example. In reality, the land surface is not a perfect Lambertian reflector (Liang and Strahler 2000) --- most surfaces in the optical (reflective and thermal) portions of the EMS are strongly anisotropic2. The geometry of sun, target and sensor, and the size, shape and spacing of elements (e.g. trees over bare ground) controls the amount of shadow contributing to the signal (Hall et al. 1995). This effect, termed the bidirectional reflectance distribution function (BRDF) (Burgess and Pairman 1997; Deering 1989), is characteristic of vegetation structure in reflective remotely sensed images. This means that in addition to the albedo of the scene elements and their relative proportions, the spatial distribution and the pixel size of observation affect the signal measured by remote sensing instruments (Jupp 1989b; Jupp 1989a; Woodcock and Strahler 1987; Walker et al. 1986; Jupp and Walker 1996). In addition, the match between the operational scale and the resolution partly controls the strength of relationships between surface properties and remote sensing (Friedl 1997). Image information content is minimal when the resolution is approximately equal to the operational scale (Woodcock et al. 1988b, Woodcock et al. 1988a). Other factors affecting the reflective portion of the EMS include changes in the observed signal due to changes in the atmospheric component of the signal, including atmospheric precipitable water (Choudhury and DiGirolamo 1995; Hobbs 1997), and atmospheric aerosols, changes in the response of the sensor over time (Mitchell 1999).
12.1.2. Thermal Remote Sensing The thermal portion of the EMS ranges nominally from 3.75 to 12.5 µm. The observed surface temperature is a function of the radiant energy emitted by the surface which is remotely sensed, be it land, ocean or cloud top. Models have been developed to allow surface temperature to be extracted from thermal remote sensing data. Prata et al. (1995) review the algorithms and issues, including land emissivity estimation, involved in the calculation of land-surface temperature, denoted Ts. 1
Albedo is the fraction of light that is reflected by a body or surface
2
Anisotropic: having different physical properties in different directions
Thermal remote sensing is an observation of the status of the surface energy balance (SEB) at a specific time of day. The SEB is driven by the net radiation at the surface. During the daytime, the net radiation is usually dominated by incoming shortwave radiation from the sun, the amount reflected depending on the albedo of the surface. There are also longwave components that well up and down. At the ground surface, the net all-wave radiation is balanced between the sensible, latent and ground heat fluxes. The ground heat flux averages out over long periods of time; thus, the SEB represents the balance between the sensible and latent heat fluxes. During the day, Ts is partially dependent on the relative magnitude of the sensible and latent heat fluxes (Quattrochi and Luvall 1999; McVicar and Bierwirth 2001; McVicar and Jupp, 1999, 2002). Remotely sensed thermal data are also recorded at night. At this time, the SEB is dominated by the release from the ground of heat that was absorbed during daylight hours, which is governed by how much heat was absorbed during the day and the rate at which it is released after sunset. The environment’s capacity to store heat depends on the amount of water it is storing.
12.1.3. Microwave remote sensing
The microwave portion of the EMS ranges nominally from 0.75 to 100 cm. Radio signals have wavelengths that are included in these bands. These systems can either be active or passive. Both passive and active microwave observations have been used to determine the near-surface soil moisture for a number of experimental field sites. RADAR is an active system based upon sending a pulse of microwave energy and then recording the strength, and sometimes polarisation, of the return pulses. The way the signal is returned provides information to determine characteristics of the landscape. RADAR has been used in the determination of near-surface soil moisture. Earth’s surface using eight spectral bands in the reflective portion of the EMS. ERS was launched and is operated by the European Space Agency. RADARSAT is a synthetic aperture RADAR (SAR) on board a Canadian satellite.
12.1.4. Earth Satellites The major earth resource satellites are LANDSAT (USA) with wavebands in VIS, NIR, SWIR, and TIR, SPOT (France) with wavebands in VIS and NIR, and MODIS (USA) with viewing the entire Earth's surface every 1 to 2 days, acquiring data in 36 spectral bands, or groups of wavelengths. SPOT uses a push broom linear array of charged coupled device detectors which cover only the VIS and NIR spectral regions. For agricultural meteorological applications LANDSAT, SPOT and MODIS data are useful for characterizing the land surface.
12.1.5. Weather Satellites The remote sensing of weather and climate by polar-orbiting and geostationary satellites is regarded as the single
most significant breakthrough for monitoring the earth’s weather, climate and vegetation in the last quarter of a century. It is fortunate that these satellites when are not viewing clouds are recollecting data on terrestrial vegetation and ocean temperature. Hence they provide both meteorological and vegetation information, the two essential inputs into agricultural meteorology. The possible use of air and satellite borne radar will promote the remote sensing monitoring system during cloudy day. This part of remote sensing technique is only in the research phase. Two broad types of meteorological satellite are in common use. One is the NOAA polar orbiting satellites in low earth orbit of 750 km. The other is the Geosynchronous Meteorological Satellite (GMS) with an altitude of approximately 36000 km, launched by the USA, USSR, ESA and Japan, China and India to form a global atmospheric monitoring system.
12.1.6. Derived products All information produced by the satellite is elaborated for the extraction of the desired information (Table 1 and 2). There are many methods, algorithms and procedures to derive fundamental data for agrometeorological applications from remote sensing. Among the existing indices, the most extensively used are: - The sea surface temperature (SST) - The land surface temperature (LST) - The normalized difference vegetation index (NDVI) - Optimal index of current plant cover and its variation with time. INSERT TABLE 1 and 2 HERE
12.2. Introduction to Geographic Information Systems A Geographic Information System (GIS) can be defined in a broader sense as a collection of hardware, software, data, organizations, and professionals to represent and analyze geographic data. A GIS references real-world spatial data elements (also known as graphic or feature data elements) to a coordinate system. These features are usually can be separated into different thematic types (e.g. soils and meteorological data). A GIS can also store attribute data, which is descriptive information of the map features. This attribute information is placed in a database separated from the graphics data but linked to them. A GIS allows the examination of both spatial and attribute data at the same time.
Also, a GIS lets users search the attribute data and relate it to the spatial data,
and vice versa. Therefore, a GIS can combine geographic and other types of data to generate maps and reports, enabling users to collect, manage, and interpret location-based information in a planned and systematic way.
In
short, a GIS can be defined as a computer system capable of assembling, storing, manipulating, and displaying geographically referenced data.
12.2.1. Storing geographic data GIS provides an organized method of storing spatial data. It stores the characteristics of features (attribute component) in a database, then links the attributes to features (spatial component) that it displays on a map. GIS software stores the information about each feature as a record in a table and organizes the attributes into columns, it displays the linked features as a theme in a view. This stored information (features and attributes) can then be manipulated, retrieved and analyzed using GIS methods and tools. A GIS stores sets of features and attributes as layers. However, all the layers it stores do not necessarily fit together. The layers may be stored in different formats, different map projections, and different resolutions. In each of these cases, the data sets won’t fit together or "line up." Managing spatial data includes converting data sets into formats that can be used by the GIS system, changing a map’s projection to match the projection of other maps and most importantly, documenting the data and processes, also known as metadata. A GIS provides several methods of retrieving information. One method is creating an attribute query. An attribute query is a set of criteria for a specific attribute. The database searches within a specified column for records that match the criteria you set, and retrieve those records. This type of query is a basic function of most databases. It takes some thought to design a query statement that retrieves (selects) only those records that you need, leaving behind (unselected) those that you do not need. While attribute query allows you to set criteria for a certain value or range of values for an attribute in the database, a spatial query selection allows you to retrieve records by selecting features on a map. Various tools are used to select the features on the map and through their linkage in the database, The records associated with the selected features are retrieved. Different GIS software packages have different methods of retrieval. Regardless, query is a very useful tool that allows the user to retrieve the information he has previously stored. Spatial data may be stored in either raster or vector formats with a GIS. In principle, both can be used to represent point, area and continuous spatial objects, but in practice raster formats tend to be used to represent continuous spatial objects and vector formats are used to represent point and area objects.
12.2.2 Raster format In a raster representation, geographic space is divided into an array of cells (a matrix), as shown in Figure 1a. All geographic representation is then expressed by assigning attributes to these cells. These cells are sometimes called pixels. The colours of cells in the figure represent different values on a nominal scale. One of the commonest forms of raster data comes from remote sensing satellites. (Pixel sizes from satellitederived data vary from 5 m resolution to over 1 km resolution – with the area of pixels ranging from 25 m2 to over 1 km2). When information is represented in raster format each component cell is assigned a single attribute value and all detail about variation within the component cell is lost. When creating raster data several rules may be applied to
specify how a cell will be coded: in most situations the attribute with the largest share of the cell’s area gets the cell attribute value. In other circumstances the rule is based on the central point of the cell and the attribute value of the central point is assigned to the cell. Although the largest share rule is almost always preferred, the central point rule is commonly used because it is quick to calculate. 1.2.2.3 Vector format
In a vector representation, all objects are represented as points connected by straight lines. An area is captured as a series of points or vertices connected by straight lines, as shown in Figure 1b. The use of straight lines explains why areas in vector representations are often called polygons. To capture an area object in vector form, you need only specify the locations of the points that form the vertices of the polygon area. A raster representation, on the other hand, would require a listing of all the cells that form the area. To create a close approximation of an area in raster format it would be necessary to resort to using very small cell sizes and the number of cells would increase proportionally with the level of detail required. In some circumstances the precision of vector representations might be crude, since many geographic phenomena cannot be located with high accuracy. In these cases raster representations may be a more honest representation of the inherent quality of the data. Also, various methods exist for compressing raster data that can greatly reduce the capacity needed to store a given data set. Table 1 summarizes the features of raster and vector representations in a GIS.
FIGURE 1a. Examples of raster data format in GIS
FIGURE 1b. Example of Vector data format in GIS
Table 3 Features of raster and vector representations in a GIS
Factor
Raster
Vector
Source
Remote sensing
Printed maps, digitized maps
Resolution
Fixed
Variable
Volume of data
Depends on size of cells used and
Depends on level of detail
spatial extent of study area Applications
Environmental applications
Social, economic, administrative
Ease of calculation
Efficient
Relatively inefficient
Numerous thematic surveys in the form of soil survey reports, climatic maps, groundwater surveys, land cartography etc. do exist. Integrated land and water resource information systems, based on GIS technology, will play a major role in linking multi-disciplinary, geographically referenced databases of different resolutions. Digital Elevation Models (DEMs) play an increasingly important role in this integration: topographic information on a grid basis, that is three-dimensional representation of the land, started from contour lines. This is a basic information layer in Agrometeorology for GIS applications (Fig. 2)
FIGURE 2. Triangulated Irregular Network (TIN), Digital Elevation Model (DEM) and remotely sensed DEM.
12.2.3 Requirements of GIS In a GIS all the information can be linked and processed simultaneously, obtaining a syntactical expression of the changes induced in the system by the variation of a parameter. There are two types of archives: static and dynamic. This technology allows the contemporary updating of geographical data and their relative attributes, producing a fast adaptation to the real conditions and obtaining answers in near real-time. The system requires preliminary basic information that, in the agrometeorological sector, is often furnished by the historical archives of different disciplines: geography, meteorology, climatology, agronomy, etc. Data import requires time and attention, mainly because this information will provide the basic knowledge of the territory and on the individual parameters, and it is difficult to modify in a second time. A realistic reproduction of the terrain elevation, allows many other parameters, like hydrology, sunshine duration, etc. This layer is basic to all the others, especially for the agrometeorologist. Other elements can be introduced, such as text, photographs, film, etc, to complete the informative layer.
12.2.4. Basic components of a GIS The function of an information system is to improve a user’s ability to make decisions in research, planning and management. An information system involves a chain of steps from the observation and collection of data through their analysis to yield information to their use in some decision making progress. In this context, a GIS may be viewed as a major sub-system of an information system. A computer-based GIS may itself be viewed as having five component sub-systems, including a) Data encoding and input processing b) Data management c) Data retrieval d) Data manipulation and analysis e) Data display.
12.3 Integration of Remote Sensing and GIS Remote sensing technology would not be complete without merging with GIS. Remote sensing represents a technology for synoptic acquisition of spatial data and the extraction of scene-specific information. Demand for remote sensing as a data input source to spatial database development has increased tremendously during the last few years. RS derived products are particularly attractive for GIS database development because they can provide cost-effective, wide area coverage in a digital format that can be directly input into a GIS. Because RS data are typically collected in a raster data format, the data can be cost-effectively rectified or converted to a vector format for subsequent spatial data analysis or modeling applications. A leading agrometeorological weather service, using advanced data collection and analysis tools like remote
sensing and GIS, must be equipped with sophisticated devices, but above all must have efficient and trained staff. In developing countries, there remains a risk that sparse high level agrometeorological personnel and sparse money are geared towards modern specializations for which the absorption capacity of their products is too small with agricultural decision makers. Problems and priority agrometeorological services need to be defined first. Methodologies come second but will be essential if properly made available and applied Our understanding of associated data processing errors, especially for integrating multiple spatial data sets, lags far behind. It is necessary to clearly identify the types of error that may enter into the process, understand how the error propagates throughout the processing flow, and procedures need to be developed to better quantify and report the error using standardized techniques. Performing spatial data analysis operations with data of unknown accuracy, or with incompatible error types, will produce an output product of low confidence and limited use in the decision making process.
12.3.1 Data integration Remote sensing and GIS error associated with the data acquisition, processing, analysis, conversion, and final product presentation can have a significant impact on the confidence of decisions made using the data. The process of integrating remote sensing data into a GIS usually includes the following analytical procedures: data acquisition, data processing, data analysis, data conversion, error assessment, and final product presentation. Error may be transferred from one data process step to the next and manifest in the final product, accumulate throughout the process in an additive or multiplicative fashion, or an individual process error can be overshadowed by other errors of greater magnitude. In theory, the amount of error entering the system at each step can be estimated. In practice, however, error is typically only assessed at the conclusion of data analysis (i.e. the final product), if it is assessed at all. Usually, the decision maker is provided graphic final products, statistical data, or modeling results with little or no information concerning the confidences of the information. This limits the confidence of the implemented decision(s).
12.3.2 Data Acquisition Some data acquisition errors are common to any form of data collection and may be introduced from a number of sources. Some of these sources cannot be controlled such as atmospheric conditions and the natural variability of the landscape which will result in mixed pixels (depending on pixel resolution). It is important to have an understanding of the type and amount of error possible from all data acquisition sources and to control it whenever possible. The processing of multiple data layers in a GIS database is predicated upon accurate spatial registration between data layers. Therefore, it is absolutely critical that all remotely sensed data be geometrically accurate and congruent with the GIS database. Illumination geometry can affect image quality and subsequent analyses. Ideally, illumination geometry is constant or nearly constant throughout an image. In practice, however,
acquisition needs dictate a relatively wide instantaneous field-of-view (IFOV), resulting in a range of illumination-measurement geometries
12.3.3 Data Processing
12.3.3.1.Geometric Rectification Simple polynomial-based algorithms have proven adequate for satellite imagery, where geometric distortions are minimal. Adaptive or discrete techniques such as finite element programs are required to remove the complex distortions that result from aircraft instability. If the geometry is not taken into account, this can lead to area estimation errors (Van Niel and McVicar, 2001). The geometric correction of digital remote sensor data usually involves some type of resampling, e.g., nearest neighbor, bilinear, cubic convolution . How these and other resampling algorithms affect the radiometric integrity of the data and its spatial appearance need to be more fully understood. Techniques to better automate or fine-tune geometric processing have been developed using different methods of multiple image spatial cross-correlation. However, broader application of these useful techniques requires development of more sophisticated image processing environments. Current software menu driven or “toolkit” approaches are too primitive and tedious for routine production processing. Additionally, producers of geo-corrected remotely sensed imagery need to assess the accuracy of the outputs against independent geographical features to determine if the imagery meets positional accuracy standards (Van Niel and McVicar, 2002).
12.3.3.2. Data Conversion Processing of spatial data in image processing often involves some form of data conversion. It is possible to resample the data to such a degree that the geometric and radiometric attributes of the resampled data have a poor relationship with the original data. A good example of resolution degradation during data conversion is when remotely sensed data are classified and then spatially filtered in order to increase classification accuracy. Once filtered, the spatial precision of resulting products may be reduced from that of the original measurements. Similarly, in GIS analysis of slope and aspect calculated from digital elevation models, the resulting value is representative of a neighborhood rather than being directly relatable to an individual pixel. These types of data conversions must be catalogued, studied, and their cumulative impact quantified when incorporated into GIS. It is inevitable that data would need to be converted between raster and vector formats. Raster format is simply data arranged as regularly spaced, equal sized grids. Satellite data and digital elevation models are common examples of raster data. These data are easily stored in a computer as a matrix of numbers. Vector data are more complex than raster data. Vector data maintain the true shape of a polygon using a series of arcs and nodes. Vector data are more aesthetically pleasing and are the preferred methods of data display for most GIS thematic maps containing polygons. Unfortunately, there can be significant error introduced either by converting from raster to vector format or from vector to raster format. The amount of this error depends on the algorithm used in the conversion process, the complexity of features, and on the grid cell size and orientation used for the raster representation. Failure to
consider this potential error can introduce considerable problems into any analysis.
12.3.3.3 Data Analysis In RS/GIS processing flow, data analysis involves the exploration of relationships between data variables and the subsequent inferences which may be developed. This stage of error accumulation will focus on the validity of statistical techniques. Difficulties in statistical analysis of spatially based environmental data sources involve the typical assumptions of the general linear model, compounded by the effects of spatial autocorrelation. Data analysis will also be subject to errors arising from variability in analyst expertise. Agrometeorological data commonly violate assumptions of independence for measured parameters and error variance. The tendency of adjacent or nearly adjacent samples to have similar values in environmental datasets, i.e., spatial autocorrelation, may violate the independence of samples required in classical statistics. This may result in underestimated sample variance and inflated confidence estimates. The effects of spatial autocorrelation in remotely sensed data sources should be considered. Flexible statistical tools need to be identified to take into account the particular difficulties of spatial environmental data sets and organized into a usable software environment. This would encourage adequate consideration of statistical assumptions in the development of more accurate information products. In addition to statistical validity, the classic problem in GIS-based data analysis of misregistered polygon boundaries continues to plague us. Registration error might be seen as being somewhat distinct from the positional errors involved in the various independent data products. It is imperative that the temporal nature of remotely sensed phenomena be catalogued and judgment be made concerning the optimum time period during which they are collected and their degree of longevity (Van Niel and McVicar, 2004a).
12.3.3.4. Classification System Thematic data layers created using remote sensing data generally require the use of some type of classification system to facilitate categorization of the data for subsequent GIS spatial data analysis. Some of the potential sources of classification system induced errors are: the inability of classification systems to categorize mixed pixels, transition zones or dynamic systems; poorly defined or ambiguous class definitions; human subjectivity; and the lack of compatibility among different classifications systems used with both remote sensing and traditional data types. Classification system induced error is of particular significance when dealing with natural systems. In a mixed pixel and transition or dynamic process situations, it is particularly important that detailed field verification data be collected to adequately describe the variation within a system to minimize classification system related error. The problem of poorly defined or ambiguous class definitions is a common problem that often introduces an element of error. Inconsistency in classification schemes can cause serious problems, rendering certain thematic coverages unusable in combination (Van Niel et al., 2005; Van Niel and McVicar, 2004a). Additionally, to optimize the information content when classifying agricultural systems with no mixed cropping (i.e., each field is planted with one crop during one growing season) GIS boundaries (considered ‘fine-scale’ vector data) can be combined with outputs from classified remotely sensed imagery (considered ‘coarse-scale’ raster data) to
improve the overall classification accuracy (Van Niel and McVicar, 2003).
12.3.3.5. Data generalization Data generalization is routinely performed during remote sensing analysis for two purposes; spatial resolution and spectral or thematic data reduction. Spatial generalization involves pixel resampling prior to analysis and resampling or grouping after analysis to meet a minimum map unit. Resampling to a spatial resolution finer than the original data commonly results in substantial error. Spectral generalization may be performed by filters which either enhance certain features such as edges, or homogenize similar pixels. Since filters alter the original pixel values, errors such as accurate location of edges or loss of spectrally similar yet unique resources may occur. It is also common to resample a classified data set to a minimum map unit. With the recent trend of transferring raster-based remotely sensed data into a vector-based GIS, it is important to minimize the number of polygons which must be created in the vector form. Generalization of this form may result in inaccurate boundaries and the inclusion of small resources within a larger area resource class.
12.3.3.6 Error Assessment Ideally an error assessment is performed after each phase of the analysis. However, project funds and schedules rarely provide the opportunity to perform such a thorough error assessment. Typically in remote sensing projects, error assessments are only performed after completion of data analysis, and usually only address thematic and locational accuracy. Locational accuracy typically refers to how well the georeferencing algorithms correctly placed pixels into a map coordinate projection, and not the accuracy of thematic or class boundaries. Most assessments were derived from the same data used to train the classifier. Training and testing on the same data set results in overestimates of classification accuracy. Rigorous guidelines must be developed to insure that these fundamental non-spatial specific error assessment problems do not continue.
12.3.3.7. Sampling Sample size is an important consideration when assessing the accuracy of remotely sensed data which are to be used in a GIS. Each sample point collected is expensive and therefore sample size must be kept to a minimum and yet it is important to maintain a large enough sample size so that any analysis performed is statistically valid. Sampling scheme is also an important part of an accuracy assessment. Selection of the proper scheme is critical to generating an error matrix that is representative of the entire classified image. Poor choice in sampling scheme can result in significant biases being introduced into the error matrix which may over- or under-estimate the true accuracy. The opinions about the proper sampling scheme to use vary greatly and include everything from simple random sampling to stratified systematic unaligned sampling.
12.3.3.8. Spatial Autocorrelation Spatial autocorrelation occurs when the presence, absence, or degree of a certain characteristic affects the presence, absence, or degree of the same characteristic in neighboring units. This condition is particularly important in accuracy assessment if an error in a certain location can be found to positively or negatively influence errors in surrounding locations. Surely these results should affect the sample size and especially the sampling scheme used in accuracy assessment. Therefore, additional research is required to quantify the impact of spatially autocorrelated imagery or classification products when subjected to error evaluation procedures.
12.3.3.9 Location accuracy In remote sensing, locational accuracy may be reported as the Root Mean Square Error (RMSE) resulting from the georeferencing algorithms which rectify images to map coordinates. The RMSE is the square root of the squared errors mean and reflects the proportion or number of pixel, plus or minus, that the image control points differ from the map or reference control points. However, the RMSE does not truly reflect the locational accuracy of all pixels within an image; the RMSE only addresses the control points. The most accurate means of examining locational accuracy is too costly to implement; a ground survey with differential GPS.
12.3.3.10 Final Product Presentation Error The goal of most remote sensing/GIS investigation is to produce a product which will quickly and accurately communicate important information to the scientists or decision-maker. This product may take many forms including thematic maps and statistical tables. There used to be sources of geometric (spatial) and thematic (attribute) error in the final map products and statistical summaries. Geometric error in the final thematic map products may be introduced through the use of 1) different scale base maps over a region, 2) different national horizontal datum in the source materials, and 3) different minimum mapping units which are then resampled to a final minimum mapping unit. It is imperative that improved map legends be developed which include cartobibliograhic information on the geometric nature of the original source materials. This is the only way a reader can judge the geometric reliability of the final thematic map products. The highest accuracy of any GIS output product is only as accurate as the least accurate file in the data base. Thus although the final map may look uniform in its accuracy, it is actually an assemblage of information from diverse sources. It is important for the reader to know what these sources are through a thematic reliability diagram. There is a great need to standardize the design and function of thematic reliability diagrams. Fundamental cartographic design principles must be followed especially when constructing the class interval legends for thematic maps
12.3.3.11 Decision Making The decision maker is often presented with remote sensing/GIS derived map and/or statistical presentation products for use in the decision making process. In most situations, adequate information concerning the lineage of thematic data layers and associated thematic and geometric accuracies is not provided. Ideally, in addition to the above mentioned information, the decision maker needs an estimate of the overall accuracy and confidence of the data product used in the process. However, GIS map and statistical data are being used by decision makers with little to no knowledge of the potential sources of error, and no information concerning the accuracy and confidence level of final presentation products. It is imperative that the RS and GIS communities educate the decision makers to better understand the potential error sources associated with remote sensing/GIS data products. As the decision makers become more knowledgeable of the issues related to data accuracy and confidence, they will begin to demand data concerning consumer accuracy by provided with all final presentation products. Decision makers and data analysts can no longer work in isolation if the use of remote sensing and GIS technologies are to become data sources on which decisions are based.
12.3.3.12 Implementation Decisions based on data of substandard accuracy and/or inappropriate confidence levels, has an increased probability of resulting in incorrect implementation actions. The obvious implications of an incorrect implementation decision are an erroneous resource management action which can have serious consequences on the resource itself. As remote sensing/GIS derived products are increasingly being utilized as a basis for implementation decisions concerning resource management and regulatory issues, there is a high potential for an explosion in the number of litigation cases in the short to medium term. A major challenge to the remote sensing and GIS communities will be the ability to adequately defend the accuracy and reliability (confidence) of products used by decision makers in implementation processes..
12.3.3.13. Theory of Vegetation Indices RVI is the Ratio Vegetation Index (Jordan, 1969; Pearson and Miller, 1972). A common practice in remote sensing is the use of band ratios to eliminate various albedo effects. In this case the vegetation isoline converge at origin. Soil line has slope of 1 and passes through origin, it range from 0 to infinity. And it is calculated as follow:
RVI = ρNIR/ρred
(2)
NDVI is the Normalized Difference Vegetation Index (Kriegler, 1969; Rouse et al., 1973) and it is the common vegetation index referring to. This index can vary between -1 and 1. In this case vegetation isolines are considered to be convergent at origin and soil line slope is 1 and passed through origin. It is calculated as:
NDVI = ρNIR-ρred / ρNIR-ρred
(3)
VIs assume that external noise (soil background, atmosphere, sun and view angle effect) is normalized, but this assumptions is not always true. The relative percentage of sunlit, shaded soil and plants components is highly dependent upon the view angle. Qi et al. (1995) studied the effect of multidirectional spectral measurements on the biophysical parameter estimation using a modeling approach. When the bidirectional effect is transformed from reflectance domain into vegetation index domain, it could be reduced (Jackson et al., 1990; Huete et al., 1992) or increased (Kimes et al., 1985; Qi et al., 1994b), depending on the vegetation types and solar zenith angles. Qi (1995) suggested that when bidirectional effect is a major concern (NDVI/NDVIo > 1) it is better to use NIR rather than NDVI, and that bidirectional effect on vegetation indices must be quantified before a quantitative VI-LAI relationship can be used. The Green Normalized Vegetative Index (GNDVI) is a modification of the NDVI where the Red portion is substituted by the reflectance in the Green band (Gitelson et al., 1996). DVI is the Difference Vegetation Index, (Richardson and Everitt (1992), but appears as VI in Lillesand and Kiefer (1994). Vegetation isolines are parallel to soil line. Soil line has arbitrary slope, passes through origin, and index range is infinite.
DVI = ρNIR-ρred
(4)
PVI is the Perpendicular Vegetation Index (Crippen, 1990), and it is sensitive to atmospheric variation.
In this
case vegetation isolines are parallel to soil line. Soil line has arbitrary slope, passes through origin and the index range from -1 to 1. PVI = 1/√a2+1 (ρNIR - aρred -b)
(5)
Where a and b are the coefficient derived from the soil line: NIRsoil= a REDsoil + b. WDVI is the Weighted Different Vegetation Index (Clevers, 1988) and like PVI is sensitive to atmospheric variation (Qi et al., 1994). Vegetation isolines are parallel to soil line. Soil line has an arbitrary slope and passes through origin, vegetation index range is inifinite.
WDVI = ρNIR - aρred
(6)
Where a is the slope of the soil line. Huete (1988) proposed a Soil Adjusted Vegetation Index (SAVI) to account for the optical soil properties on the plant canopy reflectance. SAVI involves a constant L to the NDVI equation. The index range is
SAVI = ρNIR - ρred / (ρNIR+ρred+L) (1+L)
from -1 to +1.
(7)
The constant L is introduced in order to minimize soil-brightness influences and to produce vegetation isolines independent of the soil background (Baret and Guyot, 1991). This factor can vary from 0 to infinity and the range depends on the canopy density. For L=0 SAVI is equal to NDVI, for L tends to infinity, SAVI is equal to PVI. However for intermediate density L was found equal to 0.5. Huete (1988) suggested that there maybe two or three optimal adjustment factor (L) depending on the vegetation density (L=1 for low vegetation; L=0.5 for intermediate vegetation densities; L=0.25 for higher density). TSAVI is the Transformed Adjusted Vegetation Index (Baret et al., 1989), and it is a measure of the angle between the soil line and the vegetation isoline. The soil line has arbitrary slope and intercept. The interception between soil line and vegetation isoline occur somewhere in the third quadrant. Baret and Guyot (1991) have proposed an improving of the initial equation as follow: TSAVI = a(ρNIR - aρred –b) / [aρNIR + ρred – a b + χ (1 + a2)]
(8)
Where a and b are soil line parameters (slope and intercept of the soil line) and χ has been adjusted so as minimize background effect, and its value is 0.08. TSAVI values ranging from 0 for bare soil and is close to 0.70 for very dense canopies as reported from Baret and Guyot (1991). At 40% green cover, the noise level of the NDVI is 4 times the WDVI and almost 10 times the SAVI, corresponding to a vegetation estimation error of +/- 23% for the NDVI, +/- 7% cover for the WDVI, and +/2.5% for the SAVI. Therefore the SAVI is a more representative vegetation indicator than the other Vis, but an optimization of the L factor will further increase his value (Qi et al., 1994). Qi et al. 1994, developed a Modified Soil Vegetation Index (MSAVI). This index provide a variable correction factor L. Geometrically vegetation isolines don’t converge to a fixed point as SAVI, and soil line has not fixed slope and passes through origin. Correction factor is based on calculation of NDVI and WDVI as shown by equations 9 and 10:
MSAVI = ρNIR - ρred / (ρNIR+ρred+L) (1+L)
(9)
where L is calculated as follow:
L = 1 – 2 a * NDVI * WDVI
(10)
This term is computed to explain the variation of L among different types of soils, moreover L varies with canopy cover, and it’s range varies from 0 for very sparse canopy to 1 for very dense canopy. To further minimize the soil effect Qi et al. 1994, use an L function with boundary condition of 0 and 1 (Ln= 1-MSAVIn-1) and an MSAVI equal to:
MSAVIn = [(ρNIR - ρred ) / ρNIR + ρred + 1 – MSAVI n-1] * (2-MSAVI n-1)
(11)
The final solution for MSAVI is: MSAVI = 2 ρNIR + 1 – [(2 ρNIR + 1)2 –8(ρNIR - ρred) ]0.5 / 2
(12)
OSAVI is the Optimized Soil Adjusted Vegetation Index. This index has the same formulation of the SAVI family indices, but the value L or X as refered by Rondeaux et al. (1996) is the optimum value that minimizing the standard deviations over the full range of cover.
OSAVI = ρNIR - ρred / (ρNIR+ρred+0.16) (1+0.16)
(13)
GESAVI is the Generalized Soil Adjusted Vegetation Index. This index is based on an angular distance between the soil line and the vegetation isolines. GESAVI is not normalized and vary from 0 to 1 (from bare soil to dense canopies).Vegetation isolines are neither parallel nor convergent at the origin.
Vegetation isolines intercept the
soil line at any point depending on the vegetation amount.
GESAVI = ρNIR - ρred b - a / ρred + Z
(14)
Z is the soil adjustment coefficient, and its based on the assumption that vegetation isolines intercept soil line at any point in the third quadrant. Z decrease when vegetation cover increase. However, practically, Z consider vegetation isolines convergent in a point. At least this hypotesis may be limited for dense canopies (Gilabert et al., 2002). To normalize soil effects Z value is found at 0.35. Indices that include the Mid-InfraRed Band (MIR) are: Stress related Vegetation Index (STVI) (Gardener, 1983): STVI = ρMIR * ρred / ρNIR
(15)
Cubed ratio index (CRVI) (Thenkabail et al.,1994): CRVI = (ρNIR / ρMIR)3
(16)
The VIs that account for soil effect, do not consider atmospheric conditions, sensor viewing angle, solar illumination conditions. Kaufman and Tanré (1992) developed the Atmospherically Resistant Vegetation Index (ARVI) and the Soil and Atmospherically Resistant Vegetation Index (SARVI and SARVI2) where the reflectances are corrected for molecular scattering and ozone absorption. Liu and Huete (1995) incorporated a soil adjustment and atmospheric resistance concepts into a Modified Normalized Vegetation Index (MNDVI). SARVI2 as well as ARVI, SARVI are able to remove smoke effect and cirrus clouds from images (Huete et al., 1996).
12.4 Operational AgroMeteorological Products employing GIS and Remote Sensing Nowadays, public agencies, research laboratories, academic institutions, private and public services, have established their own GIS. Due to increasing pressure on land and water resources, land use management and forecasting (crop, weather, fire, etc.) become more essentials every day. GIS are, therefore, an irreplaceable powerful tool at the disposal of decision-makers. In Agrometeorology, to describe a specific situation, all relevant information available on the territory may be used: water availability, soil types, land cover, climatic data, geology, population, land-use, administrative boundaries and infrastructure (highways, railroads, electricity or communications systems). Each informative layer provides to the operator the possibility to consider its influence to the final result. However, more than the overlap of the different themes, the relationship of the numerous layers is reproduced with models (ranging from simple “indicator” formula, such as the USLE to physically process based models). The final information is extracted using graphical representation or precise descriptive indexes. Developed countries use agricultural and environmental GIS to plan the times and types of agricultural practices, territorial management activities, population security, to monitor devastating events and to evaluate damages. More than the classical applications, such as crop yield forecasting, uses such as those of the environmental and human security are becoming increasingly important. For instance, effective forest fires prevention needs a series of management strategy information on an enormous scale. The analysis of data, such as the vegetation coverage with different levels of inflammability, the presence of urban agglomeration, the presence of roads and many other aspects, allows the mapping of the areas where risk is greater. The use of other informative layers, such as the position of the control points and resource availability (staff, cars, helicopters, airplanes, fire fighting equipment, etc.), can help the decision-makers in the management of the territorial systems. Obviously some data sets, such as the under pinning DEM and temporally invariant (or static), whereas other data sources such a weather conditions (either near real-time observations or short-term forecasts) are temporally dynamic.
12.4.1 Assessment of meteorological and agronomic conditions to aid decisions on drought using remote
sensing. The following section is based on the material presented in McVicar and Jupp (1998). Precipitation and solar radiation are meteorological conditions that can be mapped and monitored by the meteorological remote sensing community which could directly assist in the scientific process to provide advice on the occurrence of drought. Precipitation Remote sensing provides additional information to the existing network of ground based precipitation gauges, for mapping the extent and amount of precipitation. It is unlikely that remotely sensed data will replace the existing network of precipitation gauges. There are several remote sensing techniques which have potential to assist in mapping the extent of precipitation patterns. Several reviews provide background on the use of satellite remote sensing to estimate precipitation (Arkin and Ardanuy, 1989; Barrett and Beaumont, 1994; Petty, 1995; Rasmussen and Arkin, 1992). Precipitation-sized ice particles present in some clouds, higher than the freezing level, result in a lowering of the return signal from clouds in the microwave region of the EMS relative to the nominal background value (Petty, 1995). This is especially evident in the 85.5 GHz band from SSM/I sensor. It is not the liquid layer of rain overlying the surface which directly affects the signal. The tops of clouds are colder than the land surface and provide an indication of where the clouds are. For convective weather systems cold cloud top temperatures imply the presence of precipitation-sized ice particles from which there is more likelihood that rain will result. Cloud top temperatures associated with frontal activity, which also bear rain, are usually warmer than the very cold temperatures measured in convective clouds. This is the physical basis for using high frequency thermal remotely sensed data to map rainfall patterns. Passive microwave remote sensing of rainfall over land was more achievable since the 1987 launch of the Special Sensor Microwave Imager (SSM/I) on board the (USA) Defense Meteorological Satellite Program (DMSP). The DMSP is a polar-orbiting sun-synchronous (meaning a revisit time of 12.0 hours) with the SSM/I scanner having a swath of approximately 1400 km. SSM/I is a four-frequency (19.35, 22.23, 37.00 and 85.50 GHz) seven channel (all are dual polarized except the 22.23 GHz channel, which is vertically polarized) passive microwave radiometer. Internationally there are numerous examples of the use of the SSM/I to estimate rainfall (e.g. Grody, 1991; Kniverton et al., 1997; Liu and Curry, 1992; Spencer et al., 1989; Wilheit et al., 1994). The skill of SSM/I data to estimate instantaneous precipitation over land for a 1.25 degree resolution cell is reported to be as high as 0.82 (Petty, 1995). However, the rainfall algorithm intercomparison needs be conducted for longer time periods over more precipitation producing conditions. International research indicates that integrating geostationary thermal measurements with other data to make rainfall rate estimates provides more promise than using remotely sensed data alone. Todd et al. (1996) used a temporally and spatially varying threshold, over small time and space scales, to improve the identification of rainfall distribution and estimation of amount. There are several integration techniques which may allow rainfall to be better predicted by combining the
thermal GMS data with other data sets (Ebert and Le Marshall, 1995). These include: 1. use of pattern recognition or visible data to determine cloud type (Ebert, 1987); 2. integrating shortwave infrared (SWIR) based inferences of cloud top droplet size may be linked to the presence of rainfall (Rosenfield and Gutman, 1994); and 3. combining the outputs from numerical weather prediction (NWP) to include some information about current meteorological conditions (Grassotti and Garand, 1994). Herman et al. (1997) have developed an operational system using this approach for Africa to provide 10-day estimates for the entire continent. The determination of the amount of precipitation solely from GMS data is unlikely. What appears possible is to link the ‘where’ and ‘when’ capabilities of remote sensing with the ‘how much’ from ground based measurements. This link may be made through physical based models or though statistical methods which combine rainfall gauge data with GMS data. Whichever approach is used complete validation of rainfall amounts derived from GMS-based estimates needs to be undertaken. Hence GMS data are a significant information source for overcoming the large distance between rainfall measuring stations. Using the high spatial resolution offered by remote sensing should assist in rainfall mapping for drought events especially in areas with a sparse rainfall measurement network. However, remote sensing only provides a snapshot which may have a revisit time of, at best hourly (geostationary satellites) to every 12 hours (polarorbiting sun-synchronous satellites). During that time clouds will move and intense periods of rainfall may occur and be over before the next revisit time. Providing accurate, precise and thoroughly validated space-time images of precipitation derived from using remote sensing is, and should continue to be, a major research area for issues like drought, climate prediction and thorough understanding of the global hydrologic cycle. Solar Radiation Solar radiation is a major determinant of plant growth, via photosynthesis, which in turn affects soil moisture via transpiration from the leaves. Loss of soil moisture also occurs through direct evaporation from the soil. Reflective measurements acquired by the GMS satellite may be used to estimate insolation over a region on a daily basis. Weymouth and Le Marshall (1994) incorporate the following physical parameters to estimate insolation using GMS data: 1. Rayleigh scattering; 2. absorption by water vapor; 3. absorption by ozone; 4. isotropic reflection and absorption by clouds; and 5. regularly updated surface albedo. For clear sky and near clear-sky conditions the average daily deviation of GMS based estimates compared to ground-based pyranomter measurements was 4.3%. This is within the error limits of well-maintained and
calibrated pyrometers. Under heavy cloud conditions the error between GMS based estimates compared to ground based pyranomter measurements increased to about 15%. The relative error may appear large; however, the absolute error is small since the amount of incoming solar radiation is low due to the heavy cloud cover conditions. Agronomic Conditions There are several ways in which remote sensing can assist in mapping and monitoring agronomic conditions of direct relevance to drought. These include mapping vegetation type and monitoring vegetation condition, moisture availability and soil moisture. This section will be divided into four main sections: 1. Vegetation Condition Monitoring with Reflective Remote Sensing; 2. Environmental Condition: Monitoring with Thermal Remote Sensing; 3. Soil Moisture: Monitoring with Microwave Remote Sensing; and 4. Environmental Stress: Combining Thermal and Reflective Remote Sensing.
Vegetation Condition : Monitoring with Reflective Remote Sensing There are a number of parameters that can be used to characterize and assess vegetation condition and cover and changes of this spatially and temporally. There are several promising approaches which are described in the following categories: 1. only remotely sensed imagery; 2. relating remotely sensed images with meteorological parameters; 3. crop yield modelling; and 4. inversion of plant growth models.
Only remotely sensed imagery A number of authors, with varying degrees of integration with GIS, have analysed the difference between two years of NDVI data. Approaches vary from visual display (Hendricksen, 1986) to statistical correlations (Peters et al., 1991) to image differences between 2 years (Lozana-Garcia et al., 1995; Reed, 1993). The last 3 papers analysed drought in the context of the 1987 and 1988 in contiguous USA. The next level of complexity is to scale the NDVI response of one image to the range of responses. The Vegetation Condition Index proposed by Kogan (1990) is defined as:
VCIj =
NVDIj − NDVImin *100% NDVImax − NDVImin
where: VCIj is the image of vegetation condition index values for date j; NDVIj is the image of NDVI values for date j; NDVImax is the image of maximum NDVI values from all images within the data set;
and
NDVImin is the image of minimum NDVI values from all images within the data set. The VCIj is a percentage of NDVI values at time j with respect to the maximum NDVI amplitude on a pixel by pixel basis. The VCI may be thought of as being closely related to the vegetation condition in a specific region. If AVHRR data are recorded over enough years, where the extremes in climate variability are sampled, then the VCI may indicate potential crop yields or carrying capacities within given agricultural systems. The primary aim of developing the VCI was to assess changes in the NDVI signal through time due to weather conditions, by reducing the influence of ‘geographic’ (Kogan, 1990) or ‘ecosystem’ (Kogan, 1995c) variables, meaning climate, soils, vegetation type and topography. This provides a mechanism to compare values across different landscapes, for example between rangeland and rainforest, to determine the changes in NDVI signal due to the prevailing weather conditions. An assumption in the calculation of the VCI is that reliable, calibrated AVHRR data be used to form the NDVI. Abrupt changes in land cover, for example woodland clearing, mean that the interpretation of this index is more problematic. Having reliable updated baseline maps of land cover will assist in the interpretation of the VCI in such cases. The VCI has been used to determine drought and hence poor vegetation growth and corresponding low yields for spring wheat in Kazakhstan (Gitelson et al., 1996; Kogan, 1995b), cotton in China (Kogan, 1995b) and barley production in Southern Russia (Kogan, 1995b). Liu and Kogan (1996) found that the NDVI was highly correlated with water deficit and rainfall for Cerrado (Savanna grassland) and Caatinga (woodland and open woodland) which both grow in areas with district wet-dry seasons. For 4 sites the NDVI explained 46 to 61 % of the variance of rainfall amount with a one month time-lag from August 1981 to July 1987. Within the time frame of the analysis there was no reporting to ensure that the land cover, and hence the response of the NDVI signal as a function of weather, had remained constant. Liu and Kogan (1996) defined drought in four ways. These are: 1. monthly rainfall less than 50mm; 2. NDVI lower than 0.18; 3. monthly rainfall departure lower than 50% of the mean (July 1985-June 1987) and 50 mm lower than the mean value (July 1987 - June 1992); and 4. VCI lower than 36%. The area of the total grain producing region with a VCI less than 0.36 was found to be highly related to the reduction in grain production of summer crops in Argentina and Brazil. Such analysis illustrates the potential of the VCI as an indicator of crop growth conditions. More detailed analysis, such as using cumulative VCI of the crop production areas, may provide a predictive ability. Liu and Kogan (1996) state that the VCI provides a “better” indicator of regional drought, when compared to the other types of drought delineation, 1-3 above. The response of the VCI is spatially and temporally different from the other delineations, but an analytical comparison with the operational output of drought declared by Governments would be a valuable addition to
confirm if differences are improvements. Moreover, water stress is only one cause of low green plant cover leading to NDVI signals. For instance in the southern highlands of New South Wales of Australia hydrologic drought may break in May, following autumn rains, but the pastoral drought may continue until September due to air temperature limiting plant growth. To take other environmental variables into account (e.g. air temperature, solar radiation and crop phenology) it may be best to stratify the NDVI by time. That is, use the maximum and minimum NDVI for the month, or season, of interest. McVicar and Jupp (1998) suggest stratifying the NDVI response by time, using the same scaling as the VCI, defines the monthly vegetation condition index (MVCI). This is defined as, using January as an example:
MVCIj, Jan =
NDVIj, Jan − NDVImin, Jan NDVImax, Jan − NDVImin, Jan
where: MVCIj,Jan is the image of monthly vegetation condition index values for date j, which
falls within the month
of January; NDVIj,Jan is the image of NDVI values for the jth image recorded in January; NDVImax,Jan is the image of maximum NDVI values from all images acquired in
January; and
NDVImin,Jan is the image of minimum NDVI values from all images acquired in
January.
This can be defined for any month, or season, within the data set. This allows the NDVI signal from January for year j to be compared to the range of all January NDVI signals within the data set available. A similarity between the MVCI, and NDVIdiff used by Reed (1993), and the NDVI divergence from a long term average analysis technique presented by ERIN, below, is the level of temporal stratification. In Australia NDVI has been used by a number of groups to make inferences about the changes in vegetation condition occurring across the Australian landscape. The analysis of AVHRR NDVI data is performed operationally and assists in the decision making process for drought. Cridland et al. (1995) analysed the four years of NDVI data, by plotting the NDVI signal as a time series. The height, in NDVI units, from a varying baseline to the maximum peak within the growing season has to be calculated. This green ‘flush’ is the response of the landscape to rainfall. It is defined as the maximum NDVI for a growing season subtracted from the baseline. The baseline was varied so that the influence of perennial cover in the NDVI signal was accounted for. The baseline is defined as the minimum from the previous year. The vegetation response or ‘flush’ recorded as the maximum for a particular year is then considered relative to the absolute maximum ‘flush’ within the four (or more) years of data. For 1994 this would be calculated using the following equation:
NDVImax,1994 − NDVIbaseline' 93−' 94 max NDVImax, year − NDVIbaseline' year
, which can be rewritten as
flush94 . max flushpopulation
By mid-1998 there should be the ability to directly measure photosynthetic activity using the high resolution spectral vegetation indices such as the Physiological Reflectance Index (PRI) (Gamon et al., 1992). The PRI has been shown to be linked more closely to plant physiological response than the NDVI. These direct measures will be available for the entire Australian continent once MODIS is launched, in mid 1998, it has a much finer spectral resolution than is currently available. The PRI has been shown to be closely correlated to levels of accessory pigments called xanthophylls involved in dissipation of excess photochemical energy during the plant’s CO2 assimilation process (Gamon et al., 1992). The PRI may be more sensitive to stresses, including drought, for vegetation communities that have strategies other than dropping leaves due to dry conditions. This would be true for overstory components of woodland and forests and may also be true of shrubs.
12.4. 2. Operational uses of remote sensing and GIS for irrigation scheduling The canopy variables needed for calculating the crop water requirements under standard conditions (disease-free, adequate fertilization soil water conditions) can be either extracted from tables or estimated from field and/or remote observations. Field observations are routinely used by Irrigation Advisory Services, but they often lack of objectivity in the evaluation and they are difficult to carry out over extensive areas. To this respect, the potentiality of remote sensing techniques in irrigation and water resources management is now widely acknowledged. Several algorithms for retrieving biophysical parameters of vegetation, such as LAI, biomass density and canopy roughness from remote sensing data with different spatial and temporal resolution have been successfully tried out in many different environments. On this baseline, experimental studies have assessed the direct correspondence between the spectral response of cropped surfaces and the corresponding value of evapotranspiration and crop coefficient Kc (Bailey, 1990; Bausch, 1995; Bausch and Neale, 1987;Choudry et la.,2000; D’Urso and Menenti, 1995, One important advantage of deriving crop coefficients from spectral measurements is that Kc values do not depend on variables such as planting date and density, but on the effective cover; as such, the spectral Kc value includes the variability within the same crop type due to actual farming practices. Within the DEMETER project (D’Urso personal communication), two different procedures for the operational estimation of crop water requirement from remotely sensed data have been developed and tested to support irrigation advisory services. The first procedure is based on the relationship between the NDVI and the value of the basal crop coefficient; the second procedure, named “Kc-analytical”, is based on the direct application of the Penman-Monteith equation with canopy parameters estimated from satellite imagery, in analogy to the direct calculation proposed by F.A.O. Details on both procedures can be found at the reference (D’Urso et al.,2006; FAO,1998, Moran et al., 1991). In the “Kc-NDVI”, an empirical relationship between the basal crop coefficient Kcb and the vegetation index NDVI is derived, considering a fractional vegetation cover fc=0.8 for a canopy at full development. Experimental data have been used to derive the following linear relationship:
K c = 1.25NDVI + 0.2
(1)
A correction should be applied when calculating Kc for the late season phase, because fc remains nearly constant in that phase. The analytical approach for mapping the crop coefficient Kc is based on the direct application of the PenmanMonteith equation, adopted in FAO-56 procedure (Allen et al., 1998). The vegetation parameters required in this schematization, namely the surface albedo, r, the leaf area index, LAI and the crop height, hc, are obtained from the processing of E.O. data. The calculation is performed assuming that canopy resistance is at minimum value, i.e. 70 sm-1 (potential conditions). Using ground-based meteorological data, the Kc values for each pixel are calculated. For the estimation of r from remote sensing data we have to solve three main problems: the directional integration of spectral radiance detected by the sensor, the spectral integration to obtain the planetary albedo, that is at top-of-atmosphere height, and the correction of atmospheric effects in each spectral band for deriving the surface albedo r from the planetary albedo rp. The current sensor capabilities, such as those of Landsat Thematic Mapper used in this study, impose several simplifications. Considering that radiance measurements are performed at different wavelengths, the spectral integration is approximated in discrete form, as expressed by the following relationship: ∞
r =π∫ 0
K ↑ (λ )
K ↓ (λ )
λn
dλ ≅ π ∑ λ1
K λ↑ ( d 0 )
2
(2)
Eλ0 cosϑ 0
In Eq.2 the spectral reflected radiance, K↑λ (W m-2), and the extraterrestrial solar irradiance, E0λ (W m-2), are integrated values over the width of each spectral band λi; ϑ0 and d0 are the solar zenith angle and the sun-earth distance in Astronomical Units. By grouping these quantities in a set of band-coefficients (which are sensordependent), Eq.(Bausch, 1995) can be simplified in the following expression:
r = ∑ λ wλ ρ λ where
λ = 1,2,...., n
(3)
represent the spectral reflectance (corrected for atmospheric effects) in the generic band.
Simple and feasible approaches based on empirical relationships between LAI and nadir-viewing measurements in the red and infrared bands has been have been defined by several authors. These methods implicitly assume that all other factors, except LAI, influencing the spectral response of canopy are fixed. In analogy with LAI, some correlation between vegetation indices and canopy roughness parameters may be found. Moran et al. 1994[8] tried out a purely empirical relationship linking the roughness length z0m of alfalfa to the ratio of reflectance in near-infrared and red bands. Nevertheless, it should be considered that when the radiation component in the surface energy balance is dominant, as it happens most frequently during the irrigation season at mid-latitude regions, the association of a mean crop height (constant) to each land-use class derived from satellite data, is a satisfactory compromise in areas where the absolute accuracy of ra,H is of minor concern in the calculation of daily values of ETp. The application of both approaches for Kc calculation requires a pre-processing of remotely sensed data, composed of three main steps:
1.
inter-satellite calibration
2.
atmospheric correction
3.
geometric correction and image re-sampling
Semi-automatic procedures have been developed in order to elaborate Kc maps from remotely sensed data in the minimum possible time. Pre-processing requires approximately half of the elaboration time of the entire process. Once geo-referenced surface reflectance has been calculated in each pixel, the algorithms for determining Kc are quite straightforward for both approaches described above. High fragmentation of land parcels, such as in Mediterranean agriculture systems, requires high-resolution imagery to resolve the smallest plots. One of the major limiting factors is thus the inadequate repeat cycles of high-resolution satellites. However, it is possible to obtain a revisit time of 1-5 days using the full set of currently available high-resolution satellites (ASTER, IKONOS, Landsat-5 TM, and SPOT). The resulting average resolution of 5-7 days is sufficient for crop water requirements calculation at both district scale and farm scale. When multiple sources of EO data are used to evaluate temporal crop evolution, factors such as sensor calibration differences between the various satellite systems, atmospheric condition and illumination-view geometry can affect pixel value. Thus, an atmospheric correction and a scene inter-calibration has to be performed in order to reduce reflectance variation due to non-surface factors (sensor, atmospheric and geometric conditions) so that variation in reflectance between date and different sensors could be related to actual change in crop conditions. An example of the derived product is shown in figure 4, where Kc raster maps derived by using different sensors are shown. Parcel boundaries are shown in overlay. It can be noticed that Landsat spatial resolution is still acceptable for this study-area for the smallest parcels. The six images acquires during two months of the central part of crop phonological cycle were adequate to monitor accurately the crop development. The maximum gap between two consecutive acquisitions was of ten days. That allows a good interpolation of the data to describe the crop development. See Van Niel and McVicar (2004b) for a review of how remote sensing can be used in irrigated rice based agricultural systems.
FIGURE 4. Time-series of Kc raster maps from multi-sensor acquisitions (Landsat-5 and IKONOS). Planting date of maize crop was set between June 3rd and 21st.
12.4.3. Operational uses of remote sensing and GIS for Soil and Crop Management 12.4.3.1. Evapotranspiration
All objects on the Earth’s surface emit radiation in the thermal-infrared (TIR) part of the spectrum (~ 8 to 14 µm). This emitted energy has proven useful in assessing crop water stress because the temperature of most plant leaves are mediated by soil water availability and its effect on crop evaporation (Jackson, 1982; Hatfield et al., 1983; Moran et al., 1989b; Pinter et al., 2003). In recent years, there has been much progress in the remote sensing of some of the parameters that can contribute to the estimation of evapotranspiration (ET). These include surface temperature, surface soil moisture, vegetative cover and incoming solar radiation. The surface temperature can be estimated from measurements at the thermal infrared wavelengths of the emitted radiant flux, that is the 10.5 and 12.5 µm. The microwave emission and reflection or backscatter from soil, primarily for wavelengths between 5 and 21 cm, are dependent on the dielectric properties of the soil, which are strong functions of the soil moisture content. Thus, measurements of these microwave properties can be used to obtain estimates of the surface soil moisture. Crop stress, due to water deficiency, crop diseases, is often shown with a decrease in the transpiration rate of the crop. Several studies have been carried on estimating ET with remote sensing data (Reginato, 1985; Jackson et al., 1987; Moran, et al., 1992, 1994, 1995; Maas, 1992,1993a, 1993b; Carlson et al., 1995, Hunsaker et al., 2003). A combination of remote sensing data and soil-plant-atmosphere models is commonly seen in the literature for ET estimation. The location of the “red edge” obtained with hyperspectral measurements shows potential for early detection of water stress (Shibayama et al., 1993). “Stress-Degree-Day” (SDD; Idso et al., 1977b), “Crop Water Stress Index” (CWSI; Idso et al., 1981; Jackson et al., 1981), “Non-water-stressed baseline (Idso et al., 1982), “Thermal Kinetic Window” (TKW; Mahan and Upchurch, 1988), “Water Deficit Index” (WDI, Moran et al., 1994), and the “Normalized Difference Temperature Index (NDTI McVicar and Jupp, 1999; 2002) are indices that measure plant stress induced by water stress. These indices have been used in research on more than 40 different crop species (Gardner et al., 1992a; Gardner et al., 1992b). Most studies have shown that the thermal infrared is more sensitive to water stress than is reflectance in visible or NIR. However, the reflective portion of the spectrum and VIs also respond to plant water stress status when the canopy changes architecture through the leaf rolling or wilting (Moran et al., 1989a) or alters the senescence rate (Pinter et al.,1981). Thermal plan water stress indices provide valuable information and adequate lead time to schedule irrigations, and allow the onset of stress conditions to be detected more rapidly in dryland areas. Thermal indices can overestimate water stress when canopy cover is full and the sensors view a combination of cool plant and warm soil temperatures. The WDI a combination of VI and TIR (Moran et al., 1994; Clarke, 1997 and Clarke et al, 2001) seems to have overcome this problem since it accounts for the amount of plant cover through the VI part of the index A cost benefit study by Moran (1994) shows that irrigation scheduling with thermal infrared sensors on aircraft is both practical and affordable if growers join together to purchase the images. Hatfield (1984c) found that spatial
variation of surface temperature in wheat changed with the degree of water availability. One alternative tool for a spatially variable irrigation can be to mount infrared sensors on irrigation booms to provide the capability to vary irrigation amounts as the unit travels across the field. VIs can be then used as surrogates for crop coefficients (Kcb). Crop coefficients are usually obtained from curves or tables and they lack flexibility to account for spatial and temporal crop water needs caused by uneven plant population, unusual weather patterns, non uniform water application, nutrient stress or pest pressures (Bausch and Neale, 1987; Choudhry et al., 1994; Pinter et al., 2003).
12.4.3.2 Soil Salinity Remote sensing can also be used to map areas of soils that have been contaminated by salt. The principle behind this application is that salt in the soil produces an unusually high surface reflectance (Leone et al., 2001). Salted areas can also be identified by detecting areas with reduces biomass or changes in spectral properties of plant growing in affected areas (Barnes et al, 2003). Leone et al., 2001 evaluated the impact of soil salinity induced through irrigation with saline water on plant characteristics and assessed the relationships between these characteristics and spectral indices. They showed that soil salinity had a clear impact on plant characteristics and significant relationships between chlorophyll content, biomass, NDVI and red edge peak. Studies have also shown an increase in canopy temperature of plants exposed to excessive salts in irrigation water (Howell et al., 1984a; Wang et al, 2002b), suggesting the possibility of pre-visual detection of stress which can manage with the appropriate measure of leaching or irrigation with good quality water.
12.4.3.3. Remote Sensing in Precision Agriculture Direct Application The past research efforts on remote sensing have provided a rich background of potential application to sitespecific management of agricultural crops. In spite of the extensive scientific knowledge, there few examples of direct application of remote sensing techniques to precision agriculture in the literature. The reasons are mainly due to the difficulty and expense of acquisition of satellite images or aerial photography in timely fashion. With the progress in GPS and sensor technology direct application of remote sensed data is increasing. Now an image can be displayed on the computer screen with real-time position superimposed on it. This allows for navigation in the field to predetermined points of interest on the photograph. Blackmer et al., (1995) proposed a system for N application to corn based on photometric sensors mounted on the applicator machine. They showed that corn canopy reflectance changed with N rate within hybrids, and the yield was correlated with the reflected light. Aerial photographs were used to show areas across the field that did not have sufficient N. The machine reads canopy colors directly and applies the appropriate N rate based on the canopy color of the control (well fertilized) plots (Blackmer and Schepers, 1996; Schepers et al., 1996). Management zones can be extracted using VIs maps and with the use of a GIS be viewed over a remotely sensed
image. The computer monitor displayed the image along with the current position as the applicator machine moved on the field. When interfaced with variable rate sprayer equipment, real time canopy sensors could supply site-specific application requirements improving nutrient use efficiency and minimizing contamination of groundwater (Schepers and Francis, 1998). Indirect Applications The most common indirect use of remote sensing images is as a base map on which other information is layered in a GIS. Other indirect applications include use of remotely measured soil and plant parameters to improve soil sampling strategies, remote sensed vegetation parameters in crop simulation models, and use in understanding causes and location of crop stress such as weeds, insect, and diseases. Moran et al. (1997) in their excellent review on opportunities and limitations for image-based remote sensing in precision agriculture, classify the information required for site specific management in information on seasonally stable conditions, information on seasonally variable conditions, and information to find the causes for yield spatial variability and to develop a management strategy. The first class of information includes conditions that do not vary during the season (soil properties) and only need to be determined at the beginning of the season. Seasonally variable conditions, instead, are those that are dynamic within the season (soil moisture, weeds or insect infestation, crop diseases) and thus need to be monitored throughout the entire season for proper management. The third category is comprehensive of the previous two to determine the causes of the variability. Remote sensing can be useful in all three types of information required for a successful precision agriculture implementation. Muller and James (1994) suggested a set of multi-temporal images to overcome the uncertainty in mapping soil texture due to differences in soil moisture and soil roughness. Moran et al. (1997) also suggested that multi-spectral images of bare soil could be used to map soil types across a field.
12.4.3.4 Crop growth and intercepted radiation Remote sensing techniques have also been applied to monitor seasonally variable soil and crop conditions. Knowledge of crop phenology is important for management strategies. Information on the stage of the crop could be detected with seasonal shifts in the “red edge” (Railyan and Korobov, 1993), bidirectional reflectance measurements (Zipoli and Grifoni, 1994), and temporal analysis of NDVI (Boissard et al., 1993; Van Niel and McVicar 2004a). Moreover Wiegand et al. (1991) consider them as a measure of vegetation density, LAI, biomass, photosynthetically active biomass, green leaf density, photosynthesis rate, amount of photosynthetically active tissue and photosynthetic size of canopies.
Aparicio et al. (2000) using three VIs (NDVI; Simple Ratio; Photochemical Reflectance Index) to estimate changes in biomass, green area and yield in durum wheat. They results suggest that under adequate growing conditions, NDVI may be useful in the later crop stage, as grain filling, where LAI values are around 2.
SR,
under rainfed condition, correlated better with crop growth (total biomass or photosynthetic area) and grain yield than NDVI. This fact is supported by the nature of relationship between these two indices and LAI. SR and LAI show a linear relationship, compared to the exponential relationship between LAI and NDVI. However the utility
of both indices, as suggested by the authors, for predicting green area and grain yield is limited to environments or crop stages in which the LAI values are < 3. They found that in rainfed conditions, the VIs measured at any stage were positively correlated (P < 0.05) with LAI and yield. Under irrigation, correlations were only significant during the second half of the grain filling. The integration of either NDVI, SR, or PRI from heading to maturity explained 52, 59 and 39% of the variability in yield within twenty-five genotypes in rainfed conditions and 39, 28 and 26% under irrigation. Shanahan et al. (2001) use three different kinds of VIs (NDVI, TSAVI, GNDVI) to asses canopy variation and its resultant impact on corn (Zea mays L.) grain yield. Their results suggest that GNDVI values acquired during grain filling were highly correlated with grain yield, correlations were 0.7 in 1997 and 0.92 in 1998. Moreover they found that normalizing GNDVI and grain yield variability, within treatments of four hybrids and five N rates, improved the correlations in the two year of experiment (1997; 1998). Correlation, however, increases with a net rate in 1997 from 0.7 to 0.82 rather than in 1998 (0.92 to 0.95). Therefore, the authors suggest that the use of GNDVI, especially acquiring measurements during grain filling is useful to produce relative yield maps that show the spatial variability in field, offering an alternative to use of combine yield monitor. Raun et al. (2001) determined the capability of the prediction potential grain yield of winter wheat (Triticum aestivum L.) using in-season spectral measurements collected between January and March. NDVI was computed in January and March and the estimated yield was computed using the sum of the two postdormancy NDVI measurements divided by the Cumulative Growing Degree Days from the first to the second reading. Significant relationships were observed between grain yield and estimated yield, with R2 = 0.50 and P > 0.0001 across two years experiment and different (nine) locations. In some sites the estimation of potential grain yield, made in March and measured grain yield made in mid-July differed due to some factors that affected yield. The capability of VIs to estimate physiological parameters, as fAPAR,
is studied on other crops, faba bean
(Vicia faba L.) and semileafless pea (Pisum sativum L.) that grows under different water condition, as an experiment followed by Ridao et al. (1998) where crops see above grew both under irrigated and rainfed conditions. They have computed several indices (RVI, NDVI, SAVI2, TSAVI, RDVI, PVI) and linear, exponential and power relationship between VI and fAPAR were constructed to assess fAPAR from VIs measurements. During the pre-LAImax phase, in both species, all VIs correlated highly with fAPAR, however R2 at this stage did not differ significantly between indices that consider soil line (SAVI2 and TSAVI) and those that did not consider it (NDVI, RVI, RDVI). In post-LAImax phase the same behaviour was observed. All VIs are affected by the hour of measurement at solar angles grater than 45°. Authors conclude that simple indices as RVI and NDVI, can be used to accurately assess canopy development in both crops, allowing good and fast estimation of fAPAR and LAI.
12.4.3.5 Nutrient Management Appropriate management of nutrients is one of the main challenges of agriculture productions and at the same
environmental impact. Remote sensing is able to provide valuable diagnostic methods that allow for the detection of nutrient deficiency and remedy it with the proper application. Several studies have been carried out with the objective of using remote sensing and vegetation indices to determine crop nutrient requirements (Schepers et al., 1992; Blackmer et al., 1993; Blackmer et al., 1994; Blackmer et al., 1996a; Blackmer et al., 1994b; Blackmer and Schepers 1996; Daughtry et al. (2000). Results from these studied concluded that remote sensing imagery can be a better and quicker method compared to traditional method for managing nitrogen efficiently. Bausch and Duke (1996) developed a N reflectance index (NRI) from green and NIR reflectance of an irrigated corn crop. The NRI was highly correlated to with an N sufficiency index calculated from SPAD chlorophyll meter data. Because the index is based on plant canopy as opposed to the individual leaf measurements obtained with SPAD readings, it has great potential for larger scale applications and direct input into a variable rate application of fertilizer. Ma et al. (1996) studied the possibility to evaluate if canopy reflectance and greenness can measure changes in Maize yield response to N fertility. They have derived NDVI at three growing stage: preanthesis, anthesis and postanthesis. NDVI is well correlated with leaf area and greenness. At preanthesis NDVI showed high correlation with field greenness. At anthesis correlation coefficient of NDVI with the interaction between leaf area and chlorophyll content was not significant with yield. Ma et al. (1996) summarized that reflectance measurements took prior to anthesis predict grain yield and may provide in-season indications of N deficiency. Gitelson et al. (1996) pointed out that in some conditions, as variation in leaf chlorophyll concentration; GNDVI is more sensitive than NDVI. In particular is the green band, used in the computing GNDVI that is more sensitive than the red band used in NDVI. This change occurs when some biophysical parameters as LAI or leaf chlorophyll concentration reach moderate to high values. Fertility levels, water stress and temperature can affect the rate of senescence during maturation of crops. In particular Adamsen et al. (1999) used three different methods to measure greenness during senescence on spring wheat (Triticum aestivum L.): digital camera, SPAD, hand-held radiometer.
They derived G/R (green to red)
from digital camera, NDVI from an hand-held radiometer and SPAD readings was obtained from randomly selected flag leaves. All three methods showed the similar temporal behaviour. Relationship between G/R and NDVI showed significant coefficient of determination and their relationship were described by a third order polynomial equation (R2 = 0.96; P < 0.001). Relation is linear until G/R > 1, when canopy approach to maturity (G/R < 1) NDVI is still sensitive to the continued decline in senescence than did G/R. This fact suggests that the use of the visible band is limited in such conditions. However authors found that G/R is more sensitive than SPAD measurements. Daughtry et al. (2000) have studied the wavelengths sensitive to leaf chlorophyll concentration in Maize (Zea
mays L.). VIs as NIR/Red, NDVI, SAVI and OSAVI, have shown LAI as the main variable, accounting for > 98% of the variation. Chlorophyll, LAI, and their interaction accounted for > 93% of the variation in indices that compute the green band. Background effect accounted for less than 1% of the variation of each index, except for GNDVI, which was 2.5%. Serrano et al., (2000) studied the relationship between VIs and canopy variables (aboveground biomass, LAI canopy chlorophyll A content and the fraction of intercepted photosynthetic active radiation (fIPAR) for a wheat crop growing under different N supplies. The VIs-LAI relationships varied among N treatments. The authors also showed that VI were robust indicators of fIPAR independently of N treatments and phenology. Li et al. (2001) studied spectral and agronomic responses to irrigation and N fertilization on cotton (Gossypium hirsutum L.) to determine simple and cross correlation among cotton reflectance, plant growth, N uptake, lint yield, site elevation, soil water and texture. NIR reflectance was positively correlated with plant growth, N uptake. Red and middle-infrared reflectance increased with site elevation. Li et al. (2001) found that soil in depression areas contains more sand on the surface than on upslope areas. This behaviour modified reflectance patterns. As a result, a dependence on sand content was shown by NDVI with higher values in the depression areas and lower values in areas where the soil had more clay. In addition cotton NIR reflectance, NDVI, soil water, N uptake and lint yield were significantly affected by irrigation (P < 0.0012). The N treatment had no effect on spectral parameters, and interaction between irrigation and N fertilizer was significant on NIR reflectance (P < 0.0027). Wright (2003) investigated the spectral signatures of wheat under different N rates, and the response to a midseason application at heading. VIs were computed (RVI, NDVI, DVI, GNDVI) and spectral data were compared with pre-anthesis tissue samples and post-harvest grain quality. The author found that imagery and tissue samples were significantly correlated with pre-anthesis tissue samples and post-harvest grain quality. The second application of N at heading improved protein only marginally. GNDVI was significantly correlated with nitrogen content of plants. VIs used in the study, whether from satellite or aircraft correlated well with preseason N and plant tissue analysis, but had lower correlation with protein. Osborne et al. (2002a; 2002b) demonstrated that hyperspectral data in distinguishing difference in N and P at the leaf and canopy level, but the relationship were not constant over all plant growth stages.
Adams et al. (2000)
have detected Fe, Mn, Zn and Cu deficiency in soybean using hyperspectral reflectance techniques and proposing a Yellowness Index (Adam et al. 1999) that evaluated leaf chlorosis based on the shape of the reflectance spectrum between 570 nm and 670 nm. 12.4.3.6 Pest Management Remote sensing has also shown great potential for detecting and identifying crop diseases (Hatfield and Pinter, 1993) and weeds. Visible and NIR bands can be useful for detecting healthy plants versus infected plants because diseased plant react with changes in LAI, or canopy structure. Malthus and Madeira, 1993, using hyperspectral
information in visible and NIR bands, were able to detect changes in remotely sensed reflectance before disease symptoms were visible to the human eye. Weed management represent an important agronomic practice to growers. Weeds compete for water, nutrient, light and often reduce crop yield and quality. Decisions concerning their control must be made early in the crop growth cycle. Inappropriate herbicide application can also have the undesirable effect on the environment and a side effect to the crop. With the advent of precision agriculture, there has been a chance from uniform application to the adoption of herbicide-ready crop and to apply herbicide only when and where needed. This kind of approach is economically efficient and environmentally sound but site-specific herbicide management requires spatial information on the weeds. The discrimination between crops and weeds is usually accomplished based on the differences in the visible/NIR spectral signatures of crops and specific weeds (Gausman et al., 1981; Brown et al., 1994) or by acquiring images when weed coloring is particularly distinctive. Richardson et al., (1985) demonstrated that multispectral aerial video images could be used to distinguish uniform plot of Johnsongrass and pigweed from sorghum, cotton and cantaloupe plots. Several other authors have utilized spectral imagery to separate crops from weeds based on spectral signatures of species and bare soil (Hanks and Beck, 1998) or based on the leaf shape determine by the machine vision technology (Franz et al., 1995; Tian et al., 1999). Basso et al., 2004 used the handheld radiometer CropScan to determine if a wheat field with various level of pappy (Papever Rhoeas) infestation could be detected by the multispectral radiometer. The study showed that the reflectance in the Red and NIR of the highly infested areas with pappy of the durum wheat field was significantly different from the no infestations of lower levels of weed presence. Remote sensing can also be used to determine herbicide injury to the crop for insurance purposes (Hickman et al., 1991; Donald, 1998a, Donald 1999b).
To improve application efficiency of herbicides, Sudduth and Hummel
(1993) developed a portable NIR spectrophotometer for use in estimating soil organic matter as part of the estimation procedure for the amount of herbicide to be sprayed. Several studies have also been carried out using remote sensing for identifying and managing insects, mite and nematode populations. Such studies have been able to demonstrate that remote sensing is able to detect actual changes in plant pigments caused by pest presence, damages by pest and to identify areas susceptible to infestation. Riedell and Blackmer (1999) infested wheat seedlings with aphids and after 3 weeks they measured the reflectance properties of individual leaves. The leaves of the infected plants had lower chlorophyll concentration and displayed significant changes in reflectance spectra at certain wavelengths (500 to 525, 625 to 635 and 680 to 695 nm). This study in combination with others (Cook et al., 1999; Elliot et al., 1999; Willers et al., 1999) suggests the potential usefulness of canopy spectra for identifying outbreaks in actual field situations and to guide field scouts to specific areas for directed sampling. Site specific pesticide application can reduce the impact of toxic chemicals on the environment by 40 percent (Dupont et al., 2000).
12.4.3.7. Selection of growth traits
The use of morphological and physiological traits as indirect selection criteria for grain yield is an alternative to breeding approach. Future wheat yield improvements may be gained by increasing total dry matter production (TDM). VIs have been proposed as an appropriate and nondestructive method to assess total dry matter and LAI. Aparacio et al., (2000) and (2002) investigated whether VIs could accurately identify TDM and LAI in durum wheat and serve as indirect selection criteria in breeding programs. They found that the best growth stages for growth traits appraisal were stages 65 and 75 of the Zadock scale (Zadok et al., 1974). VIs accurately tracked changes in LAI when data were analyzed across a broad range of different growth stages, environment and genotypes. Since VIs lack of predictive ability for specific environment/growth stages combinations, their value as indirect genotypes selection criteria for TDM or LAI was limited. Ma et al., (2001) showed that canopy reflectance measured between R4 and R5 stages in soybean adequately discriminates high from low yielding genotypes and providing a reliable and fast indicator for screening and ranking soybean genotypes based on the relationship between NDVI and grain yield.
12.4.3.8 Crop Yield Estimation Remote sensing can provide valuable information of yield assessment and show spatial variation across the field. There are two approaches for yield estimation; the first is a direct method in which predictions are derived directly from remote sensing measurements (Figure 6). The second method is an indirect one, where remotely sensed data are incorporated into simulation model for crop growth and development either as within season calibration checks of model output (LAI, biomass) or in a feedback loop used to adjust model starting conditions (Maas, 1988). The direct method for prediction yield using remote sensing can be based on reflectance or thermal-based. Both methods have been applied with case of successes on various crops like corn, soybean, wheat, alfalfa (Tucker et al, 1979; Tucker et al., 1981; Idso et al., 1977; Pinter et al., 1981). Hatfield (1981) in his survey of 82 different varieties of wheat was not able to find a consistent relationship between spectral indices and yield. Hatfield (1983b) coupled frequent spectral reflectance and thermal observation in a more physiological method to predict yields in wheat and sorghum. This method requires TIR daily measurements during grain filling period to estimate crop stress. Shanahan et al., (2001) demonstrated that the time of corn pollination was not a good growth stage to estimate yield because of the various that can cause tassel emergence dates to vary.
Yang et al., (2000) found similar
results, concluding that images from images taken at grain filling can provide good relationships between VIs and yield. Reliability of imagery for use in yield estimation decreases as the time before harvest increases because there are more opportunities for factors like various nature of stresses to influence yield. Aase and Siddoway (1981) had cautioned that the relationships of spectral indices to yield were dependent upon normal grain-filling conditions for the crop. Similar results were found by Basso et al., 2004 (personal communication, unpublished data) where the NDVI images on a rainfed durum wheat field showed different
correlation to yield depending on the time of the image selected. In this specific case, spatial variability of soil texture and soil water uptake by plants affected by drought varied at anthesis presenting different scenarios from the one predicted by the NDVI estimation.
12.4.4. Operational uses of remote sensing and GIS for assessing Environmental Sensitive Area for Desertification Risk Soil, vegetation, climate and management are the main factors affecting environmental sensitivity to degradation, through their intrinsic characteristics or by their interaction on the landscape. Different levels of degradation risks may be observed in response to particular combinations of the aforementioned factors. For instance, the combination of inappropriate management practices and intrinsically weak soil conditions will result in a degradation of the environment of a severe level, while the combination of the same type of management with better soil conditions may lead to negligible degradation. A weighted multiplicative model within a GIS has been developed in order to assess the environmental sensitivity (ES) of Basilicata Region (Italy) by taking into account the particular set of environmental and socio economic conditions and their relationships. Furthermore through spatial analysis, mayor contributing factors to degradation have been identified at across Basilicata region. Environmental Degradation or Sensitivity (ES) has been modelled as the multiplicative effect of soil, climate, vegetation and management, indicated in the model as quality layers. ES = (Quality Index1 * Quality Indexr2 * Quality Index3 * Quality Indexn) (1/n) In turn each, Quality Index (QI) represents the result of interactions among the elementary factors listed above according to the following equation: QI = (Information Layer1 * Information
Layer2 * Information Layer3 * Information Layern) (1/n)
Each layer represents a single variable and measures how such variable relates on its own to the general environmental sensitivity. Details of the ES model can be found in Basso et al., (2000). Grouping ES values into three classes, using a natural break classification method, low, high and severe risk classes were identified. Over 50 % of Basilicata surface is exposed to high risk, and 7 % to severe risk, while the remaining 37 % is exposed to low risk as summarized in Fig.4
FIGURE 4. Climate quality map; Vegetation quality map; Environmental Sensitivity; Risk Class and Contribution factor to spatial distribution of Desertification Risk in Basilicata region (Lat 40°.6’, Long, 16°.6’) Southern Italy.
12.4.5
Aquaculture and Remote Sensing Mangroves form an important vegetation belt along the coastal regions and its presence has
considerable influence in maintaining the proper environment balance. Destruction of mangroves will, in the long run, have a serious impact on the coastal ecosystem. Mangroves to some extent protect land from cyclones and its ill-effects on crops, cattle and human habitation. A study by scientist of the NRSA, Hyderabad, using satellite remote sensed data covering Andhra Pradesh ecosystems to delineate many of the cultivated areas and mangroves, revealed that in 1973, the area under prawn cultivation was almost negligible and by 1985and 1990, it became more than five times in Guntur and 10 times in Krishna district. In this study, one could also notice the decrease in the area of mangrove vegetation suggesting that the increase in prawn cultivation may have to some extent, affected mangroves. Mangroves, essential for maintaining
ecological balance in the coastal ecosystems, are being destroyed in some
areas as in Kottapalem in Repalle Mandal in Guntur district, for extending prawn culture(Narayan,1999). Remote sensing data not only help in the estimation of the area under brackish water aquaculture along the whole Indian coast, but can also help in selecting the suitable locations for prawn/shrimp farming without a serious threat to mangrove systems which are essential in the coastal areas. In Krishna district the prawn cultivation has increased from 1973 to 1992, while the mangroves have reduced during the same period as could be observed from satellite remote sensed data of this period. In terms of statistics, prawn cultivation was nil while the area under mangroves was about 5884 ha in Krishna district in 1973. In 1992, the area under prawn cultivation increased to 6005 ha, while the area under mangroves decreased to 5479 ha.
12.4.6
Operational use of remote sensing for identification of fishing zones With the advent of remote sensing methods, many ocean related applications including fisheries can be
studied using satellite and aircraft data. One of the important parameters that can be measured with sufficient accuracy is the Sea Surface Temperature (SST), which has been related to the concentration of fish population. Here, it has been shown how sea surface temperature can be mapped on a regular basis, and pass it on to the fishermen who could concentrate on high potential areas and improve the catch (Das, 2004). SST derived from NOAA-AVHRR satellite serves as a very useful indicator of prevailing and changing environmental conditions and is one of the important parameters which decides suitable environmental conditions for fish aggregation. SST images obtained from satellite imagery over three or four days are composited and the minimum and maximum temperatures are noted down. obtain maximum contrast of the thermal information.
These values are processed to
These are filmed to prepare relative thermal gradient
images. From these images features such as thermal boundaries, relative temperature gradients with an accuracy of 1 degree centigrade, contour zones, eddies and upwelling zones are identified.
These features are
transferred using optical instruments to corresponding sectors of the coastal maps prepared with the help of Naval Hydrograph charts.
Later, the location of the Potential Fishing Zone (PFZ) with reference to a particular
fishing centre is drawn by identifying the nearest point of the thermal feature to that fishing centre.
The
information extracted consists of distance in kilometres, depth (for position fixing) in metres and bearing in degrees with reference to the North for a particular fishing centre. The PFZ maps thus prepared are sent through facsimile transmission (FAX) to major fishermen’s associations of India, unions and Governmental organizations of India such as the Central Marine Fishery Research Institute (CMFRI), Fishery Survey of India (FSI), and state fishery departments of all maritime states of India, including the union territories of Andaman and Nicobar and Lakshadweep Islands every Monday and Thursday. The Department of Ocean Development has already set up fax machines at the ports of Balasore, Bhubaneswar, Thiruvananthapuram and Malpe (Karnataka State) and has plans to extend such facilities to all major fish landing centres along the Indian coast. Wherever fax facility is not available, descriptive information is sent.
12.4.7
Forest Management through Remote sensing The presence scenario in the use of remote sensing in forest studies indicates that for forest cover
monitoring or surveillance, the use of this technology at the national and state level and mapping on a scale of 1:25 000 is possible. This is operationally being done by the Forest Survey of India, to produce and report information on a biennial basis. Such information has been found to be quite reliable and can be generated within a short span of time and, if necessary, in digital format. Besides this, the study of spatial distribution of forest types could also be necessary and useful, particularly to provide a better insight to the forest officials as to what type of species need to be introduced either by replacement of natural forest or regeneration of degraded forests in the area in a more scientific manner. Preparation of forest type maps and their distribution at the block level on a scale of 1:50 000, is now possible using satellite remote sensed data to aid forest managers.
12.5 Collaborations for resource sharing in RS & GIS The transfer of the new techniques of processing and interpreting remote sensing data from developed to developing countries is limited by many factors, such as the cost of receiving equipment, the restrictive or very difficult access to the archives of satellite images and data, the shortage of qualified staff, etc. The situation has changed to better in recent years, thanks to the availability of long series of satellite data; for example the data archives from NOAA (USA) and Meteor (Russia), contain information for more than 15 years. In this way, the researcher and user, have the possibility to apply the traditional techniques, approaches and procedures and to local studies. The access to the archives and the transfer of information, software and so on is becoming simpler, especially with the use of INTERNET tools (e.g, see Schmidt et al 2006 and the references therein). International organisations and in particular the WMO also play an active role in coordinating efforts connected with receiving, processing, disseminating and using remote sensing data. WMO’s Commission on Basic Systems (CBS) has recently established a Working Group on SATellites (WGSAT) that can be the place designed for such activity. The main goal is the development of common working strategies and the improvement of regional and global management capability of satellite data. For this reason, particular emphasis is placed on data compatibility and integration between different sources of data. WGSAT has supported a project aimed at developing the receiving stations (both hardware and software) at a reasonable cost to be available for developing countries. The WGSAT has discussed a possible process to improve the use of satellite data from the present global satellite observing system, and has proposed a set of recommended actions. The WMO strategy to improve satellite system utilization is made up by four components: a) the strategic vision; b) the long-term strategic goal; c) the major objectives; d) the methodology to meet the objectives.
The strategic vision, to improve satellite system utilisation, is the prospect of substantially improved transfer to communities around the world of the benefit of meteorological science and technology, via rapidly evolving global and regional communications networks. This will allow improved access to satellite data and services and the interactions between developed and developing countries. The long-term strategic goal for the next 15 years is to improve systematically the utilisation of satellite systems by National Meteorological and Hydrometeorological Services with emphasis on improving utilization in the developing countries. The major strategic objectives are: - to focus on the needs of the developing countries; - to improve access to satellite data through increased effectiveness in the distribution of satellite system data and products at major hubs, in particular those maintained by the satellite operators, WMO WMC’s, RSMCs and other entities as appropriate; - to improve the use of satellite data through increased capabilities in its applications by direct involvement of existing WMO Member expertise. The methodology to improve satellite system utilization is based on an iterative process to assess continuously the status of the use of satellite data and services and their impact on the various applications and, therefore, to identify limitations and deficiencies. The necessary steps to improve the utilisation shall be developed and implemented through the use of specific projects.
12.6. References Allen RG., Pereira LS., Raes D., Smith M. 1998. Crop evapotranspiration - Guidelines for computing crop water requirements.' FAO Irrigation and Drainage Paper 56, Rome, Italy Aparicio, N., Villegas, D., Casadeus, J., L. Araus, J., Royo, C. 2000. Spectral Vegetation Indices as Nondestructive Tools for Determining Durum Wheat Yield. Agron. J. 92: 83-91. Aparicio, N., Villegas, D., J.L. Araus, J., Casadeus, J.Royo, C. 2002. Relationship between Growth Traits and Spectral Vegetation Indices in Durum Wheat. Crop Sci. 42-1547-1555. Arkin, P.A. and Ardanuy, P.E., 1989. Estimating Climatic-Scale Precipitation from Space: A Review. J of Climate, 2: 1229-1238. Bailey, J.O. (1990). The Potential Value of Remotely Sensed Data in the Assessment of Evapotranspiration and Evaporation. Remote Sensing Reviews, 4(2):349-377. Barnes, E. .M., Sudduth,K.A., Hummel,J.W., Lesch,S.M., Corwin, D.L., Yang,C., Daughtry,C.S.T., and Bausch,W.C. 2003. Remote- and ground-based sensor techniques to map soil properties." Photogrammetric Engineering and Remote Sensing 69(6): 619-630 Barrett, E.C. and Beaumont, M.J., 1994. Satellite Rainfall Monitoring: an overview. Remote Sensing Reviews, 11: 23-48. Basso, B., D. Cammarano, P.De Vita. 2004 Remotely Sensed Vegetation Indices: Theory And Applications For
Crop Management. Ital. J. of Agrometeorology (1) 36-53 Basso, F., E. Bove, S. Dumontet, A. Ferrara, M. Pisante, G. Quaranta, M. Taberner. 2000. Evaluating environmental sensitivity at the basin scale through the use of geographic information systems and remotely sensed data: an example covering the Agri basin_Southern Italy. Catena (40) 19-35 Bausch W.C., Neale C.M.U. 1987. Crop coefficients derived from reflected canopy radiation: a concept. Trans ASAE 30(3):703–709 Bausch, W.C., Duke, H.R. 1996. Remote sensing of plant nitrogen status in corn, Transactions of the ASAE, 39(5):1869-1875. Bayarjargal Y Karnieli A, Bayasgalan M, Khudulmur S, Gandush C, Tucker CJ. 2006. A comparative study of NOAA-AVHRR derived drought indices using change vector analysis. Remote Sensing of Environment 105, 9-22. Blackmer, T. M., Schepers, J. S. 1995. Use of a chlorophyll meter to monitor N status and schedule fertigation of corn. J. Production Agriculture 8, 56-60. Blackmer, T.M., Schepers, J.S. 1994. Techniques for monitoring crop nitrogen status in corn. Commun. Soil Sci. Plant Anal. 25:1791–1800 Blackmer, T.M., Schepers, J.S., Meyer, G.E. 1996a. Analysis of aerial photography for nitrogen stress within corn fields, Agronomy Journal, 88(5):729-733. Blackmer, T.M., Schepers, J.S. Varvel, G.E., alter-Shea, E.A.1996. Nitrogen deficiency detection capabilities of light reflection from irrigated corn canopies. Agron. J. 88:1–5. Boissard, P., Pointel, J. G., Huet, P., 1993. Reflectance, green leaf area index and ear hydric status of wheat from anthesis until maturity. Int. J. Remote Sens. 14, 2713-2729. Brown, R. B., Steckler, J. P. G. A., Amderson, G. W., 1994. Remote sensing for identification of weeds in no-till corn. Trans. ASAE. 37, 297-302. Burgess, W. and Pairman, D. 1997. Bidirectional reflectance effects in NOAA AVHRR data. International Journal of Remote Sensing, 18, 2815--2825. Carlson, T.N., Taconet, O. Vidal, A. Gilles, R.R. Olioso, A. Humes, K. 1995. An overview of the workshop on thermal remote-sensing held at La-Londe-Les-Maures, France, September 20-24, 1993, Agricultural and Forest Meteorology, 77(3-4):141-151. Choudhury BJ, Amhed NU, Idso SB, Reginato RJ, Daughtry CS (1994) Relations between evaporation coefficients and vegetation indices studied by model simulations. Remote Sens Environ 50:1-17 Choudhury, B.J. and DiGirolamo, N.E. 1995. Quantifying the effect of emissivity on the relation between AVHRR split window temperature difference and atmospheric precipitable water over land surface. Remote Sensing of Environment, 54, 313--323. Clarke,T.R. 1997. An empirical approach for detecting crop water stress using multispectral airborne sensors. Hortechnology 7(1) 9-16 Clarke, T.R., Moran, M.S., Barnes, E.M. Pinter, P.J,Jr., Qi, J. 2001. Planar domain Indices: A method for measuring a quality of a single component in two-component pixels, Proceedings IEEE International Geoscience and Remote Sensing Symposium, 09-13 July, Sydney, Australia, unpaginated CD-ROM. Congalton RG (2001) Accuracy assessment and validation of remotely sensed and other spatial
information. International Journal of Wildland Fire 10, 321-328. Cook, C.G., Escobar, D.E., Everitt, J.H. Cavazos, I., Robinson, A.F. Davis, M.R.1999. Utilizing airborne video imagery in Kenaf management and production, Industrial Crops and Products, 9:205-210. Cridland, S., Burnside, D. and Smith, R.C.G., 1995. The NDVI-Use in Rangeland Management, Proceedings of the 5th International Rangelands Congress, Salt Lake City, Utah, USA, pp. 105-106. Curran, P.J., Dungan, J.L. and Gholz, H.L. 1992. Seasonal LAI in slash pine estimated with landsat TM. Remote Sensing of Environment, 39, 3--13. Das, H.P. 2004, Satellite based Agro-Advisory Services. Proc. Of WMO Training Workshop on satellite Remote Sensing and GIS applications in Agricultural Meteorology, AGM-8, WMO/TD No . 1182, pp. 347-359. Daughtry, C. S. T., Walthall, C. L., Kim, M. S., de Colstoun, E. B., McMurtrey III, J. E. 2000. Estimating Corn Leaf Chlorophyll Concentration from Leaf and Canopy Reflectance. Remote Sens. Environ. 74 : 229239Deering, D.W. 1989. Field measurements of bidirectional reflectance. In: Asrar, G., ed., Theory and Application of Optical Remote Sensing. New York, John Wiley and Sons, pp. 14--65. Donald, W.W., 1998a. Estimated soybean (Glicine max) yield loss from herbicide damage using ground cover or rated stunting, Weed Science, 46:454-458. Donald, W.W., 1998b. Estimating relative crop yield loss resulting from herbicide damage using crop ground cover or rated stunting, with maize and sethoxydim as a case study, Weed Research, 38:425-431. Dupont, J.K., Campenella, R. , Seal, M.R., Willers, J.L., Hood, K.B. 2000. Spatially variable insecticide applications through remote sensing, Proceedings Beltwide Cotton Conferences, 04-08 January, San Antonio, Texas (National Cotton Council of America and The Cotton Foundation), 1:426-429. D’Urso G., Menenti M. (1995). Mapping crop coefficients in irrigated areas from Landsat TM images; Proceed. European Symposium on Satellite Remote Sensing II, Europto, Paris, sett.’95; SPIE, Intern. Soc. Optical Engineering; 2585: 41-47 D’Urso G., Osann Jochum M.A., Moreno J. (Eds.), (2006). Earth Observation for vegetation monitoring and water management, Conference Proceedings Naples, 2005, 9-10 Nov., published by American Inst. Physics, ISBN -7354-0346-5/06; Vol. 852 Ebert, E.E., 1987. A pattern recognition technique for distinguishing surface and cloud types in polar regions. Journal of Climate and Applied Meteorology, 26: 1412-1427. Elliot, N.C., Kieckhefer, R.W., Lee, J.H., French, B.W. 1999. Influence of within-field and landscape factors on aphid predator populations in wheat. Landscape Ecology 14:239- 252. FAO (1998). Crop evapotranspiration. Guidelines for computing crop water requirements. Irrigation and drainage Paper no.56, Food and Agriculture Organization., Rome Frannz, E., Gebhardt, M.R. Unklesby, K.B. 1995. Algorithms for extracting leaf boundary information from digital images of plant foliage, Transactions of the ASAE, 38(2):625-633. Ebert, E.E. and Le Marshall, J.F., 1995. An evaluation of infrared satellite rainfall estimation techniques over Australia. Australian Meteorological Magazine, 44: 177-190. Friedl, M.A. 1997. Examining the effects of sensor resolution and sub-pixel heterogeneity in spectral vegetation
indices: implications for biophysical modelling. In: Quattrochi, D.A. and Goodchild, M.F., eds, Scale in Remote Sensing and GIS. Boca Raton, CRC Press Inc., pp. 113--139. Gamon, J.A., Peuelas, J. and Field, C.B., 1992. A Narrow-waveband spectral index that tracks diurnal changes in photosynthetic efficiency. Remote Sensing of Environment, 41: 35-44. Gardner, B.R., Nielsen, D.C. Shock, C.C., 1992a. Infrared thermometry and the crop water-stress index: 1. History, theory, and base-lines, Journal of Production Agriculture, 5(4):462-466. Gardner, B.R., Nielsen, D.C. Shock, C.C., 1992b. Infrared thermometry and the crop water-stress index: 2. Sampling procedures and interpretation, Journal of Production Agriculture, 5(4):466-475. Gausman, H.W., Menges, R.M., Richardson, A.J, Walter, H.R. Rodriguez, R., Tamez, S. 1981. Opticalparameters of leaves of 7 weed species, Weed Science, 29(1):24-26. Gitelson, A., Kogan, F., Spivak, L., Zakarin, E. and Lebed, L., 1996. Estimation of Seasonal Dynamics of Pasture and Crop Productivity in Kazakhastan Using NOAA/AVHRR Data, Proceedings of International Geoscience and Remote Sensing Symposium (IGARSS), pp. 209-211. Grassotti, C.G. and Garand, L., 1994. Classification-based rainfall estimation using satellite data and numerical forecast model fields. Journal of Applied Meteorology, 33: 159-178. Grody, N.C., 1991. Classification of snow cover and precipitation using the Special Sensor Microwave Imager. Journal of Geophysical Research, 96: 7423-7435. Hall, F.G., Shimabukuro, Y.E. and Huemmrich, K.F. 1995. Remote sensing of forest biophysical structure using reflectance decomposition and geometric reflectance models. Ecological Applications, 5, 993--1013. Hanks, J. E.. Beck J. L. 1998. Sensor-controlled hooded sprayer for row crops. Weed Technol. 12:308-314. Hatfield, J.L., 1981. Spectral behaviour of wheat yield variety trials, Photogrammetric Engineering & Remote Sensing, 47:1487-1491. Hatfield, J.L., Vauclin, M., Vieira, S.R., Bernard, R. 1984c. Surface temperature variability patterns within irrigated fields, Agricultural Water Management, 8:429-437. Hatfield,J.L., Pinter Jr.,P.J. Remote sensing for crop protection. Crop Protection 12(6):403 414 (1993). Hendricksen, B.L., 1986. Reflections on drought: Ethiopa 1983-1984. International Journal of Remote Sensing, 7: 1447-1451. Herman, A., Kumar, V.B., Arkin, P.A. and Kousky, J.V., 1997. Objectively determined 10-day Africian rainfall estimates created for famine early warning systems. International Journal of Remote Sensing, 17: 21472159. Hickman, M.V., Everitt, J.H., Escobar, D.E., Richardson, A.J. 1991. Aerial-photography and videography for detecting and mapping Dicamba injury patterns. Weed Technology, 5(4):700-706. Hobbs, T.J. 1997. Atmospheric correction of NOAA-11 NDVI data in the arid rangelands of Central Australia. International Journal of Remote Sensing, 18, 1051--1058. Honghui, L., Xiaohuan, Y. and Naibin, W. 1999. Remote sensing based estimation system for winter wheat yield in North China Plain. Chinese Geographical Journal, 9, 40--48. Howell, T.A., Hatfield, J.L..Yamada, H Davis, K.R. 1984a. Evaluation of cotton canopy temperature to detect crop water stress, Transactions of the ASAE, 27(1):84:88. Hunsaker,D.J., Pinter Jr.,P.J., Barnes,E.M., Kimball,B.A. Estimating cotton evapotranspiration crop coefficients
with a multispectral vegetation index. Irrigation Science 22:95-104 (2003) Idso, S.B., 1982. Non-water-stressed baselines – A Key to measuring and interpreting plant water-stress, Agricultural Meteorology,27 (1-2):59-70. Idso, S.B., Jakson, R.D., Reginato, R.J. 1977a. Remote-sensing for agicultural water management and crop yield production, Agricultural Water Management, 1(4):299-310. Idso, S.B., Jakson, R.D. Reginato, R.J. 1977b. Remote-sensing of crop yields. Science, 196(4285):19-25. Idso, S.B., Jackson, R.D., Pinter, P.J., Jr., Reginato, R.J. and Hatfield, J.L., 1981. Normalizing the stress-degreeday parameter for environmental variability. Agricultural Meteorology, 24: 45-55. Jackson, R.D., M.S. Moran, L.W. Gay, and L.H. Raymond, 1987. Evaluating evaporation from field crops using airborne radiometry and ground-based meteorological data, Irrigation Science, 8(2):81-90. Jackson, R.D., Idso, S.B., Reginato, R.J. and Pinter, P.J.J., 1981. Canopy temperature as a crop water stress indicator. Water Resources Research, 17: 1133-1138. Jackson, R.D., 1982. Soil Moisture Inferences from Thermal Infra-red Measurements of Vegetation Temperatures. IEEE Transactions on Geoscience and Remote Sensing, GE-20(3): 282-286. Jupp, D.L.B., Strahler, A.H. and Woodcock, C.E. 1989a. Autocorrelation and regularization in digital images I. Basic theory. IEEE Transactions on Geoscience and Remote Sensing, 26, 463--473. Jupp, D.L.B., Strahler, A.H. and Woodcock, C.E. 1989b. Autocorrelation and regularization in digital images II. Simple image models. IEEE Transactions on Geoscience and Remote Sensing, 27, 247--258. Jupp, D.L.B. and Walker, J. 1996. Detecting structural and growth changes in woodlands and forests: The challenge for remote sensing and the role of geo-optical modelling. In: Gholz, H.L., Nakane, K. and Shimoda, H., eds, The Use of Remote Sensing in the Modeling of Forest Productivity. Dirdrecht, The Netherlands, Kluwer Academic Publications, pp. 75--108. Kaufman, Y.J. and Tanre, D. 1992. Atmospherically Resistant Vegetation Index (ARVI) for EOS-MODIS. IEEE Transactions on Geoscience and Remote Sensing, 30, 261--270. Kniverton, D.R., Barrett, E.C. and Dodge, J.C., 1997. The development of composite alogrithms for global rainfall estimation using data from the DMSP SSM/I. International Journal of Remote Sensing, 18: 517534. Kogan, F.N., 1990. Remote Sensing of weather impacts on vegetation in non-homogeneous areas. International Journal of Remote Sensing, 11: 1405-1420. Kogan, F.N., 1995b. AVHRR Data for Detection and Analysis of Vegetation Stress, Proceedings of The 1995 Meteorological Satellite Data Users' Conference, Winchester, United Kingdom,, pp. 155-162. Kogan, F.N., 1995c. Droughts of the late 1980s in the United States as Derived from NOAA Polar-Orbiting Satellite Data. Bulletin of the American Meteorological Society, 76(655-668). Leone, A.P., Menenti, M., Sorrentino, G. 2001. Reflectance Spectometry to Study Crop Response to Soil Salinity. Ital. J. Agron., 4,2, 75-85. Leprieur, C., Kerr, Y.H. and Pichon, J.M. 1996. Critical assessment of vegetation indices from AVHRR in a semiarid environment. International Journal of Remote Sensing, 17, 2549--2563. Li, H., Lascano, R.J., Barnes, E.M., Booker, J. L., Wilson, T., Bronson, K.F. Segarra E. 2001. Multispectral reflectanceof cotton related to plant growth, soil water and texture,and site elevation. Agron. J.
93:1327–1337 Liang, S. and Strahler, A.H. eds. 2000. Land Surface Bidurectional Reflectance Distribution Function (BRDF); Recent Advances and Future Prospects, Harwood Academic Publishers. Lillesand, T. M., Kiefer, R. W., 1994. Remote sensing and Image Interpretation, 3rd edition. Wiley. New York. Liu, G. and Curry, J.A., 1992. Retrieval of precipitation from satellite microwave measurement using both emission and scattering. Journal of Geophysical Research,
97:
9959-9974.
Liu, W.T. and Kogan, F.N., 1996. Monitoring regional drought using the Vegetation Condition Index. International Journal of Remote Sensing, 17: 2761-2782. Lozana-Garcia, D.F., Fernandez, R.N., Gallo, K.P. and Johannsen, C.J., 1995. Monitoring the 1988 severe drought in Indiana, USA using AVHRR data. Int ernational Journal of Remote Sensing, 16: 1327-1340. Lu, H., Raupach, M.R. and McVicar, T.R. 2001. Decomposition of Vegetation Cover into Woody and Herbaceous Components Using AVHRR NDVI Time Series. Canberra, Australia, CSIRO Land and Water, pp. 44. Ma, B. L., Morrison, M. J., Dwyer, L. M. 1996. Canopy Light Reflectance and Field Greenness to Assess Nitrogen Fertilization and Yield of Maize. Agron. J. 88: 915-920. Maas, S.J., 1993a. Within-season calibration of modelled wheat growth using remote-sensing and field sampling, Agron. J.,85(3):669-672. Maas, S. J., 1993b. Parameterized model of gramineous crop growth: II. Within-season simulation calibration. Agron. J..85, 354-358. Maas, S.J., 1988. Use of Remotely-Sensed Information in Agricultural Crop Growth Models. Ecological Modelling, 41: 247-268. Mahan, J.R., Upchurch, D.R. 1988. Maintenance of constant leaf temperature by plants. 1. Hypothesis limited homeothermy, Environmental and Experimental Botany, 28:351-357. Maselli, F., Conese, C., Petkov, L. and Gilabert, M. 1992. Use of NOAA-AVHRR NDVI data for environmental monitoring and crop forecasting in the Sahel. Preliminary Results. International Journal of Remote Sensing, 13, 2743--2749. McVicar, T.R. and Jupp, D.L.B. (1998) The current and potential operational uses of remote sensing to aid decisions on Drought Exceptional Circumstances in Australia: A Review. Agricultural Systems. 57, 399468. McVicar, T.R., Jupp, D.L.B. and Williams, N.A. 1996b. Relating AVHRR vegetation indices to LANDSAT TM leaf area index estimates. Canberra, ACT, CSIRO, Division of Water Resources, pp. 33. McVicar, T.R., Walker, J., Jupp, D.L.B., Pierce, L.L., Byrne, G.T. and Dallwitz, R. 1996c. Relating AVHRR vegetation indices to in situ leaf area index. Canberra, ACT, CSIRO, Division of Water Resources, pp. 54. McVicar, T.R. and Bierwirth, P.N. 2001. Rapidly Assessing the 1997 Drought in Papua New Guinea using Composite AVHRR Imagery. International Journal of Remote Sensing, 22, 2109--2128. McVicar, T.R., Zhang, G., Bradford, A.S., Wang, H., Dawes, W.R., Zhang, L. and Lingtao, L. 2000. Monitoring Regional Agricultural Water Use Efficiency for Hebei Province on the North China Plain. Australian Journal of Agricultural Research. 53, 55-76.
McVicar, T.R., Davies, P.J., Qinke, Y. and Zhang, G. (2002) An Introduction to Temporal--Geographic Information Systems (TGIS) for Assessing, Monitoring and Modelling Regional Water and Soil Processes. McVicar, T.R., Rui, L., Walker, J., Fitzpatrick, R.W. and Changming, L. (eds.) 2003. Regional Water and Soil Assessment for Managing Sustainable Agriculture in China and Australia. ACIAR
Monograph
84,
pp
205-223.
http://www.eoc.csiro.au/aciar/book/PDF/Monograph_84_Chapter_16.pdf McVicar, T.R. and Jupp, D.L.B. (2002) Using covariates to spatially interpolate moisture availability in the Murray-Darling Basin: a novel use of remotely sensed data. Remote Sensing of Environment (Invited paper in a Special Issue: Biophysical Remote Sensing). 79, 199-212. McVicar, T.R. and Jupp, D.L.B. (1999) Estimating one-time-of-day meteorological data from standard daily data as inputs to thermal remote sensing based energy balance models. Agriculture and Forest Meteorology. 96, 219-238. McVicar, T.R., Van Niel, T.G., Li, L.T., Hutchinson, M.F., Mu, X.M. and Liu, Z.H. (2007) Spatially Distributing Monthly Reference Evapotranspiration and Pan Evaporation Considering Topographic Influences. Journal of Hydrology. (In Press), http://dx.doi.org/10.1016/j.jhydrol.2007.02.018 Mitchell, R.M. 1999. Calibration status of the NOAA AVHRR solar reflectance channels: CalWatch Revision 1. CSIRO Atmospheric Research Technical Paper No. 42, Melbourne, pp. 20. Mo, X.G., Liu, S. X., Lin, Z.H., Xu, Y.Q. and McVicar, T.R. (2005). Prediction of crop production, water consumption and water use efficiency with a SVAT-crop growth model using remotely sensed data on the North China Plain. Ecological Modelling. 183, 301-322. Moran, M. S., Inoue, Y., Barnes, E. M. 1997. Opportunities and Limitations for image-Based Remote Sensing in Precision Crop Management. Remote Sens. Environ. 61: 319-346. Moran, M. S., Maas, S. J., Pinter, P. J. Jr., 1995. Combining remote sensing and modelling for estimating surface evaporationand biomass production. Remote Sens. Rev.. 12, 335-353. Moran, M.S., Clarke, T.R., Inoue, Y. and Vidal, A., 1994. Estimating crop water deficit using the relation between surface-air temperature and spectral vegetation index. Remote Sensing of Environment, 49: 246-263. Moran, M.S., Pinter, P.J. Jr., Clothier, B.E. Allen, S.G. 1989a.Effect of water-stress on the canopy architecture and spectral indices of irrigated alfalfa, Remote Sens. Environ. 29:251-261. Moran, M.S., Jackson, R.D., Raymond, L.H. Gay, L.W., Slater, P.N. 1989b. Mapping surface-energy balance components by combining Landsat thematic mapper and ground-based meteorological data, Remote Sens. Environ. 30(1):77-87. Moran M.S., Jackson R.D. (1991). Assessing the spatial distribution of evapotranspiration using remotely sensed inputs. Journal Environmental Quality, 20, 725-737 Muller, E., James, M., 1994. Seasonal variation and stability of soil spectral patterns in a fluvial landscape. Int. J. Remote Sens.. 15, 1885-1900. Nemani, R.R. and Running, S.W. 1989. Testing a theoretical climate-soil-leaf area hydrologic equilibrium of
forests using satellite data and ecosystem simulation. Agricultural and Forest Meteorology, 44, 245--260. Oliva R.N., Steiner J.J., Young W.C., III. Red clover seed production: I. Crop water requirements and irrigation time. Crop Sci. 1994;34:178-184 Penuelas, J., Filella, I. 1998. Visible and Near-Infrared Reflectance Techniques for Diagnosing Plant Physiological Status. Trends in Plant Science Vol. 3, No 4: 151-155. Peters, A.J., Rundquist, D.C. and Wilhite, D.A., 1991. Satellite detection of the geographic core of the 1988 Nebraska drought. Agricultural and Forest Meteorology, 57: 35-47. Petty, G.W., 1995. The Status of Satellite-Based Rainfall Estimation Over Land. Remote Sensing of Environment, 51: 125-137. Pinter, P.J. Jr., Ritchie, J.C., Hatfield, J.L. Hart, G.F. 2003. The Agricultural Research Service’s remote sensing program: An example of interagency collaboration, Photogrammetric Engineering and Remote Sensing, 69(6):615-618. Pinter, Jr., P.J., R.D. Jackson, Idso, S.B., Reginato R.J. 1981. Multidate spectral reflectances as predictors of yield in water stressed wheat and barley. Int. J. Remote Sens. 2(1):43- 48. Pinter, P.J., Stanghellini, M.E., Reginato, R.J., Idso, S.D., Jenkins, A.D. and Jackson, R.D., 1979. Remote Detection of biological stresses in plants with infrared thermometry. Science, 205: 585-587. Prata, A.J., Caselles, V., Coll, C., Ottle, C. and Sobrino, J. 1995. Thermal remote sensing of land surface temperature from satellites: Current status and future prospects. Remote Sensing Reviews, 12, 175--224. Quattrochi, D.A. and Luvall, J.C. 1999. Thermal infrared remote sensing for analysis of landscape ecological processes: methods and applications. Landscape Ecology, 14, 577--598. Railyan, V. Ya., Korobov, R. M., 1993. Red edge structure of canopy reflectance spectra of Triticale. Remote Sens. Environ.. 46, 173-182. Rasmussen, M.S. 1998a. Developing simple, operational, consistent NDVI-vegetation models by applying environmental and climatic information: Part I: Assessment of net primary productivity. International Journal of Remote Sensing, 19, 97--117. Rasmussen, M.S. 1998b. Developing simple, operational, consistent NDVI-vegetation models by applying environmental and climatic information: Part II: Crop yield assessment. International Journal of Remote Sensing, 19, 119--139. Rasmussen, E.M. and Arkin, P.A., 1992. Observing Tropical Rainfall from Space: a review. In: J.S. Theon, T. Matsumo, T. Sakata and N. Fugono (Editors), The Global Role of Tropical Rainfall. Deepak, Hampton, VA, pp. 105-188. Raun, W.R., Solie, J.B., Johnson, G.V. Stone, M.L., Mullen, R.W., Freeman, W.E. Thomason, and E.V. Lukina. 2001. Improving nitrogen use efficiency in cereal grain production with optical sensing and variablerate application. Agron. J. 94:815–820. Reed, B.C., 1993. Using remote sensing and Geographic Information Systems for analysing landscape/drought interaction. International Journal of Remote Sensing, 14: 3489-3503. Reginato, R.J., and J. Howe, 1985. Irrigation Scheduling using crop indicators Journal of Irrigation and Drainage Engineering. ASCE, 111(2):125-133.
Richardson, A.J., Meneges, R.M. Nixon, P.R. 1985. Distinguishing weed from crop plants using video remotesensing. Photogrammetric Engineering & Remote Sensing, 51(11):1785-1790. Ridao E., Conde, J. R. Inés Mìnguez, M. 1998. Estimating fAPAR from Nine Vegetation Indices for Irrigated and Nonirrigated Faba Bean and Semileafless pea Canopies. Remote Sens. Environ.. 66, 87-100. Roderick, M.L., Chewing, V. and Smith, R.C.G. 2000. Remote sensing in vegetation and animal studies. In: ‘t Mennetje, L. and Jones, R.M., eds, Field and Laboratory Methods for Grassland and Animal Production Research. Wallingford UK, CABI, pp. 205--225. Rosenfield, D. and Gutman, G., 1994. Retrieving microphysical properties near the tops of potential rain clouds by multispectral analysis of AVHRR data. Atmospheric Research, 34: 259-283. Schepers, J.S.,Francis, D.R. 1998. Precision agriculture– What's in our future. Commun. Soil Sci. Plant Anal.29:1463–1469. Schepers, J.S., Francis, D.R., Vigil, M., Below, F.E. 1992. Comparison of corn leaf nitrogen concentration and chlorophyll meter readings, Commun. Soil Sci. Plant Anal. 23(17-20):2173-2187. Schepers, J.S., Blackmer, T.M., Wilhelm, W.W., Resende, M. 1996. Transmittance and reflectance measurements of corn leaves from plants with different nitrogen and water supply. Journal of Plant Physiology, 148(5):523-529. Schmidt M., King E.A., and McVicar, T.R. (2006) A user-customized Web-based delivery system of hypertemporal remote sensing datasets for Australasia. Photogrammetric Engineering and Remote Sensing. 72, 1073-1080. Sellers, P.J. 1985. Canopy reflectance, photosynthesis and transpiration. International Journal of Remote Sensing, 6, 1335--1372. Serrano, L., Filella, I., Penuelas, J. 2000. Remote Sensing of Biomass and Yield of Winter Wheat under Different Nitrogen Supplies. Crop Sci. 40: 723-731. Shanahan, J. F., Schepers, J. S., Francis, D.D., Varvel, G. E., Wilhelm, W. W., Tringe, J. M., Schelemmer, M. R., Major, D. J. 2001. Use of Remote Sensing Imagery to Estimate corn Grain Yield. Agron. J. 93: 583-589. Smith, R.C.G., Adams, J., Stephens, D.J. and Hick, P.T. 1995. Forecasting wheat yield in a Mediterranean-type environment from the NOAA Satellite. Australian Journal of Agricultural Research, 46, 113--125. Spencer, R.W., Goodman, H.M. and Hood, R.E., 1989. Precipitation Retrieval over land and ocean with the SSM/I: Identification and characteristics of the scattering signal. Journal of
Atmospheric and Ocean
Technology, 6: 254-273. Thenkabail, P.S., Ward, A.B., Lyon, J.G. and Merry, C.J. 1994. Thematic Mapper vegetation indices for determining soybean and corn growth parameters. Photogrammetric Engineering and Remote Sensing, 60, 437--442. Tian, L., J.F. Reid, Hummel, J.W. 1999. Development of a precision sprayer for site-specific sorghum growth and yield variations using air-borne multispectral digital imagery. Trans. ASAE, 43(6):1927-1938. Todd, M.C., Barrett, E.C. and Beaumont, M.J., 1996. Estimation of Daily Rainfall at the Pixel Scale over the Upper Nile River Basin Using Meteosat Infrared Satellite Images, Proceedings of the 8th Conference on Satellite Meteorology and Oceanography, Atlanta, Georgia, pp. 282-286. Tucker, C.J., B.N. Holben, Jr., J.H. Elgin, and J.E. McMurtrey III, 1981. Remote sensing of total dry matter
accumulation in winter wheat, Remote Sensing of Environment, 13:461. Tucker, C.J. 1979. Red and photographic infrared linear combinations for monitoring vegetation. Remote Sensing of Environment, 8, 127--150. Van Niel T.G., McVicar, T.R., and Datt, B. (2005) On the relationship between training sample size and data dimensionality:
Monte Carlo analysis of broadband multi-temporal classification. Remote
Sensing of Environment. 98, 468-480. Van Niel T.G. and McVicar T.R. (2004a) Determining temporal windows of crop discrimination with remote sensing: a case study in south-eastern Australia. Computers and Electronics in Agriculture. 45, 91-108. Van Niel, T.G. and McVicar, T.R. (2004b). Current and potential uses of optical remote sensing in rice-based irrigation systems: a review. Australian Journal of Agricultural Research. 55, 155-185. Van Niel T.G. and McVicar T.R. (2003) A simple method to improve field-level rice identification: Toward operational monitoring with satellite remote sensing. Australian Journal of Experimental Agriculture. 43, 379-387. Van Niel, T.G. and McVicar, T.R. (2002). Experimental evaluation of positional accuracy estimates of a linear network using point- and line-based testing methods. International Journal of Geographic Information Science. 16, 455-473. Van Niel, T.G. and McVicar, T.R. (2001) Assessing Positional Accuracy and its Effects on Rice Crop Area Measurement: An Application at Coleambally Irrigation Area. Australian Journal of Experimental Agriculture. 41, 557-566. Walker, J., Jupp, D.L.B., Penridge, L.K. and Tian, G. 1986. Interpretation of vegetation structure in Landsat MSS imagery: a case study in disturbed semi-arid eucalypt woodlands. Part 1. field data analysis. Journal of Environmental Management, 23, 19--33. Wang, D., Poss, J.A.,.Donovan, T.J, Shannon, M.C., Lesch, S.M. 2002b. Biophysical properties and biomass production of elephant grass under saline conditions, Journal of Arid Environments, 52(4):447-456 Weymouth, G. and Le Marshall, J.F., 1994. An Operational System to Estimate Insolation Over the Australian Region. In: J.F. Le Marshall and J.D. Jasper (Editors), The Technical Proceedings of the 2nd Pacific Ocean Remote Sensing Conference, PORSEC '94,, Melbourne, Australia, pp. 443-449. Wiegand, C. L., Richardson A. J., Escobar, D. E., Gerbermann, A. H., 1991. Vegetation Indices in Crop Assesments. Remote Sens. Environ.. 35, 105-119. Wilheit, T., Adler, R., Avery, S., Barrett, E.C., Bauer, P., Berg, W., Chang, A., Ferriday, J., Grody, N., Goodman, S., Kidd, C., Kniveton, D., Kummerow, C., Mugnai, A., Olson, W., Petty, G., Sabata, A., Smith, E. and Spencer, R., 1994. Algorithms for the retrieval of rainfall from passive microwave measurements. Remote Sensing Reviews, 11: 163-194. Willers, J.L., Seal, M.R. Luttrel, R.G. 1999. Remote sensing, line-intercept sampling for tarnished plant bugs (Heteroptera: Miridae) in Mid-South cotton, Journal of Cotton Science, 3:160-170. Woodcock, C.E. and Strahler, A.H. 1987. The factor of scale in remote sensing. Remote Sensing of Environment, 21, 311--332. Woodcock, C.E., Strahler, A.H. and Jupp, D.L.B. 1988a. The use of variograms in remote sensing: I scene
models and simulated images. Remote Sensing of Environment, 25, 323--348. Woodcock, C.E., Strahler, A.H. and Jupp, D.L.B. 1988b. The use of variograms in remote sensing: II real digital images. Remote Sensing of Environment, 25, 349--379. Wright, D. L. Jr, 2003. Using Remote Sensing to Manage Wheat Grain Protein. Final report for Utah State University Affiliated Research Ceneter, Earth Science Applications Directorate NASA Science Enterprise. Pp. 1-8. Zadoks, J.C., Chang, T.T., Konzak., C.F. 1974. A Decimal Code for the Growth Stages of Cereals. Weed Research 14:415-421 Zipoli, G., Grifoni, D., 1994. Panicle contribution to bidirectional reflectance factors of a wheat canopy. Int. J. Remote Sens.. 15, 3309-3314.
