Figure 1: Relationship of ground measured LAI on wheat fields with IRS LISS-III derived NDVI at Indore (Madhya Pradesh, India) (Pandya et al., 2003)

1:1 Line

LAI derived from LISS III

Figure 2: Comparison between LISS-III derived LAI and MODIS LAI at Indore (MP, India) (source: Pandya et al., 2003)

Rastogi et al. (2000) tested Price model on farmers fields during 199697 season in Karnal (Haryana, India) and 1997-98 in Delhi using IRS LISS-III data and estimated wheat attenuation coefficients. The root mean square error (RMSE) between RS estimates and ground measured LAI ranged between 0.78-0.87 when LAI was in the range of 1-4, while for higher LAI range (46), the RMSE varied from 1.25 to 1.5 in two sites. Such errors can severely reduce utility of a model using field-level LAI as input.

Crop simulation models

Crop simulations models are based on physical plant processes and simulate the effects of change in growing environment on plant growth and development on a daily basis. A crop simulation model is a simple representation of a crop and is explanatory in nature. The processes essentially modeled are phenology, photosynthesis and dry matter production, dry matter partitioning, in simulation models aimed at potential production. Those aiming at crop-specific behaviour include modules for phyllochron, branching pattern and potential flowers/ grain filling sites. The response to water and nutrition limited environment is added by introducing models of soil water balance and uptake and transpiration by crop, and nitrogen transformations in soil, uptake and remobilization within plant, respectively. Models of effects of weeds and pests are being developed and could be available in new generation of crop simulation models.

In dynamic crop simulation models, three categories of variables recognized are, state, rate and driving variables. The state variables are quantities like biomass, amount of nitrogen in soil, plant, soil water content, which can be measured at specific times. Driving variables, or forcing functions, characterize the effect of the environment on the system at its boundaries, and their values must be monitored continuously, e.g., meteorological variables. Each state variable is associated with rate variables that characterize their rate of change at a certain instant as a result of specific processes. These variables represent flow of material or biomass between state variables. Their value depends on the state and driving variables according to rules that are based on knowledge of the physical, chemical and biological processes that take place during crop growth.

Under the International Benchmark Sites Network for Agrotechnology Transfer (IBSNAT) project a computer software package called the Decision Support System for Agrotechnology Transfer (DSSAT) was developed which integrates 11 crop simulation models (CERES cereal, CROPGRO legume and other models) with a standardized input and output (Jones, 1993) and has been evaluated/ used in a number of countries. Use of CERES-Wheat included in DSSAT for regional wheat yield prediction has been demonstrated recently in India (Nain et al, 2004).


Introduction to GIS

Burrough and McDonnell (1998) has defined GIS as a powerful set of tools for collecting, storing, retrieving at will, transforming and displaying spatial data from the real world for a particular set of purposes. The three major components of GIS are (i) computer hardware, (ii) computer software and (ii) digital geographic data. The information stored within a GIS is of two distinct categories. The spatially referenced information that can be represented by points, lines, and polygons, that are referenced to a geographic coordinate system and is usually stored in either raster (grid-cell) or vector (arc-node) digital format. The second category of information stored in a GIS is attribute data or information describing the characteristics of the spatial feature.

Using RS & GIS for crop monitoring

The use of GIS along with RS data for crop monitoring is an established approach in all phases of the activity, namely preparatory, analysis and output. In the preparatory phase GIS is used for (a) stratification/zonation using one or more input layers (climate, soil, physiolgraphy, crop dominance etc.), or (b) preparing input data (weather, soil and collateral data) which is available in different formats to a common format. In the analysis phase use of GIS is mainly through operations on raster layers of NDVI or computing VI profiles within specified administrative boundaries. The final output phase also involves GIS for aggregation and display of outputs for defined regions (e.g., administrative regions) and creating map output products with required data integration through overlays.

Wade et al. (1994) described efforts within National Agricultural Statistics Service (NASS) of U.S. Department of Agriculture (USDA) of using NOAA AVHRR NDVI for crop monitoring and assessment of damage due to flood and drought by providing analysts a set of map products. Combining satellite data in a GIS can enhance the AVHRR NDVI composite imagery by overlaying State and county boundaries. The use of raster-based (grid-cell) capabilities of ARC/INFO (GRID) for the generation of difference image helps compare a season with previous year or average of a number of years. Overlaying a crop mask helps in highlighting only effects on crops. Application of frost isolines is made to help analysts to locate average dates of the first frost for possible crop damage. Generation and overlay of contours of precipitation data generated using TIN function of ARC/INFO also is an aid to interpreting NDVI difference image.

Interfacing crop simulation models to GIS

Crop simulation models, when run with input data from a specific field/ site, produce a point output. The scope of applicability of these simulation models can be extended to a broader scale by providing spatially varying inputs (soil, weather, crop management) and policy combining their capabilities with a Geographic Information System (GIS). The main purpose of interfacing models and GIS is to carry out spatial and temporal analysis simultaneously as region-scale crop behaviour has a spatial dimension and simulation models produce a temporal output. The GIS can help in spatially visualizing the results as well as their interpretation by spatial analysis of model results.

While GIS and modeling tools have existed for so long, the integration, including the conceptual framework is being given attention only recently. Hartkamp et al. (1999) have reviewed GIS and agronomic modeling and suggested that 'interface' and 'interfacing' be used as umbrella words for simultaneously using GIS and modeling tools, and 'linking', 'combining' and 'integrating' as suitable terminology for degree of interfacing. These correspond to loose, tight and embedded coupling, respectively, as used by Burrough (1996) and Tim (1996). While there is a continuum between linking and combining, the terms are explained below:

(a) Linking: Simple linkage strategies use GIS for spatially displaying model outputs. A simple approach is interpolation of model outputs. An advanced strategy is to use GIS functions (interpolation, overlay, slope, etc.) to produce a database containing inputs of the model and model outputs are also exported to the same database. Communication between GIS and model is through identifiers of grid cells or polygons in input and output files, which are transferred in ascii or biary format between GIS and model (Figure 3a). Such an approach is not able to utilize full potential of the system and suffers from limitation due to (a) dependence on formats of GIS and model, (b) incompatibility of operating environments and (c) not fully utilizing the capabilities of GIS.

(b) Combining: Combining also involves processing data in a GIS and displaying model results, however, the model is configured with GIS and data are exchanged automatically. This is done with facilities in GIS package of macro language, interface programmes, libraries of user callable routines (Figure 3b). This requires more complex programming and data management than simple linking. Example of combining is AEGIS (Agricultural and Environmental GIS) with ArcView (Engel et al., 1997).

(c) Integrating: Integration implies incorporating one system into the other. Either a model is embedded in GIS or a GIS system is included in a modeling system. This allows automatic use of relational database and statistical packages (Figure 3c). This requires considerable expertise, effort and understanding of the two tools.

Calixte et al. (1992) developed a regional agricultural decision support system, known as Agricultural and Environmental Geographic information System (AEGIS) that uses the DSSAT capabilities within ARC/INFO GIS for regional planning and productivity analysis. AEGIS allows the user to select

Figure 3: Organizational structure for (a) linking, (b) combining and (c) integrating GIS and crop models (Hartkamp et al., 1999)

various combinations of crop management practices over space and evaluate potential crop production. Engel et al. (1997) modified the AEGIS into AEGIS/WIN (AEGIS for Windows) written in Avenue, an object-oriented macro scripting language, which links the DSSAT (Version 3) with the geographical mapping tool ArcView-2. Thornton et al. (1997a) developed spatial analysis software for the most recent release of the DSSAT, Version 3.1. This software standardized the links between crop models and GIS software and this allowed developers to make use of whatever GIS software is most suitable for a particular purpose, while ensuring that basic links to the DSSAT system and the crop models are the same. The spatial analysis software has two modules: (i) a geostatistical module to interpolate maps and produce probability surfaces from a network of data points, and (ii) a utility that calculates agronomic and economic output statistics from model simulations and maps the results as polygons. Another effort is development of a SPATIAL-EPIC linked to Arc/Info (Satya et al, 1998).

Demonstrated applications of CSM interfaced with GIS

Current wide range of applications of interfacing of GIS and modelling are reviewed by Hartkamp et al. (1999) and covers spatial yield calculation (regional and global), precision farming, climate change studies, and agro-ecological zonation, etc.

CGMS (Crop Growth Monitoring System) of MARS (Monitoring Agriculture with Remote Sensing)

This project of European Union uses WOFOST model and Arc/Info for operational yield forecasting of important crops (Meyer-Ro ux and Vossen, 1994). The Crop Growth Monitoring System (CGMS) of the MARS integrates crop growth modelling (WOFOST), relational batabase ORACLE and GIS (ARC/INFO) with system analytical part for yield forecasting (Bouman et al, 1997). There are databases on soil, weather, crop, and yield statistics that cover the whole of EU. The system-analytical part consists of three modules: agrometeorological module, a crop growth module and a statistical module. The meteorological module takes care of the processing of daily meteorological data that are received in real time to a regular grid of 50x50 km for use as input by crop growth model or for assessment of 'alarm' conditions. The crop growth module consists of the dynamic simulation model WOFOST in which crop growth is calculated and crop indicators are generated for two production levels: potential and water-limited. In CGMS, WOFOST is run on a daily bases for each so-called 'simulation unit', i.e. a unique combination of weather, soil, and crop (mapping) units. In the statistical module, crop indicators (total above ground dry weight and dry weight storage organs) calculated with WOFOST are related to historical yield statistics through regression analysis in combination with a time-trend, for at least 15 years of simulated and historical data (Vossen, 1995). The resulting regression equations per crop per region are used to make actual yield forecast. CGMS generates on a 10 day and monthly basis three types of output on current cropping season: (i) Maps of accumulated daily weather variables on 50x50 km grid to detect any abnormalities, e.g. drought, frost, (ii) Maps of agricultural quality indicators based on comparison of simulated crop indicators with their long-term means, (iii) Maps and tables of yield forecasts.

Precision Farming

Han et al. (1995) developed an interface between PC ARC/INFO GIS and SIMPOTATO simulation model to study potato yield and N leaching distribution for site-specific crop management (precision farming) in a 50 ha field. The GIS input layers, corresponding to important distributed input parameters for the model, were irrigated water/N layer, soil texture layers and initial soil N layers. For each unique sub-area stored in the GIS database, the interface program extracts the attribute codes of that sub-area from the GIS database, converts the attribute codes to the input parameters of the SIMPOTATO and sends them to the model. After running the model, the interface program retrieves the output data (potato yield and N leaching), converts them to the attribute codes and stores the output data in the GIS database.

Agro-ecological Zonation

Aggarwal (1993) used WTGROWS to simulate potential and water-limited wheat yields for 219 weather locations spread all over the country. The district boundaries (as polygons) and model input parameters of soil, weather stations and agro-ecological regions were stored in ARC/INFO GIS. The model outputs of potential and rainfed productivity were stored in GIS as polygon attribute data. Based on potential and rainfed productivity, the districts were classified into 10 iso-yield zones and represented as map using GIS.

Evaluating Agricultural land use options

Aggarwal et al. (1998) studied the agricultural land use option for the state of Haryana using symphonic use of expert knowledge, simulation modelling, GIS and optimization techniques. The study area was divided into agro-ecological land units by overlaying maps of soil, soil organic carbon and climatic normal rainfall in raster GIS IDRISI. The original soil mapping units based on 19 soil properties were reclassified based on soil texture, level and extent of salinity and sodicity, slope and ground water depth. The organic carbon and normal rainfall maps were generated by inverse square interpolation of observed data points followed by segmentation. The CSM for specific crops have been linked to GIS layers of administrative boundaries, physiographic features, climate, soil and agroclimatic zones and GCM outputs to study effect of future climatic changes on crop potential / productivity (Bacsi et al., 1991; Carter and Saarikko, 1996).

Was this article helpful?

0 0

Post a comment