HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS		QUICK SEARCH:   [advanced] Author: Keyword(s): Year:  Vol:  Page:

This Article Abstract Figures Only Full Text (PDF) Free Alert me when this article is cited Alert me if a correction is posted Services Similar articles in this journal Similar articles in ISI Web of Science Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via ISI Web of Science (33) Citing Articles via Google Scholar Google Scholar Articles by Hartkamp, A.D. Articles by Hoogenboom, G. Search for Related Content PubMed Articles by Hartkamp, A.D. Articles by Hoogenboom, G. Agricola Articles by Hartkamp, A.D. Articles by Hoogenboom, G. Related Collections Remote Sensing Other Models

SIMULATION & MODELING

Interfacing Geographic Information Systems with Agronomic Modeling

A Review

^a Natural Resources Group, CIMMYT, Lisboa 27, Apartado Postal 6-641, 06600 México, DF, Mexico
^b Dep. of Biological & Agricultural Engineering, Univ. of Georgia, Griffin, GA 30223-1797 USA

d.hartkamp{at}cgiar.org

	ABSTRACT

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
Agronomic models are traditionally used for point or site-specific applications due to limitations in data availability as well as computer technologies. Interfacing geographic information systems (GIS) with agronomic models is attractive because it permits the simultaneous examination of spatial and temporal phenomena. The objective of this review is to examine strategies for interfacing GIS with agronomic models. It considers the diverse terminology in use, programming approaches, issues of data and scale, and existing applications. Linking is defined as merely passing input and output between a GIS and a model, combining is defined as automatic data exchange and GIS tool functions, and integrating is defined as embedding a model in a GIS or vice versa. Due to differences in research objectives, spatial and temporal scales, data sources or formats, and the natural processes being modeled, there is no universal approach for interfacing. Because of the detailed input requirements for agronomic models, expanding the models from a point-based application to a spatial application can greatly increase the volume of input data. This review suggests that a major challenge in interfacing GIS to models lies in developing systems that handle spatial processes by implying interactions among spatial units. Moreover, extensive data requirements must be satisfied, while also ensuring data quality control.

Abbreviations: ascii, American Standard Code for Information Interchange • GIS, geographic(al) information system(s)

	INTRODUCTION

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
GEOGRAPHICAL INFORMATION SYSTEMS (GIS) facilitate the storage, manipulation, analysis, and visualization of spatial data. Most process-based agronomic models have examined temporal variation using point data from specific sites, and model outputs thus are site-specific. Agriculture is a spatial activity, however, and there is growing interest in placing site-specific information in a spatial and long-term perspective. Precision agriculture requires models that calculate spatial variation in crop growth at a scale of meters and with a time scale appropriate for management decisions, often hours or days (NRC, 1997). Efficient targeting of germplasm or production practices requires models that calculate germplasm x environment interactions on a regional scale (e.g., 1 to 5 km), usually through modeling at daily time scales (e.g., Chapman and Barreto, 1996). Climate change research calculates global effects with models that are often run on scales of 50 to 100 km, on the basis of multicentury time scales (NSTC, 1998). Furthermore, there is increasing interest in understanding how processes with a spatial component, such as runoff and lateral flow of solutes, affect system behavior. The interaction of both spatial and temporal issues seems best handled through interfacing agronomic models with GIS.

Geographical information systems have existed for almost three decades, but only in the last 10 years have applications been widely used in agriculture and natural resource management (Burrough, 1986). In the 1980s, the number of applications grew as a result of vendor-driven efforts to show the capabilities of GIS (Suan-Pheng, 1993), and vendors' perceptions of the market guided the development of these applications (Dangermond, 1991). During the 1990s, as access to powerful computer technology became less costly, the number of GIS applications specific for research and development has increased. Consequently, a new generation of problems and issues have surfaced that are more pertinent to researchers and particular research objectives than to GIS developers per se (Suan-Pheng, 1993).

An example of such a new issue is the adding of time as a fourth dimension to GIS capabilities. The time dimension can be included in GIS analyses in two ways. In the first approach, time-series of historic data from surveys or remote sensing can be examined as a series of overlays (Marble, 1984). These static spatial snapshots may be analyzed with the help of statistical procedures (Croft and Kessler, 1996), such as Markov chains (Stoorvogel, 1995; Tomlinson Associates, 1987). Such analyses can document past trends, but their predictive power is weak, especially for new production practices or conditions. The second approach, that avoids this shortcoming by using process-based models to represent variation with time, is emphasized in this paper. The resulting model outputs may be viewed as a time series in GIS.

Use of the words `model' and `modeling' in relation to GIS can cause confusion. Firstly, the focus of this paper is on simulation modeling, as opposed to spatial and environmental modeling. Spatial modeling often refers to techniques such as reclassification, overlay, and interpretation (Yakuup, 1993). Environmental modeling refers to techniques ranging from interpolating climate data to the use of data models and remote sensing. These techniques do not relate to simulation modeling per se, although environmental modeling in the narrow sense also exists (e.g., simulations of groundwater flow and the fate of contaminants) (Maslia et al., 1994). Nonetheless, spatial modeling can be used to facilitate interfaces between GIS and modeling. Secondly, this paper focuses on process-based models concerned with agricultural issues (e.g., crop production, soil erosion, or water pollution), as opposed to rule-based (logical) and empirical (regression) models.

The main attraction of interfacing models and GIS is to facilitate simultaneous analysis of spatial and temporal variation in processes. Our understanding and interpretation of the simulation results can not only significantly improve by spatially visualizing the results of models (Engel et al., 1997) but, more importantly, improve by advanced spatial analyses of model results (Campbell et al., 1989; Stoorvogel, 1995). Relevant methods include multivariate analysis, spatial autocorrelation, cluster analysis to define homogeneous zones prior to modeling, point pattern analysis, and error analyses.

Despite the growing number of computer-based applications, little attention has been paid to developing conceptual frameworks for the simultaneous use of GIS and modeling. The objective of this review is to examine strategies for interfacing GIS with agronomic models. We consider the diverse terminology in use, concepts of interfacing, and issues of data, scale, and error. Examples of applications in agronomy and natural resource management are discussed, including extraneous major challenges to effective interfacing.

	Strategies for Interfacing

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
Models have been interfaced with GIS since the mid-1980s, but early efforts did not emphasize process-based models (Nyerges, 1991). Nyerges (1991) noted that GIS vendors have had few incentives to develop such complex models, because of their limited market potential. In the past, therefore, GIS–model interfaces were developed within the various research disciplines in an ad hoc manner by researchers who were not professional GIS programmers (Stoorvogel, 1995). Because of these circumstances, a conceptual framework with standards for terminology, formats, and procedures for interfacing models with GIS does not exist.

Terms frequently used in describing systems that interface GIS and models, and their definition (Longman, 1984), include the following:

• Interface: The place at which diverse (independent) systems meet and act on or communicate with each other.
  
• Link: To connect.
  
• Couple: To link together; the act of bringing together.
  
• Combine: To unite, to merge.
  
• Integrate: To unite, combine, or incorporate into a larger unit; to end segregation.
  

We suggest that interface and interfacing be used as umbrella words for the simultaneous use of GIS and modeling tools, since they do not imply a specific level of interaction between them.

We consider linking, combining, and integrating to be suitable terminology for degrees of interfacing. Burrough (1996) and Tim (1996) refer to "loose coupling," "tight coupling," and "embedded coupling," which correspond to linking, combining, and integrating, respectively. Fedra (1993) uses "deep coupling," which corresponds to integrating. Distinguishing between linking and combining can be difficult, while integration is more easily distinguished (Tim, 1996). The terms `linking', `combining', and `integrating' relate to the physical extent to which the GIS and models are interfaced.

Linking
Simple linkage strategies use GIS for spatially displaying model outputs. This approach often involves interpolation of model outputs (e.g., White and Hoogenboom, 1995). More sophisticated linkage strategies use GIS functions such as interpolation, overlay, and slope calculation to produce a database containing inputs for the model. Model outputs can be exported to the same or a separate database. Communication between the software systems is achieved through grid cell or polygon identifiers that link input and output to field locations. Simple transfer of files in ascii format or a common binary file format is usually sufficient in this strategy. The concept of linking GIS and models is presented in Fig. 1a . Limitations of this strategy often include (i) the system's dependence on either the GIS or model output format; (ii) failure to take full advantage of the functional capabilities of the GIS (e.g., spatial analysis tools); and (iii) the incompatibility of operating environments and hardware (Tim, 1996). Lam et al. (1996) and Fedra (1991) have emphasized that users cannot exploit the full potential of the systems through linking. Examples of linking are GLEAMS to ArcInfo (Stallings et al., 1992), USLE to MAP GIS (Hession and Shanholtz, 1988), and WOFOST to ArcInfo (Van Laanen et al., 1992) (Table 1) .

View larger version (26K): [in this window] [in a new window]   Fig. 1 Organizational structure for (a) linking, (b) combining, and (c) integrating geographical information systems (GIS) and models. (Adapted from Tim, 1996, and Tim and Jolly, 1994.)

Integrating
Integration implies incorporating one system into the other. Either a model is embedded in a GIS, or a simple GIS system is included in a modeling system. Aside from making use of GIS and modeling tools, integration usually involves automatic use of relational databases, expert systems, and statistical packages. Full integration implies systems developed within the same or similar data structures. File transfer and format conversions are avoided or automated and thus are invisible to the user. The development of such systems may mean starting from scratch with data organization, among other tasks. A considerable programming effort is needed to develop these software systems, not to mention a considerable mutual understanding between the GIS specialist and modeler, and so only limited attempts have been made to integrate process-based models with GIS. More often integrated systems make use of simplified models (Tim, 1996). The concept of integrating GIS with models is presented in Fig. 1c. Examples of integrating are RAISON (Lam and Swayne, 1991; Lam et al., 1996) and the interface described by Stuart and Stocks (1993) (Table 1).

Additional examples of interfaces that have been linked, combined, and integrated are presented in Table 1. Abbreviations of the model names and interface tools are listed alphabetically in Table 2 . In summary, limitations of the different strategies are related to problems of incompatibility of database structures, software, and hardware (Stoorvogel, 1995; Tim, 1996; Burrough, 1996). Linking strategies usually underuse the functional capabilities of GIS to achieve an interactivity between the GIS and models. Point models are run only for a series of locations, and there is no attempt to consider interaction between neighboring locations, such as run-off or run-on in adjacent plots. In combining and integrating, interactivity can be more readily achieved. Almost all interfacing activities require considerable effort from the developers and users (Engel et al., 1997). The ease of use, efficiency, development and maintenance costs, and necessary human resource training are important considerations for system design (Fedra, 1991; Nyerges, 1991). The amount of effort needed to develop integrated systems is large, and probably for this reason most efforts at interfacing have evolved through linking models with GIS. Stoorvogel (1995) noted that a modular approach should contribute to the transparency and flexibility of structure and procedures.

	Structural Issues Affecting the Interface Strategy

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
The physical extent to which modeling and GIS capabilities are interfaced can be viewed as a programming issue. However, there are structural issues that affect the interface strategy. These are related to the research problem or the purpose of the application and include the scale, type (linear, nonlinear), and complexity of the processes modeled and of the data sources, the format and structure of the available data, and the dynamic relations between model runs and the spatial units.

Scale and Complexity
The spatial scale of the research problem may range from the plot to the field, farm, watershed, region (intranational), nation, region (international), continent, and/or global level. The temporal scale can vary from seconds to several years or more. Related scales in interfacing include the data measurement scale, original map and GIS scale, modeling scale, data manipulation scale, natural scale of the phenomenon, and scale of application (Burrough, 1996).

Issues of scale for the GIS component of an interfaced system are straightforward. The map scale is often predefined. For instance, map scales between 1:100 000 and 1:250 000 are often recommended for regional studies, whereas scales of 1:1000 and 1:2500 are more appropriate for farm-level applications (Garrity and Singh, 1991). Unfortunately, use of detailed map scales is frequently precluded by practical constraints, such as poor data availability or inadequate computer resources. Wilson et al. (1996), Inskeep et al. (1996), and Wagenet and Hutson (1996) found that the impact of input map (data) resolution on final model calculation depends on the processes accounted for by the model.

The scale of a model is more problematic and involves three closely linked dimensions: space, time, and complexity (Penning de Vries, 1996). Model scale is often equated with the complexity of the processes that are modeled. However, this is incorrect. For example, a model of global climate patterns may simulate complex atmospheric processes but operate at a spatial resolution of 50 km or larger. Simple models of physiological processes may run on a time scale of minutes. Traditional point-based agronomic models lack an explicit spatial scale, although it is often suggested that they are valid at the plot or field scale. However, when such a model is interfaced to spatial data in a GIS, the spatial scale is usually predetermined by the scale of the spatial data or application. The temporal scale of a model may be influenced by the time span of the types of processes being modeled, data availability, computational constraints or scale of the application.

Model complexity is largely determined by the type (e.g., linear or nonlinear) and detail of the processes represented. However, it can also be influenced by data availability, computational constraints, and interests in making underlying assumptions readily understandable.

The choice of an appropriate model scale is often difficult and is a topic of active debate. Clearly, the answer to the scale and/or complexity issue should lie in the research problem or application objective itself (Boote et al., 1996). Passioura (1996) made the valuable distinction between scientific models, which are intended to improve the understanding of processes, and engineering models, which are intended to provide sound calculations for decision makers. Monteith (1996) noted that the ease of software development afforded by modern personal computers may have led some researchers to develop excessively complex models. However, this view may be contrasted with that of Leenhardt et al. (1995), who argued that, for modeling effects of spatial soil and water variability at a regional scale, simplification of models to facilitate modeling across large regions or long time scales is unjustified on theoretical grounds. Given equal experimental effort, simple approaches allow a greater spatial sampling density than more mechanistic ones, but simplifying processes can reduce the sphere of validity of the outputs. Furthermore, integrated parameters are often difficult to relate to specific measurable parameters, such as soil texture or leaf area index. Burrough (1989) related the choice of model complexity and sampling density to an economic consideration of investment. Stoorvogel (1995) noted that complex models are often avoided because of limited availability of data.

Model complexity relates not only to the processes modeled and data requirements, but also to computational requirements of the model. A simple model, such as the Universal Soil Loss Equation (USLE) (Wischmeier and Smith, 1978), can be embedded more easily in a GIS than a model with complex computational requirements, such as parallel processing, that are not provided by GIS software and hardware (Burrough, 1996).

Model complexity also affects functionality for end-users (Moore et al., 1993). The tendency to develop complex, physically based models that are difficult for users to understand is liable to grow unless common concepts and terminology are developed. Generic system-integration tools developed under a common conceptual theory have been proposed as a means to reduce the gap between theoretical GIS practitioners and discipline-oriented applications specialists (Stoorvogel, 1995).

Spatial Distribution and Type of Data
The spatial distribution of the available data can influence the strategy for interfacing. Weather stations and soil samples can have an irregular, sparse distribution. Their values can show large variation in space. Before these weather and soil data can be useful for input into a model they may need to be interpolated. Interpolation methods include kriging and cokriging (Krige, 1951; Journel and Huijbregts, 1978), splining (Hutchinson, 1991), and spatial domain methods based on state-space models (Shumway et al., 1989; Wendroth et al., 1992). The choice of spatial interpolators depends on type and distribution of the data and research objective (e.g., DeBrule, 1983).

The type of data, continuous or discontinuous, can also influence the strategy for interfacing. For example, it can be problematic to relate point measurements of discontinuous variables, such as soil taxonomic units, to a final polygon or raster structure unless borders are exactly delineated by measurement. If a variable is discontinuous, point data lose their connection to the polygon or raster unless borders are exactly delineated by measurement. In this case, Monte Carlo simulations can be used to capture the variance, as in the work of Foussereau et al. (1993) on soil variables within discontinuous soil taxonomy units. Also, fuzzy logic can be used to estimate the spatial distribution of soil types and to derive soil properties (e.g., Zhu et al., 1997).

Spatial Data Format
The appropriate spatial data format depends on the type of data or data source. Quantitative data such as climate or soil traits are often provided as interpolated surfaces in raster (gridded) format. Soil taxonomy maps and land-use data are more commonly recorded in vector (polygon) formats. The choice of raster or vector depends on the importance of spatial interactions in the process being studied and how these are handled in the model (Fraisse et al., 1994). Relative advantages and disadvantages of the two different formats are reviewed in basic texts on GIS (e.g., Burrough, 1986). Fortunately, enhancements in recent software have reduced this format incompatibility, and some interfaces can use both formats (e.g., Wilson et al., 1996). The spatial data format has consequences for subsequent analysis, particularly where spatial scales are varied. Raster data (grid cells) can be easily overlaid and can be aggregated (lumped to bigger grid cell sizes) more easily than polygon structures, which have irregular shapes. Furthermore, with polygon structures small `splinter' areas are often formed, which are hard to interpret.

Model Simulations in Relation to Type and Size of Spatial Unit
A model can be run for all spatial units (i.e., grid cells or polygons) in a study area or, to decrease the number of runs, for a subset of the spatial units. The type (interacting or noninteracting) and size of spatial units influence the selection of spatial units for the simulations.

Noninteracting Spatial Units
Noninteracting spatial units are units (grid cells or polygons) whose value does not affect the value of the neighboring unit. If the total study area is small relative to the spatial unit size, simulations may be run on all possible spatial units, and the spatial database may be used as model input without alteration. However, if the study area is large and the spatial units are small (with few classes), simulations may be run for only specific classes of units. Class values that are not farming areas, such as cities or water bodies, can be masked out. Values can be sorted and classified in an intermediate database structure, using multivariate analyses. Studying maize (Zea mays L.) yield potential of East Africa, Collis and Corbett (1997) created what they called effective environments—climate zones that were defined through cluster analysis. Model simulations were conducted for each environment only.

For large databases or small spatial units, a random subset of units can be evaluated (the Monte Carlo method). In a modified Monte Carlo, units are prestratified. For example, variable numbers of simulations are executed for different regions according to their relative importance as production areas or for specific soil types. More simulations may be executed for border cells to reduce edge effects.

Interacting Spatial Units
Spatial units are considered to interact when values of one unit affect the values of neighboring units. For interacting units, a model may have to be run for all spatial units, or else subsampling has to be carefully managed. Furthermore, the order of the simulation runs must be determined prior to running the simulation. In the case of surface run-off, the sequence of the simulations is determined by identifying the flow path over the terrain, as is done in ANSWERS (Rewerts and Engel, 1991) and ZOO (Ioannidis et al., 1997). Given that traditional agronomic models are one-dimensional and essentially ignore horizontal flow when interfaced to a GIS, this potentially important interaction is usually ignored. These applications will remain less useful in areas of variable terrain or topography.

Managing Input and Output
To analyze different scenarios, many simulations may have to be conducted for a single spatial unit. A data and programming issue that arises is how to manage long-term modeling of different inputs for one spatial unit structure. For example, if three irrigation levels and three fertilizer levels are applied to five crops for 20 years of variable soil and weather data, a single spatial unit would have 900 values for each output variable. Fortunately, advances in imaging allow outputs to be viewed as a series of images, either displayed together or as a dynamic view (animation). For example, yearly images could present differences in yield due to treatments, with summary statistics used as aids. Further development of viewing and analysis tools in this area is anticipated.

	Applications

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
In agronomy and natural resource management research, applications of interfaces of GIS and modeling have grown from primarily hydrological applications in the mid-1980s to the current wide range of applications. Ordered roughly by increasing detail of spatial scale, examples can be categorized into groups such as the following.

1. Atmospheric modeling (Lee et al., 1993).
   
2. Climate change, sensitivity and/or variability studies (Rosenzweig, 1990; Wei et al., 1994; Beinroth et al., 1998).
   
3. Agroecological characterization and zonation (Aggarwal, 1995; Bouman et al., 1994).
   
4. Regional risk analysis (Bouman, 1993).
   
5. Scenario modeling and impact assessment, ex ante and also ex post (WRR, 1992; Stoorvogel, 1995; De Koning et al., 1993; Stockle, 1996; Lam, 1993).
   
6. Hydrology, water quality, water pollution (Mamillapalli et al., 1996; Warwick and Haness, 1992; Holloway, 1992; Corwin and Loague, 1996; Kovar and Nachtnebel, 1993; Maidment, 1993).
   
7. Spatial yield calculation—regional, global (Haskett et al., 1995; Van Keulen and Stol, 1995; Karthikeyan et al., 1996).
   
8. Precision farming (spatial yield calculation) (Booltink and Verhagen, 1997a; Engel et al., 1997; Hoogenboom et al., 1993).
   

However, strict borders between the application groups do not exist. For instance, nutrient management, particularly minimizing nitrate leaching, is a cross-cutting theme, especially in Groups 5, 6, and 8. Climate change and variability may be seen as a scenario, but scenario modeling, as defined here, includes scenarios derived from policy goals (e.g., WRR, 1992). Examples of linking the DSSAT family of crop models to GIS at different spatial scales (field to regional) are presented in Table 3 .

Additionally, care must be taken not to assume that the model application remains the same when it is interfaced to GIS. The model name or acronym may remain unchanged as the application evolves. EPIC (Erosion Productivity Impact Calculator) was first developed to examine relations among crop management practices, productivity, and soil erosion. More recently the GIS-interfaced version of the model has been used in climate sensitivity, hydrology, and water quality assessment applications. Consequently, the original acronym has acquired a new meaning, Environmental Policy Integrated Climate (Ramanarayanan, personal communication, 1997). In some cases, the confusion is reduced by giving the GIS–model interface a new name, such as AEGIS (Agricultural and Environmental Geographic Information Systems), which developed within the DSSAT (Decision Support System for Agrotechnology Transfer) framework.

	Challenges for Successful Applications

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
The large number of software systems in which models and GIS have already been interfaced (Table 1) suggests that, while interfacing per se is not a trivial exercise, it is a relatively tractable software engineering problem. In this review, other challenges are apparent. These include developing interfaced systems that achieve interactivity and satisfying data requirements while ensuring data quality control by developing methods for error analysis.

Interactivity
As mentioned above in the section on strategies for interfacing, linking interfaces often does not achieve interactivity, because spatial output is created merely by interpolating point-based simulations. The points themselves are considered independent of their neighbors when the simulations are run. A challenge to interfacing applications is to make the simulations interactive, and so truly achieve spatial modeling. This interactivity between spatial units can be achieved more easily by combining and integrating efforts.

Data Availability and Information Sharing
Because of the detailed input requirements for agronomic models, expanding the models from a point-based application to a spatial application necessarily expands the need for data. Although it is not stated in the interfacing studies, data availability must be enhanced to fully realize the potential of interfacing GIS to models.

Improved data availability may be achieved through use of additional sources, encouraging data format standards, and improving information sharing management systems. Remote sensing is thought to be a potentially important additional source of data in precision agriculture (NRC, 1997) and in county-level yield mapping (Carbone et al., 1996). It has been used to help estimate parameters for model input, such as leaf area index, soil moisture, and surface soil water evaporation (Bouman, 1995; Reiniger and Seguin, 1986; Moran et al., 1995), and to evaluate and validate results of GIS–modeling efforts (Maas, 1993; Bouman, 1995; Booltink and Verhagen, 1997b).

Data availability can also be improved by reducing data incompatibility due to physical storage (format), syntactic organization (conversion, repackaging needs), quality and accuracy, or semantic interpretation. The first two problems can be resolved by standards (Evans, 1994). Efforts to standardize input formats have been promoted for crop modeling by the International Consortium for Agricultural Systems Application (ICASA) (Hunt et al., 1994; Ritchie, 1995; Tsuji et al., 1998). These standards facilitate interchange of data, thereby increasing data availability. Differences in accuracy and semantics are harder to solve and may still inhibit information sharing (Evans, 1994).

Besides technical issues, data availability and information exchange are often affected by organizational, legal, cultural, and bureaucratic factors. There is considerable discussion on whether governments should encourage data distribution on a free or subsidized basis, as opposed to charging the full costs of data collection and distribution. Economic analyses (Porter and Callahan, 1994) suggest that data contributors should receive more benefit than they currently do. Information management policies that increase the credit to the collector and ensure the responsibility of quality and documentation are necessary. However, the benefits of open data access in agricultural systems might include long-term effects on regional economies or on the natural resource base that are difficult to quantify. Porter and Callahan (1994) provided a broad review of other issues related to the organizational, legal, and bureaucratic aspects of data sharing for environmental research.

Error Analysis
Spatial data have errors due to measurement, digitization, or interpolation. Similarly, models, being simplified representations of reality, produce output with error. How these errors interact when systems are interfaced is poorly understood, and so error analysis will become increasingly important as more models are interfaced with GIS. Users become concerned about the reliability and quality of the model outputs (Loague et al., 1998). Error analysis is also useful for assessing optimal combinations of sampling density and model complexity (Leenhardt et al., 1995). Uncertainty analysis is related to error analysis. Quantifying the effects of the uncertainty of variables on modeling can provide an indication of the reliability of the resulting calculations (Bouman, 1993; Corwin and Loague, 1996).

Conventional error propagation theory can be used to assess the quality of modeling results only if they are influenced by random errors. For data or variables stored in a GIS or used as model input, sources of error are usually functions of observations, measurements, and entry. These are random errors. However, some techniques used in GIS, such as logic models (e.g., suitability classes) contain a systematic error of unknown magnitude that error propagation theory cannot address effectively (Drummond, 1987). Burrough (1986) recognized this problem and developed propagation rules for several GIS procedures.

Other attempts at error analysis for GIS have used probability modeling. This approach is problematic, because of the variety of possible spatial data processing procedures and the rigorous requirements of probabilistic data gathering. In a GIS, two major classes of error or uncertainty can be defined, those dealing with positional error (digitizing, georeferencing) and those addressing thematic uncertainty and error. For questions of spatial variability, fuzzy surfaces are used for uncertainty analysis and Monte Carlo methods for error analysis (Davis and Keller, 1996).

In simulation modeling validation (Neelamkavil, 1987), sensitivity analysis and Monte Carlo analyses can help determine error (Bouman, 1993). Bouman (1995) also suggested using remote sensing to reduce uncertainty in modeling efforts.

Methods for analyzing error propagation in GIS and model interfaces are still lacking. Hill et al. (1996) estimated error using an iterative Monte Carlo process for a range of model parameters, grid resolutions, and value estimates where the rules of Burrough (1986) were not applicable. Data resolution and model organization are often changed to interface GIS and models. Error can increase because of the aggregation. De Roo et al. (1989) found that simulations with the GIS-interfaced version of a model calculated 46% more runoff and 36% more erosion than with the original model. Stallings (personal communication 1997) found that aggregated soil data led to a 100% error in model outputs. These results suggest that there is still a poor understanding of how up- and down-scaling influence error propagation, when models are interfaced to GIS.

	Discussion and Conclusions

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
In reviewing existing GIS–model interfaces, this study identified `linking,' `combining,' and `integrating' as suitable terminology for characterizing basic strategies for interfacing agronomic models with GIS. Structural issues such as scale of models and the type, distribution, and scale of data were discussed.

Although there is an increased availability of user interfaces for linking GIS to simulation modeling, there is no guarantee that this improves science. On the contrary, in working with complex interfaces, users have fewer incentives to learn basic concepts, procedures, and the limitations of the underlying systems. Questions on the contribution of complex systems to our problem-solving capacity have been raised in crop modeling research (Passioura, 1996; Sinclair and Seligman, 1996). Both GIS and simulation models have been developed with their own conventions, procedures and limitations. However, linking them at a technical level does not guarantee improved understanding nor useful prediction (Burrough, 1996). There is a danger that calibration, validation, and error analysis will be neglected if GIS–modeling interfaces become too easy to use (Burrough, 1996).

Major challenges lie in achieving full interactivity of a GIS and a model, and in satisfying spatial data requirements while ensuring data quality control through error analysis. Qualitative and subjective procedures are often used for spatial analysis in GIS, and the resulting information loses much of its relevance and statistical validity (Stoorvogel, 1995). More quantitative quality indicators, together with spatial statistics and error analysis, are needed to improve the value of GIS–modeling interfaces.Longman Dictionary of the English Language 1984; National Science and Technology Council. 1998

	NOTES

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 
Presented at Annu. Meet. ASA, 89th, Anaheim, CA, 26–31 Oct. 1997.

Received for publication March 26, 1998.

	REFERENCES

TOP NOTES ABSTRACT INTRODUCTION Strategies for Interfacing Structural Issues Affecting the... Applications Challenges for Successful... Discussion and Conclusions REFERENCES

 

• Aggarwal P.K. Uncertainties in crop, soil, and weather inputs used in growth models: Implications for simulated outputs and their application. Agric. Syst. 1995;48:361-384.
• Beinroth F.H., Jones J.W., Knapp E.B., Papajorgji P., Luijten J. Evaluation of land resources using crop models and a GIS. In: Tsuji G.Y., et al. , ed. Understanding options for agricultural production. Dordrecht, Netherlands: Kluwer Academic Publ, 1998:293-311.
• Biesheuvel A., Hemker C.J. Groundwater modeling and GIS: Integrating MICRO-FEM and ILWIS. In: Kovar K., Nachtnebel H.P., eds. Application of geographical information systems in hydrology and water resources management. Wageningen, Netherlands: Publ. 211. IAHS, 1993:289-296.
• Blaszczynski J. Regional soil loss prediction utilizing the RUSLE/GIS interface. In: Johnson A.I., ed. Geographical information systems (GIS) and mapping: Practices and standards. Philadelphia, PA: Am. Soc. Testing and Materials, 1992:122-131.
• Booltink H.W.G., Verhagen J. Using decision support systems to optimize barley management on spatially variable soil. In: Kropff M.J., et al. , ed. Applications of systems approaches at the field level. Systems Approaches for Sustainable Agricultural Development 6. Dordrecht, Netherlands: Kluwer Academic Publ, 1997:219-233 a.
• Booltink, H.W.G., and J. Verhagen. 1997b. Integration of remote sensing, modeling and field measurements towards an operational decision support systems for precision agriculture. p. 921–929. In J.V. Stafford (ed.) Precision agriculture 1997. vol. 2. Technology, IT and management. Proc. Eur. Conf. on Precision Agriculture, 1st, Warwick, UK. 7–10 Sept. 1997. BIOS Scientific Publ., London.
• Boote K.J., Jones J.W., Pickering N.B. Potential uses and limitations of crop models. Agron. J. 1996;88:704-716.[Abstract/Free Full Text]
• Bouman, B.A.M. 1993. Uncertainty in soil and management parameters in crop growth modeling on a regional scale. p. 58–72. In B.A.M. Bouman, et al. (ed.) Agro-ecology of rice-based cropping systems. SARP Res. Proc. DLO Res. Inst. for Agro-Biology and Soil Fertility, Wageningen Agric. Univ., and IRRI, Manila, Philippines.
• Bouman B.A.M. Crop modelling and remote sensing for yield prediction. Neth. J. Agric. Sci. 1995;43:143-161.
• Bouman, B.A.M., M.C.S. Wopereis, and J.J. Riethoven. 1994. The use of crop growth models in agro-ecological zonation of rice. SARP Res. Proc. DLO Res. Inst. for Agro-Biology and Soil Fertility, Wageningen Agric. Univ., and IRRI, Manila, Philippines.
• Brisson N., King D., Nicoullaud B., Ripoche D., Darthout R. A crop model for land suitability evaluation: A case study of the maize crop in France. Eur. J. Agron. 1992;1:163-175.
• Burrough P.A. Principles of geographical information systems for land resource assessment. New York: Oxford Univ. Press, 1986.
• Burrough P.A. Matching spatial databases and quantitative models in land resource assessment. Soil Use Manage. 1989;5(1):3-8.
• Burrough, P.A. 1996. Environmental modeling with geographical information systems. p. 59–69. In A. Stein, et al. (ed.) Models in action. Proc. Seminar Ser., Quantitative Approaches in Systems Analysis 6. June 1996. DLO Res. Inst. for Agrobiology and Soil Fertility and C.T. de Wit Graduate School for Production Ecology, Wageningen, Netherlands.
• Campbell W.G., Church M.R., Bishop G.D., Mortenson D.C., Pierson S.M. The role for a geographical information system in a large environmental project. Int. J. Geogr. Inf. Syst. 1989;3(4):349-362.
• Calixte J.P., Beinroth F.H., Jones J.W., Lal H. Linking DSSAT to a GIS. Agrotechnol. Transfer 1992;15:1-7.
• Carbone G.J., Narumalani S., King M. Application of remote sensing and GIS technologies with physiological crop models. Photogramm. Eng. Remote Sens. 1996;62:171-179.
• Chapman S.C., Barreto H.J. Using simulation models and spatial databases to improve the efficiency of plant breeding programs. In: Cooper M., Hammer G.L., eds. Plant adaptation and crop improvement. Wallingford, UK: CAB Int, 1996:563-587.
• Cochrane, T.A., D.C. Flanagan, L.D. Norton, and B.A. Engel. 1997. WEPP watershed modeling using ArcView 3.0 (GIS). p. 61. In Agronomy abstracts 1997. ASA, Madison, WI.
• Collis, S.N., and J.D. Corbett. 1997. Interpolation of climate data for application of crop models and GIS. p. 63. In Agronomy abstracts 1997. ASA, Madison, WI.
• Corwin, D.L., and K. Loague (ed.) 1996. Applications of GIS to the modeling of non-point source pollutants in the vadose zone. SSSA Spec. Publ. 48. SSSA, Madison, WI.
• Croft F., Kessler B. Remote sensing, image processing, and GIS: Trends and forecasts. J. For. 1996;4(6):31-35.
• Dangermond, J. 1991. The role of software vendors in integrating GIS and environmental modeling. Keynote 3. In Proc. Int. Conf./Workshop on Integrating Geographical Information Systems and Environmental Modeling, 1st, Boulder, CO. 15–19 Sept. 1991.
• Davis, T., and C.P. Keller. 1996. Modelling and visualizing of uncertainties in environmental assessment using fuzzy logic and Monte Carlo simulation. p. 125–132. In M. Heit, et al. (ed.) Proc. GIS Applications in Natural Resources 2: Papers Presented at Symposia, Vancouver, Canada. 1992–1995. GIS World Books, Fort Collins, CO.
• DeBrule O. Two methods with different objectives: Splines and kriging. Math. Geol. 1983;15:245-257.
• Deckers F. EGIS, a geohydrological information system. In: Kovar K., Nachtnebel H.P., eds. Application of geographical information systems in hydrology and water resources management. Wageningen, Netherlands: Publ. 211. IAHS, 1993:611-619.
• De Koning G.H.J., Janssen M.J.W., Boons-Prins E.R., Diepen C.A.v., Penning de Vries F.W.T. Crop growth simulation and statistical validation for regional yield forecasting across the European Community. Wageningen, Netherlands: CABO-DLO, 1993.
• De Roo A.P.J., Hazelhof L., Burrough P.A. Soil erosion modeling using ANSWERS and geographical information systems. Earth Surface Processes Landforms 1989;14:517-532.
• Desmet P.J.J., Govers G. GIS based simulation of erosion and deposition patterns in an agricultural landscape: A comparison of model results with soil map information. Catena 1995;25:389-401.
• Desmet P.J.J., Govers G. A GIS procedure for automatically calculating the USLE LS factor on topographically complex landscape units. J. Soil Water Conserv. 1996;51(5):427-433.
• Donatelli, M.A., C.O. Stockle, G.S. Campbell, C.O. Gardi, and M. Bitelli. 1997. Regionalized use of CropSyst/GIS to evaluate cropping systems in northern Italy. p. 11. Agronomy abstracts 1997. ASA, Madison, WI.
• Drummond, J. 1987. A framework for handling error in geographic data manipulation. p. 109–118. In W.J. Ripple (ed.) 1989. Fundamentals of geographical information systems: A compendium. Am. Soc. for Photogrammetry and Remote Sensing and Am. Congr. on Surveying and Mapping, Bethesda, MD.
• Engel B.A., Srinivasan R., Rewerts C. A spatial decision support system for modeling and managing agricultural non-point source pollution. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:231-237.
• Engel T., Hoogenboom G., Jones J.W., Wilkens P.W. Aegis/win: A computer program for the application of crop simulation models across geographical areas. Agron. J. 1997;89:919-928.[Abstract/Free Full Text]
• Evans J. Sharing spatial environmental information across agencies, regions, and scales: Issues and solutions. In: Michener W.K., et al. , ed. Environmental information management and analysis: Ecosystem to global scales. London: Taylor & Francis, 1994:203-219.
• Fedra, K. 1991. GIS and environmental modeling. In Proc. Int. Conf./Workshop on Integrating Geographical Information Systems and Environmental Modeling, 1st, Boulder, CO. 15–19 Sept. 1991.
• Fedra K. GIS and environmental modeling. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:35-50.
• Foussereau X., Hornsby A.G., Brown R.B. Accounting for variability within map units when linking a pesticide fate model to soil survey. Geoderma 1993;60:257-276.
• Fraisse C.W., Campbell K.L., Jones J.W., Boggess W.G. Integration of GIS and GLEAMS for alternative dairy waste management analysis. St. Joseph, MI: Pap. 94-205. ASAE, 1994.
• Fresco L.O., Kroonenberg S.B. Time and spatial scales in ecological sustainability. Land Use Policy 1992;9(3):155-168.
• Garrity D.P., Singh V.P. Rice ecosystems analysis for research prioritization. Manila, Philippines: IRRI Workshop on Rice Research Prioritization. Tech. Pap. 91-01. IRRI, 1991.
• Georgiev, G.A., G. Hoogenboom, and K. Ragupathy. 1998. Regional yield estimation using a linked geographic information system, crop growth models and a weather network. p. 135–145. In Proc. Inst. Biol. Eng., 1st. IBE Publ., Athens, GA.
• Goddard, T., L. Kryzanowski, K. Cannon, C. Izaurralde, and T.C. Martin. 1996. Potential for integrated GIS agriculture models for precision farming systems. In Proc. Int. Conf./Workshop on Integrating GIS and Environmental Modeling, Proceedings, 3rd, Santa Fe, NM. 21–25 Jan. 1996 [CDROM computer file]. Natl. Ctr. for Geographic Information and Analysis, Santa Barbara, CA.
• Haskett J.D., Ya.A. Pachepsky, Acock B. Use of the beta distribution for parameterizing variability of soil properties at the regional level for crop yield estimation. Agric. Syst. 1995;48:73-86.
• Hession C.W., Huber K.L., Mostaghimi S., Shanholtz V.O., McCellan P.W. BMP effectiveness evaluation using AGNPS and GIS. St. Joseph, MI: Pap. 89-2566. ASAE, 1989.
• Hession C.W., Shanholtz V.O. A geographic information system for targeting nonpoint source agricultural pollution. J. Soil Water Conserv. 1988;43:264-266.
• Hijmans, R.J., and W.T. Bowen. 1997. A GIS linked simulation model for the potential adoption of drought tolerant potato. p. 23. In Agronomy abstracts 1997. ASA, Madison, WI.
• Hill M.J., Donald G.E., Vickery P.J., Furnival E.P. Integration of satellite remote sensing, simple bioclimatic models and GIS for assessment of pastoral development for a commercial grazing enterprise. Aust. J. Exp. Agric. 1996;36:309-321.
• Hinaman K.C. Use of geographical information systems to assemble input-data sets for a finite difference model of groundwater flow. Water Resour. Bull. 1993;29:401-405.
• Holloway D. Simulation models, GIS and nonpoint-source pollution. Quick Bibliogr. Ser. 1992;92-69:1-34.
• Hoogenboom G., Lal H., Gresham D.D. Spatial yield prediction. St. Joseph, MI: Pap. 93-3550. ASAE, 1993.
• Hoogenboom, G., and P.K. Thornton. 1990. A GIS for agrotechnology transfer in Guatemala. p. 61–70. In Proc. Application of Geographic Information System, Simulation Models, and Knowledge-based Systems for land use management, Blacksburg, VA. 12–14 Nov. 1990. Va. Polytechnic Inst. & State Univ., Blacksburg.
• Hunt L.A., Jones J.W., Hoogenboom G., Godwin D.C., Singh U., Pickering N., Thornton P.K., Boote K.J., Ritchie J.T. General input and output files structures for crop simulation models. In: Uhlir P.F., Carter G.C., eds. Crop modeling and related environmental data: A focus on applications for arid and semiarid regions in developing countries. Paris, France: CODATA, 1994:35-73.
• Hutchinson, M.F. 1991. The application of thin plate smoothing splines to content-wide data assimilation. p. 104–113. In BMRC Res. Rep. Ser. 27. Bur. of Meteorology, Melbourne, VIC, Australia.
• Inskeep W.P., Wraith J.M., Wilson J.P., Snyder R.D., Macur R.E., Gaber H.M. Input parameter and model resolution effects on predictions of solute transport. J. Environ. Qual. 1996;25:453-462.[Abstract/Free Full Text]
• Ioannidis, Y.E., M. Livny, A. Ailamaki, and M.C. Anderson. 1997. ZOO, A management system for complex agronomic models. p. 62. In Agronomy abstracts 1997. ASA, Madison, WI.
• Izaurralde, R.C., W.B. McGill, D.C. Jans-Hammermeister, K.L. Haugen-Kozyra, R.F. Grant, and J.C. Hiley. 1996. Development of a technique to calculate carbon fluxes in agricultural soils at the Eco-district level using simulation models and various aggregation methods. Report to Agriculture and Agri-Food Canada Greenhouse Gas Initiative in fulfillment of Contribution agreement No. 01586-3-1280/01-XSG. Spec. Publ., Dep. of Renewable Resources, Univ. of Alberta, Edmonton, Canada.
• Jordan C., Mihalyfalvy E., Garett M.K., Smith R.V. Modelling of nitrate leaching on a regional scale using a GIS. J. Environ. Manage. 1994;42:279-298.
• Journel A.G., Huijbregts C.J. Mining geostatistics. London: Academic Press, 1978.
• Karthikeyan R., Hoogenboom G., McClendon R.W. Regional yield forecasting using crop models and GIS: A conceptual framework. St. Joseph, MI: Pap. 96-5010. ASAE, 1996.
• Kovar, K., and H.P. Nachtnebel (ed.) 1993. Application of geographical information systems in hydrology and water resources management. Publ. 211. IAHS, Wageningen, Netherlands.
• Krige D.G. A statistical approach to some basic mine valuation problems on the Witwatersrand. J. Chem. Metallurg. Mining Soc. S. Afr. 1951;52:119-138.
• Lal H., Hoogenboom G., Calixte J.P., Jones J.W., Beinroth F.H. Using crop simulation models and GIS for regional productivity analysis. Trans. ASAE 1993;36(1):175-184.
• Lam D.C.L. Combining Ecological modeling, GIS and expert systems: A case study of regional fish species richness model. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:270-275.
• Lam, D.C.L., and D.A. Swayne. 1991. Integrating database, spreadsheet, graphics, GIS, statistics, simulation models and expert systems: Experiences with the RAISON system on micro computers. p. 429–459. In NATO ASI Ser. G26. Springer-Verlag. Heidelberg, Germany.
• Lam, D.C.L., D.A. Swayne, C.I. Mayfield, and D.D. Cowan. 1996. Integration of GIS with other software systems: Integration versus interconnection. In Proc. Int. Conf./Workshop on Integrating GIS and Environmental Modeling, 3rd, Santa Fe, NM. 21–25 Jan. 1996 [CDROM computer file]. Natl. Ctr. for Geographic Information and Analysis, Santa Barbara, CA.
• Lee T.J., Pielke R.A., Kittel T.G.F., Weaver J.F. Atmospheric modeling and its spatial representation of land surface characteristics. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:108-122.
• Leenhardt D., Voltz M., Ramabal S. A survey of several agroclimatic soil water balance models with reference to their spatial application. Eur. J. Agron. 1995;4:1-14.
• Loague K., Corwin D.L., Ellsworth T.R. The challenge of predicting nonpoint source pollution. Environ. Sci. Technol. 1998;32:130A-133A.
• Longman Dictionary of the English Language. 1984. Longman House, Harlow, England.
• Lücke A., Günther P., Diekkruger B., Pestemer W., Richter O. Implementation of a herbicide simulation model in a geographical information system with an example of a site-specific application. Weed Res. 1995;35:333-342.
• Luijten, J.C., and J.W. Jones. 1997. AEGIS+: A GIS-based graphical user-interface for defining spatial crop management strategies and visualization of crop simulation results. p. 23. In Agronomy abstracts 1997. ASA, Madison, Wis.
• Maas S.J. Within season calibration of modeled wheat growth using remote sensing and field sampling. Agron. J. 1993;85:669-672.[Abstract/Free Full Text]
• Maidment D.R. GIS and hydrologic modeling. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:147-167.
• Magnusson B., Söderström M. Combining crop growth models and geographical information systems for agricultural management. Acta Agric. Scand., Sect. B (Soil & Plant Sci.) 1994;44:65-74.
• Mamillapalli, S., R. Srinivasan, J.G. Arnold, and B.A. Engel. 1996. Effect of spatial variability on Basin scale modeling. In Proc. Int. Conf./Workshop on Integrating GIS and Environmental Modeling, 3rd, Santa Fe, NM. 21–25 Jan. 1996 [CDROM computer file]. Natl. Ctr. for Geographic Information and Analysis, Santa Barbara, CA.
• Marble, D.F. 1984. Geographical information systems: An overview. p. 2–8. In W.J. Ripple (ed.) 1987. Geographic information systems for resource management: A compendium. Am. Soc. for Photogrammetry and Remote Sensing and Am. Congr. on Surveying and Mapping, Bethesda, MD.
• Martin, T.C., and H. Neiman. 1996. GISMO: A user interface for linking GIS with the EPIC simulation model. In Proc. Int. Conf./Workshop on Integrating GIS and Environmental Modeling, 3rd, Santa Fe, NM. 21–25 Jan. 1996 [CDROM computer file]. Natl. Ctr. for Geographic Information and Analysis, Santa Barbara, CA.
• Maslia M.L., Aral M.M., Williams R.C., Susten A.S., Heitgerd J.L. Exposure assessment of populations using environmental modeling, demographic analysis, and GIS. Water Resour. Bull. 1994;30(6):1025-1041.
• Mohanty B.P., Van Genuchten M.Th. An integrated approach for modeling water flow and solute transport in the vadose zone. In: Corwin D.L., Loague K., eds. Applications of GIS to the modeling of non-point source pollutants in the vadose zone. Madison, WI: SSSA Spec. Publ. 48. SSSA, 1996:217-234.
• Monteith J.L. The quest for balance in crop modeling. Agron. J. 1996;88:695-697.[Abstract/Free Full Text]
• Moore I.D., Turner A.K., Wilson J.P., Jenson S.K., Band L.E. GIS and land surface–subsurface process modeling. In: Goodchild M.F., et al. , ed. Environmental modeling with GIS. New York: Oxford Univ. Press, 1993:196-230.
• Moran M.S., Maas S.J., Pinter P.J. Combining remote sensing and modeling for estimating surface evaporation and biomass production. Remote Sens. Rev. 1995;12:335-353.
• National Research Council. Precision agriculture in the 21st century: Geospatial and information technologies in crop management. Washington, DC: National Academy Press, 1997.
• National Science and Technology Council Computing, information, and communications: Technologies for the 21st century. Arlington, VA: Natl. Coord. Office for Computing, Information, and Communications, 1998.
• Neelamkavil F. Computer simulation and modeling. Chichester, UK: John Wiley & Sons, 1987.
• Nyerges, T.L. 1991. GIS for environmental modelers: An overview. In Proc. Int. Conf./Workshop on Integrating Geographical Information Systems and Environmental Modeling, 1st, Boulder, CO. 15–19 Sept. 1991.
• Olivieri L.J., Schaal G.M., Logan T.J., Elliot W.J., Motch B. Generating AGNPS input using remote sensing and GIS. St. Joseph, MI: Pap. 91-2622. ASAE, 1991.
• Papajorgji, P., J.W. Jones, J.P. Calixte, F.H. Beinroth, and G. Hoogenboom. 1993. A generic geographic decision support system for estimating crop performance. In Proc. Int. Symp. on Integrated Resource Management and Landscape Modification for Environmental Prediction. ASAE, St. Joseph, MI.
• Papajorgji P., Jones J.W., Peart R.M., Curry B. Using crop models and Geographic information systems to study the impact of climate change in the southeastern USA. Soil Crop Sci. Soc. Fla. Proc. 1994;53:82-86.
• Park S.W., Kim J.T., Lee J.J. A grass interface system for AGNPS model applications. St. Joseph, MI: Pap. 95-3243. ASAE, 1995.
• Passioura J.B. Simulation models: Science, snake oil, education, or engineering?. Agron. J. 1996;88:690-694.[Abstract/Free Full Text]
• Pawitan H. The use of PLANTGRO in forest plantation planning. In: Booth T.H., ed. Matching trees and sites. ACT, Australia: Proc. Int. Workshop. ACIAR. Canberra, 1996:97-106.
• Penning de Vries, F.W.T. 1996. Issue of scale. p. 49–56. In A. Stein, et al. (ed.) Models in action. Proc. Seminar Ser., Quantitative Approaches in Systems Analysis 6. June 1996. DLO Res. Inst. for Agrobiology and Soil Fertility and C.T. de Wit Graduate School for Production Ecology, Wageningen, Netherlands.
• Porter J.H., Callahan J.T. Circumventing a dilemma: Historical approaches to data sharing in ecological research. In: Michener W.K., et al. , ed. Environmental information management and analysis: Ecosystem to global scales. London: Taylor & Francis, 1994:193-202.
• Reiniger R., Seguin B. Surface temperature as an indicator of evapotranspiration or soil moisture. Remote Sens. Rev. 1986;1:227-310.
• Rewerts C.C., Engel B.A. ANSWERS on GRASS: Integrating a watershed simulation with GIS. St. Joseph, MI: Pap. 91-2621. ASAE, 1991.
• Ritchie J.T. International Consortium for Agricultural Systems Applications (ICASA): Establishment and purpose. Agric. Syst. 1995;49:329-335.[ISI]
• Romanowicz R., Beven K., Freer J., Moore R. Topmodel as an application module within GIS. In: Kovar K., Nachtnebel H.P., eds. Application of geographical information systems in hydrology and water resources management. Wageningen, Netherlands: Publ. 211. IAHS, 1993:211-223.
• Rosenberg N.J. Adaptation of agriculture to climate change. Clim. Change 1992;21:385-405.
• Rosenzweig C. Crop response to climate change in the southern Great Plains: A simulation study. Prof. Geogr. 1990;42:20-37.
• Sasowsky K.C., Gardner T.W. Watershed configuration and geographical information systems parametrization for SPUR model for hydrologic simulations. Water Resour. Bull. 1991;27(1):7-18.
• SathyaKumar R., Farell-Poe K.L. Predicting rural municipal nonpoint source pollution using GIS and AGNPS. St. Joseph, MI: Pap. 95-3241. ASAE, 1995.
• Savabi, M.R., L.D. Norton, and B.A. Engel. 1997. Application of WEPP and GIS-GRASS to small watersheds. p. 14. In Agronomy abstracts 1997. ASA, Madison, WI.
• Shaffer M.J., Hall M.D., Wylie B.K., Wagner D.G. NLEAP/GIS approach for identifying and mitigating regional nitrate nitrogen leaching. In: Corwin D.L., Loague K., eds. Applications of GIS to the modeling of non-point source pollutants in the vadose zone. Madison, WI: SSSA Spec. Publ. 48. SSSA, 1996:283-294.
• Shumway R.H., Biggar J.W., Morkoc F., Bazza M., Nielsen D.R. Time- and frequency domain analyses of field observations. Soil Sci. 1989;147:286-298.
• Sinclair T.R., Seligman N.G. Crop modeling: From infancy to maturity. Agron. J. 1996;88:698-704.[Abstract/Free Full Text]
• Singh U., Brink J.E., Thornton P.K., Christianson C.B. Linking crop models with a geographic information system to assist decision making: A prototype for the Indian semiarid tropics. Muscle Shoals, AL: IFDC Pap. 19. Int. Fert. Dev. Ctr, 1993.
• Srinivasan R., Arnold J.G. Integration of a basin scale water quality model with GIS. Water Resour. Bull. 1994;3(3):453-462.
• Srinivasan R., Engel B.A. GIS: A tool for visualization and analysis. St. Joseph, MI: Pap. 92-3610. ASAE, 1991.
• Stallings C., Huffman R.L., Khorram S., Guo Z. Linking GLEAMS and GIS. St. Joseph, MI: Pap. 92-3613. ASAE, 1992.
• Stockle C.O. GIS and simulation technologies for assessing cropping systems management in dry environments. Am. J. Altern. Agric. 1996;11:115-120.
• Stoorvogel J.J. Geographical information systems as a tool to explore land characteristics and land use, with reference to Costa Rica. Wageningen, Netherlands: Ph.D. diss. Wageningen Agric. Univ, 1995.
• Stuart N., Stocks C. Hydrological modeling within GIS: An integrated approach. In: Kovar K., Nachtnebel H.P., eds. Application of geographical information systems in hydrology and water resources management. Wageningen, Netherlands: Publ. 211. IAHS, 1993:319-329.
• Suan-Pheng, K., 1993. Applying GIS technology to natural resource management in developing countries: The realities behind the promises. p. 1–20. In The practice of GIS applications: Problems and challenges, Part 1. Proc. Seminar on GIS and Developing Countries, 2nd. June 1993. GISDECO, Utrecht, Netherlands.
• Thornton P.K., Booltink H.W.G., Stoorvogel J.J. A computer program for geostatistical and spatial analysis of crop model output. Agron. J. 1997;89:620-627.[Abstract/Free Full Text]
• Thornton P.K., Saka A.R., Singh U., Kumwenda J.D.T., Brink J.E., Brisson N. Application of a maize crop simulation model in the central region of Malawi. Exp. Agric. 1995;31:213-226.
• Tiktak A., van der Linden A.M.A., Leine I. Application of GIS to the modeling of pesticide leaching on a regional scale in the Netherlands. In: Corwin D.L., Loague K., eds. Applications of GIS to the modeling of non-point source pollutants in the vadose zone. Madison, WI: SSSA Spec. Publ. 48. SSSA, 1996:259-282.
• Tim U.S. Coupling Vadose zone models with GIS: Emerging trends and potential bottlenecks. J. Environ. Qual. 1996;25:535-544.[Abstract/Free Full Text]
• Tim U.S., Jolly R. Evaluating agricultural non-point source pollution using integrated geographical information systems and hydrologic/water quality model. J. Environ. Qual. 1994;23:25-35.
• Tomlinson Associates Current and potential uses of geographical information systems: The North American experience. In: Ripple W.J., ed. Fundamentals of geographical information systems: A compendium. Bethesda, MD: Am. Soc. for Photogrammetry and Remote Sensing and Am. Congr. on Surveying and Mapping, 1987:165-182 1989..
• Tsuji, G.Y., G. Hoogenboom, and P.K. Thornton (ed.) 1998. Understanding options for agricultural production. Kluwer Academic Publ., Dordrecht, Netherlands.
• Van Keulen H., Stol W. Agro-ecological zonation for potato production. In: Haverkort A.J., MacKerron D.K.L., eds. Potato ecology and modelling of crops under conditions limiting growth. Dordrecht, Netherlands: Kluwer Academic Publ, 1995:357-371.
• Van Laanen H.A.J., van Diepen C.A., Reinds G.J., Koning G.H.J., Bulens J.D., Bregt A.K. Physical land evaluation methods and GIS to explore the crop growth potential and its effects within the European Communities. Agric. Syst. 1992;39:307-328.
• Ventura S.J., Chrisman N.R., Connors K., Gurda R.F., Martin R.W. A land information system for soil erosion control planning. J. Soil Water Conserv. 1988;43:230-233.