Web-Based Ecological Decision Support System
Rabi H. Mohtar and Tong Zhai
Associate Professor and Research Associate, Purdue University
A reoccurring theme in the discussion of agriculture and food production in the coming century is sustainable development. This depends on integrated natural resource management with simultaneous optimization of agro-eco-production systems on economic, environmental, and social terms. Informatics and computer simulation models have become indispensable tools in dealing with diverse objectives in the decision making process. This paper describes our efforts in building a web-based ecological modeling system and implementation challenges. This overview illustrates the relevant technical platforms for state-of-the-art computing resources that can inform better management decisions.
Accompanying the increasing globalization of world affairs, the prospect of food security and the environment has taken and will be at the center stage in the coming several decades. A quick survey of the current and projected situations regarding agriculture, population, and water resources has been published by the United Nation FAO.1 It reveals an ominous combination of future high population, dwindling arable land, and increasing environmental pollution in most of the developing world. The prospect is more hopeful for developed countries where long-term, systematic research and extensive efforts to improve productivity and to mitigate negative environmental impacts have taken root. From across the globe, sustainable development is seen as a necessary trend that conserves land, water, and biodiversity.
In this paper, we will explore the use of computer simulation models and related information technologies that aid in the assessment, analysis, and management of agro-eco-production systems for improved productivity and fewer environmental hazards. Following an overview of the informatics and models used in natural resource management, we present a web-based hydrologic and hydraulic modeling package that is designed to bring the current generation of hydrologic models, which are utilized in most of the existing decision support systems, up to date. The features, potential usage, and future development of this web-based modeling system, in the larger context of the use of models in facilitating management, will be presented.
Informatics in Environmental and Natural Resources
Considering the highly variable nature of biological and environmental events and the need for using spatially explicit computer models in the decision making process, data must be collected from multiple locations, following carefully designed spatial patterns, and for extended periods of time. Such large volumes of data are typically stored in a relational database managed by one of the popular relational database management systems (RDBMS) such as Oracle, MySQL, MSSQL Server, and ACCESS. The use of the internet has been revolutionizing all the main components of informatics in natural resource management. In a survey of the use of digital
media in managing and delivering information in the current electronic age, Wagner (2003) showed that the use of traditional desktop databases and CD-ROMs has been in decline during the past decades, while the use of World Wide Web (WWW) internet interfaced data collection and retrieval has seen exponential and continues to grow. The most common form of organizing and delivering spatial data using WWW is the linkage of a Web-based Geographical Information System (Web-GIS) and a backend relational database. Users with a web browser, can view, query, and manipulate the presented dataset to create a new dataset, or to access more information through a geo-referenced query delivered through the Web-GIS interface to the backend database. The Web-GIS mediated data retrieval and data organization greatly improved the usability of scientific data and research findings by decision-makers. The advantage of Web-GIS data portal is that it gives structure to the otherwise disaggregated collection of data and facts. In addition to the basic on-line GIS functions of search and overlay, it also provides statistical summaries and administrative functions. Web-GIS is seen as a way for promoting grassroots monitoring, data collection, and public involvement in environmental management. One example of such use of a web-based mapping program in river health monitoring and management is described by Graham et al. (2004). All stakeholders in the catchment of interest were provided with the opportunity to be involved in water sampling and data input into a web-based mapping and scoring system that calculates key indicators for pollution and can be used by management personnel for improved environmental management.
Natural resources and their management are intrinsically complex due to the dynamic balance and interactions among coexisting biotic agents and their abiotic environments within an ecosystem. Computer models have long been used for research on complex ecosystems, forming the basis for an integrated, system approach to management planning and implementation guidance. In an essay that discusses the future generation of ecological models for the purpose of environmental protection in natural resources management, Linthurst et al. (1999) outlined the direction of more advanced future modeling system developed in the form of a must-have list. The focus of the discussion was on modeling-mediated, watershed water quality control. The overall recommendations, which followed the ecological modeling research within the EPA’s Office of Research and Development (ORD), were for: (1) a common software framework for ecological modeling to improve model(s) usability to aid in making management decisions and (2) further develop and improve watershed-scale (multi-scale) modeling to address more realistically the fate of multiple pollutants in multiple environmental media. For individual model improvement, the vision calls for the development of (1) “State-of-the-science process algorithms and component computational models with flexible scaling to provide problem-solving methodologies that are applicable at multiple geographic and temporal scales”; (2) “State-of-the-science atmospheric, terrestrial, aquatic, and biotic process models and stressors and effects models that predict real-world conditions and their incorporation into a common framework; (3) “Improved ability to interconnect one system with another system (e.g., the atmosphere and surface water ecosystems) and exchange information in between;” and (4) the linkage of ecological models to geographic information system (GIS) technology. To a large degree, the current generation of ecological models has become much more sophisticated than their predecessors in terms of dealing with coexisting biological species or pollutant agents and representation of underlying processes in hydrology, soil erosion, and nutrient cycling. However, dynamic temporal and spatial scaling of individual models is still a largely unaddressed issue. Yet, from the recommendations listed above, it is obvious that a collection of specialized models
that each operates at its own optimized scale could form an integrated modeling system, which, when coupled with algorithms to interface across the disparate scales, could be a solution to this problem. Nonetheless, the current generation of ecosystem models is mostly landscape models that can be easily applied to the most common scale in the majority of real applications, i.e. watershed scale. Hence, the scaling issue isn’t seen as urgent, at least for now. Noticeably, the above list of future developments didn’t anticipate the increasingly important WWW platform for model development and, especially, deployment. The combination of Web-GIS and ecological modeling is becoming a trend for a future generation of modeling- mediated natural resources management. Such a trend is evidenced by more and more successful integration of Web-GIS and modeling in various natural resource management applications and by the ever more powerful computers and increasing household access to broadband internet connections. Specifically, the main use of GIS in ecological modeling is the explicit spatial representation of a management unit, which could be a field, catchment/watershed, basin, or region, in increasing order of geographic magnitude. A large proportion of the current generation of ecological models is field-based, focusing within the definitive horizontal limit of edge of field, and the vertical boundary of bottom of root zone. Therefore, problems arise with the process of scaling in attempting to project field based simulation to larger geographical areas where there is much more management diversity.
In dealing with the simulation of a biological community of plants and animals, traditional ecological models tend to assume spatial homogeneity over a large management unit (watershed or basin) at a given time. Current watershed or the larger basin scale models have adopted the grid cell system to represent the spatial diversity within these geo-management units. However, the interactions among the cell elements remain predominantly those of overland runoff and groundwater flow, and subsequently eroded sediment, nutrient, and other harmful chemicals. The interaction among biological agents across cells has not been adequately represented, making it difficult to predict the spatial and temporal distribution of plant or animal communities. This is evidenced in the work by Zhai et al. (2004) on the simulation of mixed pasture growth. In their work, a pasture was regarded as a management unit where multiple forage species grow in competition for light, water, and nutrient. The species were partitioned with resources based on their relative presence in the field (this is done based on leaf area index, LAI, or biomass). The method has proved to be quite effective in predicting growth of multiple perennial forage species in the pasture. However, it lacks the ability to simulate succession in the mixed pasture of forage species, especially annual species, and their spatial pattern. This leads to further difficulty in predicting realistic grazing animal intake on mixed pasture across a growing season, considering the commonly observed preferential grazing by ruminants.
The typical set up for a web-based DSS generally follows the so-called three-tier design. Tier 1 refers to presentation layers or client interface; tier 2 represents process manager or application logic; and tier 3 are the server resources such as databases, computer simulation models, and web-enabled GIS packages (Figure 1).
Ecological models that are used to power various decision support systems should be continuously improved at the process level as new scientific understanding becomes available. For example, in an effort to adapt the EPIC model to the semiarid Northeast of Brazil, De Barros et al. (2005) further improved the EPIC model by incorporating the floral abortion effect due to dry spells in the season and devising better algorithms for resource partitioning among competing species.
In conclusion, there is a need to design a better approach to summarize various simulation results to form evaluation indices that are pertinent to the context of natural resource management projects, and at the same time are easy for stakeholders to understand. User interface design and model management in a web-based DSS directly impact the usability and adoption of these informatics- and model-based DSSs in real applications. Research such as that done by Xie (2003) on striking a balance between user friendliness and user control in an information retrieval system, and by Chiu et al. (2005) on the theoretical basis for designing better user interface for modeling systems, is needed to improve the applicability of the DSSs.
The Web-Based Ecological Modeling System
Our project aimed to create a centralized, web-based modeling environment where users can have access to hydrologic/hydraulic models and ecological models to facilitate the decision making process in projects related to runoff estimation, contaminated site remediation, reservoir siting, and pasture based production system management. All models have been under continuous development by our research group at the Department of Agricultural and Biological Engineering, Purdue University. The following models are currently included in the web-based modeling system:
2DSTREAM (2-Dimensional Surface Stream Flow Model). A finite-element-based overland flow model, 2DSTREAM was originally developed by Vieux et al. (1990a,b) and modified to account for dynamic time step for a one-dimensional
overland flow kinematic wave solution used in solving the uncoupled sets of overland flow (Jaber and Mohtar, 2002). Watershed input data such as nodal coordinates, elemental slopes, and roughness are read at the beginning of each run. The forcing function re(x) is then calculated. Using the element and force vector information, the system of equations is built, updated, and solved for new flow depth (h) values. The model has been validated for different rainfall and slope conditions (Vieux et al., 1990; Jaber and Mohtar, 2002). http://pasture.ecn.purdue.edu/~water/2dstream/.
SPARG/AIRFIX (Air Sparging and multiphase Solute Transport model). This model is developed as a practical, unsaturated flow and multiphase transport model that can be used for the design and operation of sparging and soil vapor extraction systems (Mohtar et al., 1996; Rahbeh and Mohtar, 2003). The model uses first order kinetics to represent the mass transfer of contaminants across aqueous, gaseous, solid, and NAPL phases. The theoretical basis of the model is based on the fact that soil particles are surrounded by water films which isolate the gaseous and non-aqueous phases. Therefore, the contaminant mass transfer can take place across the aqueous-solid (sorption/de-sorption), aqueous-gaseous (stripping), aqueous-NAPL (dissolution), and gaseous-NAPL (volatilization) interfaces. To simulate the air flow distribution, the model incorporates numerical code derived from the SPARG model developed by Mohtar et al. (1996) which is based on steady state unsaturated flow equations. http://pasture.ecn.purdue.edu/~water/airfix/.
SGRASIM (Silvopasture GRAzing Simulation Model) This model was built on the basis of the multi-species GRASIM model (Zhai et al., 2004a). The SGRASIM model explicitly models tree impact on understory forage growth in terms of light, rainfall interception, soil water, and nitrogen uptake/depletion. The model was field tested in an un-enhanced, mixed pasture under a black walnut tree canopy in Indiana, USA. The model simulated the multi-species forage growth in competition both under tree canopy and in open pasture. Soil water and nitrogen cycling are also explicitly modeled. Key components of the model have been field tested and are under continuous development to incorporate the latest research findings related to key processes in the pasture and grazing-based production systems. A prototype model is available for online application at http://pasture.ecn.purdue.edu/~grasim/ (Mohtar et al., 2000).
Water Harvest AHP: Water Harvesting Structure Siting/Impact analysis using the Analytical Hierarchy Protocol. Water harvesting, or rainfall collection, has been practiced at various scales in dry land and semiarid areas as a means to channel and store scarce rainfall for later use. The locations for water collecting structures are determined by a multitude of factors, which are not easily evaluated and often involve a compromise of interests. The Analytical Hierarchy Process (AHP) provides a systematic approach in conducting multi-criteria analysis and decision making. It allows for the comparison of alternatives based on the quantification of mostly qualitative characteristics of a given watershed/region. A Web-based GIS-hydrologic modeling system was designed to implement a spatial AHP process (El-Awar et al.,
2000) for selecting the most suitable and practical location for building water harvesting reservoirs. An online GIS digitizing tool helps users locate potential watershed, extract spatial data related to hydrologic characteristics to be used as input for hydrologic models, which in turn produces needed runoff estimates from sub watersheds, together with land use and land cover data and expert opinions to produce a single Reservoir Suitability Index (RSI). Ranking of RSI allows the quick determination of suitable locations. The prototype technology is available at http://pasture.ecn.purdue.edu/~water/wh/. Major features and potential application of the web-based tool is presented by Zhai et al. (2004b). http://pasture.ecn.purdue.edu/~water/wh/.
The overall design of the system is shown in Figure 2. It follows the general guidelines for web application design for a modeling system discussed previously (Figure 1). World Wide Web (WWW) interfaces have been constructed for SGRASIM, SPARG/AIRFIX, 2DSTREAM, and Water Harvest AHP. Model web interface was written in Dynamic HTML (dhtml) and server side processing is done with application specific CGI (common gateway interface) scripts written in Perl and C programming languages. To help user’s applications using these web-based models, an online workshop is developed to provide users with modularized classes that combine presentation about model theories and hands-on modeling tutorials and exercises. Users can conduct self-paced education via the URL: http://pasture.ecn.purdue.edu/~water/workshops/, where online audio and video presentations and tutorials are available for free access.
The ultimate capability of the integrated ecological modeling system is described as follows: Users can select their local area using the online GIS digitizer. Local hydrologic, production, and environmental data are then uploaded to the server, customized options are formed as model
inputs and models are executed, and finally, results are presented for user analysis. Ongoing efforts are put forth to strengthen model robustness and to streamline this process. As alluded to above in the discussion of DSSs, the scaling issue is intrinsically related to the level of detail that a modeling system uses to describe certain processes. In this regard, the next phase would be focusing on how to graft the well developed, fully distributed modeling framework to the current web environment. Since almost all ecological models have hydrological cycles as the central link among geographic areas, such discussion will be within the context of adapting the distributed hydrologic modeling to the web. Vieux et al. (2004) provided an up-to-date overview of the physically based distributed hydrologic modeling. They defined the fully distributed models as grid-cell based, or finite element based hydrologic models, such as 2DSTREAM. Based on their description of the approach they used to convert a basin-scale lumped model into a distributed sub-basin lumped model, a bottom-up approach that uses a finite element model to represent each sub-basin, and then, at higher orders of integration, uses a single cell to represent each sub-basin in the transfer and routing functions, is seen as a way to “buildup a watershed-scale process model from small-scale elements such as hillslopes or grids” (Vieux et al., 2004). Such a scheme coincides with the conceptual model for kinematic wave analogy used by the 2DSTREAM model in the current package.
The discretization of a watershed would allow the use of finite element based hydrologic models over large watersheds. When coupled with other ecologic modeling components over such grid-cell based systems, we can truly achieve the much sought scaling capacity to conduct simulations across multiple scales. However, such achievement in the web environment is not without difficulties. First and foremost is the computation overhead for using such a simulation scheme. It is well known that finite element based models typically take a long time to run to completion. Hence, on a large watershed, the waiting time of such a simulation could be too long to allow a client to stay connected to the server. Second, a flexible watershed delineation application is also needed to create the first sub-watershed boundaries as basis for further discretization. Engel et al. (2003) has achieved online watershed delineation using software associated with the MapServer, a free web application that can publish maps in various formats. However, their algorithm only allows delineation of one sub-watershed at a time, based on the user provided initial outlet point. This limitation must be resolved in the future in a well-organized Web modeling environment.
To facilitate user application, an online interactive modeling assistant service is under construction as a continuation of the online workshop described above. The end goal of the service is a web-based tool kit that can provide real time guidance to model users in basic training, data preparation, application monitoring/debugging, and modeling results analysis. The service can be invaluable in both higher education and consulting in agro-eco-production systems and environmental engineering.
The critical challenge for the agricultural industry in the next century will continue to be sustainable food production. As sustainability includes economic, environmental, and social implications, it must also involve more efficient and conscientious use of resources while improving productivity. To meet the ever-increasing demands for food and, simultaneously, prevention of environmental degradation, our role, as educators and agricultural engineers, is to
focus on the integrated conservation and protection of natural resources, including water, land, and air, and to promote research and support of biodiversity.
Computer models have long been used for knowledge integration for research on complex ecosystems. In the past, their use has been mostly confined to academia due to the difficulties in model input preparation, model execution, and result interpretation. The proposed web-based modeling system aims to provide state-of-the-art, science-based agro-ecosystem and environmental computer models to facilitate real decision making processes in production systems and natural resources management.
The web-based paradigm (Figures 1 and 2) is considered as a new generation of application development and deployment. It allows fast model deployment to a wide range of audience/platforms and centralized application development/updates without incurring additional operational or investment costs from the users. It also allows remote training and live assistance in custom model implementation. The model repository included in the current modeling package (Figure 2) is under centralized development, management, and extension to form an open, flexible, and extensible decision support system. It can be quickly reconfigured to fit an individual user’s needs or a particular application.
Future development of comprehensive and yet flexible modeling systems for decision support hinges on the following:
Digital data storage and management using Web-GIS enabled database management systems;
Logical management and organization of web-based modeling systems composed of an ensemble of independent models, each of which functions at a specific spatial and temporal scale, with the end goal of dynamic creation of user oriented applications using appropriate models and data combination;
Web-GIS mediated scaling of modeling systems to achieve a proper level of detailed simulation for specific applications;
Post modeling recommendations based on combined quantitative modeling results and other social and economic factors;
Continued model development at the process level.2
Chiu, C.M., 2005. Applying means-end chain theory to eliciting system requirements and understanding users’ perceptual orientations. Information & Management, 42(3):455-468
De Barros, I; J.R. Williams, T. Gaiser, 2005. Modeling soil nutrient limitations to crop production in semiarid NE of Brazil with a modified EPIC version II: Field test of the model Ecological Modeling, 181(4):567-580
El-Awar, F.A., M.K. Makke, R.A. Zurayk, R.H. Mohtar, 2000. A hydro-spatial hierarchy method for siting water harvesting reservoirs in dry areas. Applied Engineering in Agriculture, 16(4):395-404
Engel, B.A., J.Y. Choi, J. Harbor, and S. Pandey, 2003. Web-based DSS for hydrologic impact evaluation of small watershed land use changes. Computers and Electronics in Agriculture, 39(3):241-249
Graham, P.M., C.W.S. Dickens, and R.J. Taylor, 2004. miniSASS - A novel technique for community participation in river health monitoring and management. African Journal of Aquatic Science, 29(1):25-35
Jaber, F.H. and R.H. Mohtar, 2002. Dynamic Time Step for One-Dimensional Overland Flow Kinematic Wave Solution. Journal of Hydrologic Engineering, 7(3):1-11
Linthurst, R.A., J.H. Novak, R.F. Carousel, and S.F. Hedtke, 1999. Developing the Next Generation of Watershed Risk Assessment and Management Models: Where do we go from here? The International Society for Ecological Modeling (ISEM) online publication, March, 1999 [online] http://www.isemna.org/Pages/Publications/ecomod_mar99.pdf
Mohtar, R. H., L.J. Segerlind, and R.B. Wallace, 1996. Finite element analysis for air sparging in porous media. Fluid/Particle Separation Journal, 9(3):225-239
Mohtar, R.H., T. Zhai, and X.W. Chen, 2000. A world wide web-based grazing simulation model (GRASIM). Computers and Electronics in Agriculture, 29:243-250
Rahbeh, M. and R.H. Mohtar, 2003. Numerical Analysis of field scale air sparging remediation. 2003 ASAE Annual Meeting, Paper number 037041, Riviera Hotel and Convention Center, Las Vegas, Nevada, USA 27- 30 July 2003
Vieux, B.E., V.F. Bralts, L.J. Segerlind, and R.B. Wallace, 1990. Finite element watershed modeling: one-dimensional elements. Journal of Water Resources Planning and Management, 116(6) 803–819.
Vieux, B.E, Segerlind, L. J., and Mohtar, R. H. 1990. Surface/Subsurface Flow Equations Interactions: Identifying Sources of Groundwater Contamination, Report no. 89-G1569-02, Institute of Water Research, Michigan State University, East Lansing, MI. 45 p.
Vieux, B.E., Z.T. Cui, and A. Gaur, 2004. Evaluation of a physics-based distributed hydrologic model for flood forecasting. Journal of Hydrology, 298(1-4):155-177
Wagner, A.B., 2003. Managing Tradeoffs in the Electronic Age. Journal of the American Society for Information Science and Technology, 54(12):1160-1164
Xie, H., 2003. Supporting Ease-of-Use and User Control: Desired Features and Structure of Web-Based Online IR Systems. Information Processing & Management, 39(6):899-922
Zhai, T., R.H. Mohtar, H. Karsten, and M. Carlassare, 2004a. Modeling growth and competition of multi-species pasture system. Transactions of the ASAE, 47(2):617-627
Zhai, T., R.H. Mohtar, J.Y. Choi, B.A. Engel, F.A. El-Awar, H.W. Chung, 2004b. Web GIS and Hydrologic Modeling for Siting Water Harvesting Reservoirs. 2004 ASAE Annual Meeting, Paper number 042149, Fairmont Chateau Laurier, The Westin, Government Centre, Ottawa, Ontario, Canada, 1-4 August 2004.