Modeling Dynamic Systems Series Editors Matthias Ruth Bruce Hannon This page intentionally left blank Robert Costanza Alexey Voinov Editors Landscape Simulation Modeling A Spatially Explicit, Dynamic Approach With 116 Figures Robert Costanza Alexey Voinov Gund Institute of Ecological Economics University of Vermont Burlington, VT 05405-1708 USA robert.costanza@uvm.edu avoinov@uvm.edu Series Editors: Matthias Ruth Environmental Program School of Public Affairs 3139 Van Munching Hall University of Maryland College Park, MD 20742-1821 USA Bruce Hannon Department of Geography 220 Davenport Hall, MC 150 University of Illinois Urbana, IL 61801 USA CD_ROM included in print edition only Cover illustration: Photographs on the cover by Helena Voinov The SME software included on the CD-ROM is an open-source program Library of Congress Cataloging-in-Publication Data Landscape simulation modeling: a spatially explicit, dynamic approach / editors, Robert Costanza, Alexey Voinov p cm.—(Modeling dynamic systems) Includes bibliographical references ISBN 0-387-00835-7 (hc : alk paper) Landscape ecology—Computer simulation Landscape ecology—United States— Computer simulation—Case studies I Costanza, Robert II Voinov, Alexey III Series QH541.15.L35S63 2003 577Ј.01Ј13—dc21 2003044940 ISBN 0-387-00835-7 Printed on acid-free paper © 2004 Springer-Verlag New York, Inc All rights reserved This work may not be translated or copied in whole or in part without the written permission of the publisher (Springer-Verlag New York, Inc., 175 Fifth Avenue, New York, NY 10010, USA), except for brief excerpts in connection with reviews or scholarly analysis Use in connection with any form of information storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodo-logy now known or hereafter developed is forbidden The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights Printed in the United States of America SPIN 10920435 www.springer-ny.com Springer-Verlag New York Berlin Heidelberg A member of BertelsmannSpringer Science+Business Media GmbH Disclaimer: This eBook does not include the ancillary media that was packaged with the original printed version of the book Series Preface The world consists of many complex systems, ranging from our own bodies to ecosystems to economic systems Despite their diversity, complex systems have many structural and functional features in common that can be effectively simulated using powerful, user-friendly software As a result, virtually anyone can explore the nature of complex systems and their dynamical behavior under a range of assumptions and conditions This ability to model dynamic systems is already having a powerful influence on teaching and studying complexity The books in this series will promote this revolution in “systems thinking” by integrating computational skills of numeracy and techniques of dynamic modeling into a variety of disciplines The unifying theme across the series will be the power and simplicity of the model-building process, and all books are designed to engage the reader in developing their own models for exploration of the dynamics of systems that are of interest to them Modeling Dynamic Systems does not endorse any particular modeling paradigm or software Rather, the volumes in the series will emphasize simplicity of learning, expressive power, and the speed of execution as priorities that will facilitate deeper system understanding Matthias Ruth and Bruce Hannon v This page intentionally left blank Preface Modeling is a simplification, an abstraction of reality It is an analytical tool that we all willingly or unwillingly employ in our everyday life when we build mental, graphical, textual, or other models of reality With the advent of computers, we learned to use them to further our ability to model the world around us, and the more sophisticated and powerful the computers become and the more elaborate the software we develop, the more complexity we can afford in our models It is only during the last decade that spatial dynamic modeling became possible for analysis of large-scale real-life ecosystems The need to model ecosystem dynamics spatially is becoming even more obvious with the widespread use of spatial databases, called Geographic Information Systems Many, if not most, management decisions concerning the environment affect and are affected by the landscape and by how it evolves in space and in time This book puts you at the cutting edge of the science of spatial dynamic simulation modeling The Spatial Modeling Environment (SME) is an open-source software package that we use in conjunction with other software to create, run, analyze, and present spatial models of ecosystems, watersheds, populations, and landscapes We will walk you through the whole process of spatial modeling, starting with the conceptual design, then the formal implementation and analysis, and, finally, interpretation and presentation of the results Numerous applications and case studies will help identify the possible ambiguities and problems that a modeler should be aware of and try to avoid The book will be useful for students and researchers interested in modeling, especially in spatial modeling; it will provide ideas and software tools to translate their local understanding of systems dynamics over space It will be useful for managers and decision makers who want to know what is possible and what is not possible today in spatial modeling and how to treat the uncertainties and insufficiencies in our predictions of spatial dynamics Part I describes the theory and methods of spatially explicit modeling We introduce the Spatial Modeling Environment and demonstrate how it can be used to build simple models We also offer a collection of some basic modules that can be vii viii Preface used off the shelf to build some models of landscapes We then explore the issues of model calibration and analysis Part II is about real applications of the modeling framework that answer some pressing environmental and socio-economic problems What are the impacts of diverse climates and management plans for Barataria and Terrebonne basins in the Mississippi Delta? What can be the effects of sea-level rise? What are the patterns of the dynamic spread of fox rabies across the state of Illinois and what can be the possible disease control strategies? How growing nutrient loading and land use change affect the water quality in the Patuxent River and the Chesapeake Bay? What can be the best restoration and mitigation practices? What are the probabilities of extinction of the vireo and warbler populations, two endangered passerines at Fort Hood, an Army training installation in central Texas? How we allocate crops on a watershed to maximize the agricultural profits, yet minimize the damage to the water quality in the receiving waterways? To what extent has the combination of altered hydrology and water quality degraded the vegetative habitats and other ecological characteristics of the Everglades? What needs to be done to “restore” the Everglades? What is the future of the once abundant eelgrass meadows in the Great Bay Estuary in New Hampshire? What are the impacts of military training across time and space in the Mojave Desert on a federally threatened population of the desert tortoise (Gopherus agassizii) and its habitat? These issues reveal the scope of problems and geographic areas that we cover in these chapters A CD-ROM is provided with the book Here, we offer all of the software that is required to install the SME on your computer (SME currently runs under UNIX, Linux, and Windows; it can be installed under Darwin in Mac OS-X) The Java-based interface is platform independent In addition, we have assembled the Web pages for most of the chapters that give a detailed description of the corresponding projects and offer all of the color graphics, animations, and data to explore and better visualize the possible applications of models for decision-making and management Robert Costanza Alexey Voinov Contents Series Preface v Preface vii Contributors xi Part I Theory and Methods Introduction: Spatially Explicit Landscape Simulation Models ROBERT COSTANZA AND ALEXEY VOINOV Spatial Simulation Using the SME THOMAS MAXWELL, ALEXEY VOINOV, AND ROBERT COSTANZA 21 Modular Ecosystem Modeling ALEXEY VOINOV, CARL FITZ, ROELOF M.J BOUMANS, AND ROBERT COSTANZA 43 Calibration of Large Spatial Models: A Multistage, Multiobjective Optimization Technique 77 FERDINANDO VILLA, ALEXEY VOINOV, CARL FITZ, AND ROBERT COSTANZA ix 316 Ralf Seppelt and Alexey Voinov shows how to work on the optimization problem without any knowledge of an appropriate initial guess of the solution 12.6 Results 12.6.1 Local Optimization We start the analysis of the results with branch in the flowchart of Figure 12.8 Homogeneous control variables for each possible land use type and a certain fertilizer amount are set up: c(z) ϭ c0 ŒL, F(z) ϭ F0 for all z ŒRc Running simulation runs within the Spatial Modeling Environment (SME) for all possible combinations in the step “grid search,” we derive maps A(z), B(z), and C(z), which are used to estimate the local optimum solution The estimation of local optimum land use maps does not require much computational effort It sorts through all possible combinations in the maps A(z), B(z), and C(z) This easily allows parameter studies for the weighting parameter Due to the minimal computation effort, these steps are embedded in the geographic information system (GIS) Front End ArcView to offer simple result visualization Figure 12.9 shows a screen FIGURE 12.9 Screen shot of ArcView front end with the local optimization project 12 Landscape Optimization 317 copy of the ArcView display with a parameter study of the land use distribution in the Hunting Creek Watershed 12.6.2 Monte Carlo Simulations An analysis of these results can be performed by a Monte Carlo simulation Figure 12.8 shows Monte Carlo simulations in two boxes in the mid-column: • Monte Carlo simulation “from scratch” corresponds to the first box of branch 2; compare the first item in Section 12.5.2 It performs an analysis of the variability of the entire process and sets up an initial population for the genetic programming algorithm • The step after local optimization, which corresponds to the reallocation p1 ϭ or the disturbance of an optimum solution p1 Ͻ 1, item and of Section 12.5.3 Figure 12.10 summarizes the results of a couple of different Monte Carlo runs We used a smaller subwatershed for these analyses This allowed us to generate FIGURE 12.10 Resulting distributions of goal function values (left column) and total harvest biomass (right column) in Monte Carlo simulations for different stochastic processes (rows) 318 Ralf Seppelt and Alexey Voinov more realizations The first column in Figure 12.10 shows histograms of the global goal function results according to Eq (2) The goal function values are normalized by the goal function value derived by the local optimization Values below unity denote simulation runs where the global goal function returns values below the local optimization Values above unity denote that the local solution was improved The second column shows the histogram of the yield in the study area in US $/m2 for comparison Row one of Figure 12.10 can be identified as Monte Carlo simulation from scratch: The distribution of land use types f(c) as well as the land use type and the fertilizer amount for each grid cell are generated at random In the second row, we derived f(c) from the local optimization distribution Rows three and four in Figure 12.10 disturb the locally optimal land use pattern with the probability p1 ϭ 0.1 and p1 ϭ 0.01, respectively Obviously, a stochastic generation of land use patterns hardly identifies a map, which is optimum in terms of the performance criterion From the last two rows of Figure 12.10, one can derive that an increase in the probability p1 immediately leads to solutions which are further away from the local solution The local optimization approach seems to be very close to the global optimum However, there are land use patterns which are “better” in terms of the global optimization Are these land use patterns the ones that take into account neighborhood relationships of the control variables? 12.6.3 Statistical Analysis What are the driving parameters for the optimum solutions? The question can be answered by a simple bivariate correlation analysis based on the resulting land use and fertilizer maps and the spatial input data of the model Table 12.1 summarizes the results of four correlation analysis studies of the optimum solutions with the set of weights ϭ 0.0, 0.1, 0.2, and 1.0 However, the sample size is high (1690) and we used a nonparametric correlation according to Spearman–Rho (Davis, 1984) because a Normal distribution of the resulting parameters cannot be assumed Input data maps are the soil map and the elevation map Parameters of the soil map are porosity (meters of pore space per meters of sediment), infiltration rate (meters per day), field capacity (meters of pore space per meters of sediment), percolation rate (meters per day), and horizontal hydrologic conductivity (per day) From the elevation map, the aspects and slope are derived using GIS functions One general result is that the fertilizer application always correlates with the habitat type: Each crop gets its specific optimum amount of fertilizer Setting ϭ 0.0 neglects any ecological issues of agricultural production and fertilization The optimum solution is a land use map with the most valuable crop and a high fertilizer amount The correlation analysis shows that almost none of the important parameters for nutrient transport in soil are responsible for the land use map The value shows significant correlation with the parameters of the soil map For ϭ 1.0, c(z) and F(z) show significant correlation with all parameters of the soil map Weak correlations are also identified to the parameters of the elevation map, especially to aspect and slope This shows that spatial relationships (to TABLE 12.1 Correlation analysis of local optimization solution to underlying spatial data Porosity Infiltration rate Field capacity 0.046 0.057 Ϫ0.28 0.249 0.20 0.418 Ϫ0.027 0.270 0.045 0.067 Ϫ0.035 0.153 Percolation rate Hydrologic condictivity Elevation Aspect Slope c(z) ϭ 0.0 c(z) Correlation Significance F(z) Correlation Significance Ϫ0.44 0.073 Ϫ0.34 0.161 Ϫ0.20 0.416 0.071* 0.003 0.052 0.031 1.00 0.110** 0.000 0.197** 0.000 Ϫ0.075** 0.002 Ϫ0.044 0.069 Ϫ0.049* 0.044 0.113** 0.000 Ϫ0.001 0.968 0.028 0.250 Ϫ0.031 0.209 Ϫ0.036 0.146 0.193** 0.000 0.182** 0.000 ϭ 0.1 c(z) Correlation Significance F(z) Correlation Significance 0.130** 0.000 Ϫ0.155** 0.000 0.134** 0.000 Ϫ0.178** 0.000 Ϫ0.199** 0.000 0.102** 0.000 Ϫ0.137 0.000 0.087** 0.000 Ϫ0.95** 0.000 Ϫ0.147** 0.000 ϭ 0.2 c(z) Correlation Significance F(z) Correlation Significance 0.252** 0.000 Ϫ0.333** 0.000 0.257** 0.000 Ϫ0.323** 0.000 Ϫ0.577** 0.000 0.089** 0.000 Ϫ0.60** 0.014 Ϫ0.67** 0.006 1.000 0.176** 0.000 Ϫ0.210** 0.000 0.174** 0.000 Ϫ0.206** 0.000 Ϫ0.227** 0.000 0.021 0.387 Ϫ0.003 0.887 Ϫ0.078** 0.001 0.366** 0.000 ϭ 1.0 c(z) Correlation Significance F(z) Correlation Significance 0.408** 0.000 Ϫ0.466** 0.000 0.413** 0.000 Ϫ0.471** 0.000 Ϫ0.382** 0.000 Ϫ0.011 0.665 Ϫ0.046 0.061 Ϫ0.041 0.089 1.000 0.137** 0.000 0.173** 0.000 Ϫ0.086** 0.000 Ϫ0.077** 0.001 0.017 0.487 Ϫ0.010 0.688 0.207** 0.000 0.0159** 0.000 319 * significant correlation according 2-sided significance level 0.05 (Spearman-Rho) ** significant correlation according 2-sided significance level 0.01 (Spearman-Rho) 12 Landscape Optimization 0.162** 0.000 Ϫ0.001 0.976 1.000 320 Ralf Seppelt and Alexey Voinov neighborhood cells) find their interpretation in the locally optimum solution from the spatially explicit model Note that due to the large sample size, a small sample set is sufficient for a significant correlation In general, all correlation values are low However, statistical analysis gave a validation of the derived results 12.6.4 Genetic Algorithms The only way to answer the question about the importance of neighborhood relationships in the control variable maps is to set up an optimization procedure that can use Eq (2) for assessment As we have seen from the results of the Monte Carlo simulation, the application of the GA based on an initial population from scratch will fail because the variability of a population from scratch is much too broad The GA needs too many iterations to converge to the solution we derived by the local optimization approach (compare to Fig 12.8, upper right) We used a smaller subwatershed of the Hunting Creek for a detailed study of this behavior Figure 12.11 displays the convergence process of a GA run from scratch and the GA run from local optimum (smaller graph) Note that it took 300 generations (4500 simulation runs) to achieve 80% of the goal function value For this reason, the local optimum solution is used for the generation of the initial population [see Fig 12.8 (lower part, mid-column)] Configuring parameters of GAs is more an art than a science (Wall, 1996) From the former results, we can generate certain rules for the parameterization: A FIGURE 12.11 Development of global performance criteria values during the Genetic Programming process: (a) The GA process “from scratch”; (b) the GA process started from a local optimum solution with a stochastically “disturbed” of “mutated” population 12 Landscape Optimization 321 too broad variety in the initial population takes us away from the optimum We set p1 ϭ 0.01 for the stochastic generation of the initial population This has to be assured within each step when generating a new population: The mutation probability should be much smaller than p1, and the crossover population should be equal to zero To enable modifications in the population, we set the migration probability to a high value (0.9) The graph in Figure 12.11b shows the results of the GA run started from the local optimum solution The GA process clearly separates from the initial population and improves the optimum solution by 2% What are the changes compared to the local optimum solution? We compiled two maps to show the differences in each grid cell based on the best generation of GA set up by 15 individuals Figure 12.12 shows the results The map in Figure 12.12a shows the average fertilizer difference The map in Figure 12.12b presents the modified land use cells Only a small number of cells are changed: 43 out of 513 cells (8%) The distribution of land use types stayed the same The global optimization by GA performed a reallocation of habitats We expected a couple of solutions, which can improve the local solution A complex problem like this might have multiple solutions This would have led to various realizations (of habitat and fertilization maps), which change different cells compared to the initial solution from the local optimization However, the map in Figure 12.12b FIGURE 12.12 Analysis of GA results The maps show the difference from generation (15 individuals) of the GA to the local optimum maps for the subwatershed: (a) fertilizer amounts; (b) count of how many individuals of the population have a different land use type than the local optimum solution The small map in the center displays the extent and location of the subwatershed within the Hunting Creek watershed 322 Ralf Seppelt and Alexey Voinov shows that nearly all of the individuals in the best generation modified the same grid cells Most of the modified cells belong to the class in which more than 13 out of 16 individuals changed in the cell in the same way An explanation as to why these are the crucial cells is hard to derive by statistical approaches; however, that would allow further improvement of spatial optimization Due to a broad spectrum of input data/maps and due to a very complex network of dynamic and spatial process in the simulation model, no significant correlation can be identified Using principle-components analysis, we were not able to reduce the state space for analysis One possible approach is to perform a bivariate correlation of c(z0) and F(z0), to the neighborhood cells [northeast, north, , southwest c(z0)] Two issues make us expect illusory correlations First, analysis of the entire region, using every grid cell (and its neighbors) as a repetition, gives only a general answer, which is mainly driven by the global aspect of the global slope of the catchment Second, the spatial hydrologic algorithms assumed use linkages between cells, which are not direct neighbors (Voinov et al., 1999b) 12.7 Discussion We have described earlier as the crucial parameter to assess nutrient outflow both from the watershed as a whole and for an individual cell Therefore, we started with some sensitivity analysis to this parameter Several optimum land use and fertilizer patterns were calculated depending on different values of Figure 12.13 shows the results in an aggregated way With the value on the x axis, Figure 12.13A displays the number of cells allocated for different crops, fallow, and FIGURE 12.13 Dynamics of land use allocations and fertilizer applications in the optimal solution as a function of the environmental weighing factor 12 Landscape Optimization 323 forest Figure 12.13B shows the number of cells with agricultural use, and Figure 12.13C integrates the total amount of fertilizer applied to a grid cell If there are no concerns about the environmental conditions, ϭ 0, the optimal strategy for the region is to use it entirely to grow soybeans and corn Comparatively high market prices for soy and relatively low fertilizer demands make it the most valuable crop Increasing results in a decrease in the number of agricultural cells in the optimal solution The forest appears and then further expands as grows The total fertilizer amounts are rapidly reduced due to a more “punishing” nutrient outflow that comes from the increasing value Gradually, corn and wheat displace soybeans This is something yet to be explored because it is not directly obvious why the more profitable and less-fertilizer-demanding crop gets phased out by the less profitable and more polluting ones One possible explanation is in the timing of the growth seasons for each of the crops Currently, soybeans are assumed to be planted in July and their growth season and fertilizer application days are at a time when there is no active growth of trees that can potentially intercept the nutrient runoff This can potentially give a certain advantage to corn, with a growth season coinciding with that of forests The resulting spatial patterns are presented in Figure 12.14 Optimum land use patterns now show that as grows, forested cells appear and then expand primar- FIGURE 12.14 Spatial representation of the optimization results Note that as the importance of the environmental factors increases ( positive), the forests appear in the riparian zones and then start to further expand 324 Ralf Seppelt and Alexey Voinov ily around creeks and wetlands A further increase in results in a more heterogeneous land cover Soil moisture and nutrient content are not the only driving force for land use of forest or fallow The optimum solution is a result of all the complex interactions between the different processes in the model That is why it may sometimes be quite hard to interpret and explain the resulting allocations The presented results are certainly quite limited for two main reasons First, the goal function considered includes only the benefits of agriculture when estimating the value of a land use pattern The benefits of other land uses, such as forest or residential, are neglected The allocation of residential cells was entirely fixed, whereas the forest value appeared only in the nutrient-absorption capacity that reduced the amounts of nutrients delivered to the estuary In reality, there are certainly other values associated with forests, both market and nonmarket ones Second, there is little chance that the optimal patterns generated can be implemented on already developed landscapes Rarely will we be able to restructure a whole landscape with reforestation and complete reallocation of certain agricultural sites to achieve these optimal patterns In addition, we did not consider all of the operational costs associated with agricultural production, including prices of seeds, labor, energy, irrigation, and so forth This should not be too hard to do, but yet it has to be done There are definitely many other uncertainties that should be also studied and evaluated For example, how climatic variations affect the optimal land allocations? How much does the timing of various crops matter? Nevertheless, the results have a certain qualitative, conceptual value It was quite remarkable to see how the pattern of forests emerged as soon as we have introduced the environmental factors into the goal function ( Ͼ 0) This pattern was in perfect agreement with the widely discussed concept of riparian buffers and their role in nutrient entrapment As soon as the value of a cleaner environment becomes a concern, the riparian zones tend to become forested The size of the buffer then gradually grows as these concerns become more pronounced ( increases) Most important was to test the feasibility of the spatial optimizations in the context of fairly complex process-based simulation models Indeed, a spatial simulation model by itself is complex and its analysis is quite restricted It really becomes crucial to design some simplified techniques if we are to attempt optimization at all By generating the series of local response maps and then optimizing over the set of these maps, we have significantly decreased the amount of computations needed It is most likely that the results of this localized optimization are different from the global solution to the problem However, the tests that we have performed using the Monte Carlo analysis indicated that we should be somewhere in the vicinity of the solution Further application of genetic algorithms for global optimization (Seppelt and Voinov, 2002) only proved that the local optimization method indeed gives a good estimate for the global optimum Certainly, we should apply this conclusion with caution, because for other areas or other models, this result may be quite different For the Hunting Creek watershed, it was probably important that the area did not have significant elevation gradients, which restricted the amount of lateral interactions and increased the importance of local vertical processes We may expect that the horizontal fluxes and interac- 12 Landscape Optimization 325 tions between cells could play a more important role for steeper slopes, making local optimizations less representative of the overall spatial dynamics Acknowledgments This research has been supported by a grant from the U.S Environmental Protection Agency’s Science to Achieve Results (STAR) program (R827169) and the German Science Foundation (Deutsche Forschungsgemeinschaft, DFG) project SE-796/1-1 The Hunting Creek modeling project was also supported by a grant from the Calvert County Board of Commissioners We are grateful to Dave Brownlee from Calvert County Planning and Zoning for much needed advice and help with some data We thank Thomas Maxwell for several upgrades in the SME that were required for linkages to the optimization tools Additional thanks are due to Dagmar Söndgerath for her valuable assistance concerning the statistical analysis Results in this contribution were obtained using the GAlib genetic algorithm package, written by Matthew Wall at the Massachusetts Institute of Technology References Band, L.E., D.L Peterson, S.W Running, J Coughlan, R Lammers, J Dungan, and R Nemani 1991 Forest ecosystem processes at the watershed scale: basis for distributed simulation Ecological Modelling 56:171–196 Beven, K.J and M.J Kirkby 1979 A physically-based, variable contributing area model of basin hydrology Hydrological Sciences Bulletin 24(1):43–69 Bevers, M., J Hof, D.W Uresk, and G.L Schenbeck 1997 Spatial optimization of prairie dog colonies for black-footed ferret recovery Operations Research, 45:495–507 Bulirsch, R., A Miele, J Stoer, and K.H Well 1993 Optimal control—Calculus of variations, optimal control theory and numerical methods International Series of Numerical Mathematics, 111, Birkhäuser; Basel Burke, I.C., D.S Schimel, C.M Yonker, W.J Parton, L.A Joyce, and W.K Lauenroth 1990 Regional modeling of grassland biogeochemistry using GIS Landscape Ecology 4(1):45–54 CBP 2001 Chesapeake Bay Program Available from http://www.chesapeakebay.net R Costanza, A Voinov, R Boumans, T Maxwell, F Villa, L Wainger, and H Voinov 2001 Case study: Patuxent River watershed, Maryland In R Costanza, B Low, E Ostrom, and J Wilson, Eds Institutions, Ecosystems, and Sustainability Lewis: Boca Raton p 179–230 Costanza, R., F.H Sklar, and M.L White 1990 Modeling coastal landscape dynamics BioScience 40(2):91–107 Davis, J.C 1984 Statistical Data Analysis in Geology John Wiley & Sons: New York Engel, B.A., R Srinivasan, and C Rewerts 1993 A Spatial decision support systemor modeling and managing agricultural non-point-source pollution, in Environmental Modeling with GIS (Goodchild M.F., B.O.P., and L.T Steyaert, eds.) Oxford University Press, Oxford Fitz, H.C and F.H Sklar 1999 Ecosystem analysis of phosphorus impacts and altered hydrology in the Everglades: A landscape modeling approach 585-620 In K.R Reddy, G.A O’Connor, and C.L Schelske, Eds Phosphorus Biogeochemistry in Subtropical Ecosystems Lewis Publishers: Boca Raton, FL., pp 585–620 326 Ralf Seppelt and Alexey Voinov Jaklitsch, F (1997) Comprehensive Plan Calvert County, Maryland Planning Commission Available from http://www.co.cal.md.us/planning/compplan/compmain Krysanova, V., A Meiner, J Roosaare, and A Vasilyev 1989 Simulation modelling of the coastal waters pollution from agricultural watershed Ecological Modelling 49:7–29 Loehle, C 2000 Optimal control of spatially distributed process models Ecological Modelling, 131:79–95 Makowski, D., E.M.T Hendrix, M.K van Ittersum, and W.A.H Rossing 2000 A framework to study nearly optimal solutions of linear programming models developed for agricultural land use exploration In Ecological Modelling, 131:65–77 Martínez-Falero, E., I Trueba, A Cazorla, and J.L Alier 1998 Optimization of spatial allocation of agricultural activities Journal of Agricultural Engineering Ressources, 69:1–13 Maxwell, T and R Costanza 1995 Distributed modular spatial ecosystem modelling 5(3):247–262; special issue on advanced simulation methodologies NADP 2000 NRSP-3/National Trends Network NADP Program Office, Illinois State Water Survey, Champaign, IL Available from Available from www.nadp.sws.uiuc.edu /nadpdata/monthlyRequest.asp?siteϭMD03 Nevo, A., R Oad, and T.H Podmore 1993 An integrated expert system for optimal crop planning Agricultural Systems, 45:73–92 Randhir, T.O., J.G Lee, and B Engel 2000 Multiple criteria dynamic spatial optimization to manage water quality on a watershed scale, Transaction of American Society of Agricultural Engineers, 43(2):291–299 Sasowsky, C.K and T.W Gardner 1991 Watershed configuration and geographic information system parameterization for SPUR model hydrologic simulations Water Resources Bulletin 27(1):7–18 Seppelt, R 2000 Modelling regionalised optimum control problems for agroecosystem management Ecological Modelling 131(2–3), 121–132 Seppelt, R and A Voinov 2002 Optimization methodology for land use patterns using spatially explicit landscape models Ecological Modelling 151:125–142 Sklar, F.H., R Costanza, and J.W Day, Jr 1985 Dynamic spatial simulation modeling of coastal wetland habitat succession Ecological Modelling 29:261–281 USGS 1995 Maryland NWIS-W Data Retrieval Available from www.waterdata.usgs gov/nwis-w/MD USGS 1997 Maryland Surface-Water Data Retrieval Available from www.h2o.usgs gov/swr/MD Voinov, A., H.V Voinov, R Boumans, and R Costanza 1999 Environmental modeling of Calvert county buildout scenarios using hunting creek as a case study Report to Calvert County Board of Commissioners, Ref No [UMCES] CBL 98–166 Wall, M 1996 GALib: A Cϩϩ Library of Generic Algorithm Components, Version 2.4 MIT: Mechanical Engineering Dept URL: http://lancet.mit.edu/ga/ Index ADAM, 184, 188, 190 aggregation, 7, 56, 108–109, 112, 123, 151, 198, 202, 225, 297–298, 303, 312 agriculture, 143, 303, 324 land use change from/to, 205, 216 ANSWERS, 204 ArcView, 316 atmospheric deposition, 56, 166, 306 BIOME-BGC, 200 biological time, 60, 64 biomass, 51, 89, 135, 153, 317 eelgrass, 178, 185, 189–190 forest, 207 plant, 60, 64–66, 71, 73, 124, 147, 154–155 BTELSS (Barataria Terrebonne Ecological Landscape Spatial Simulation), 121 buildout, 217 budgets hydrologic, 158 of nitrogen and phosphorus, 148, 223–224 calibration, 6, criteria, 80 hydrologic, 102, 157–161, 168, 208–211, 305–306 local, 68, 96 multitier, 67–70, 103, 206 for nutrients, 150, 158, 161–163, 213, 308 problems, 78 of spatial models, 83, 96–102, 123–125, 127, 133, 137–138, 151–156, 168, 185, 212 of unit model, 90, 96, 180–181, 206–208 CENTURY, 200 clustering, 219, 226 composability, 44 complexity, 16 and calibration, 93, 96 models as complex hypotheses, 10 and optimization, 311, 312 crop, 207 rotation, 66 DDD (density dependent death), 286–287 decomposition, 57, 154, 178, 189 DEM (digital elevation model), 252 deposited organic material, 66 detritus, 66, 178, 187, 190 dispersal, 252, 255–257, 265–266, 269, 274, 282 DLS (dynamic landscape simulation), 249, 269 economic model, 205 eelgrass, 173 ELM (Everglades Landscape Model), 105 EPIC (Erosion Productivity Impact Calculator), 207 erosion, 32, 120, 191 error model, 84–85 327 328 Index eutrophication, 152, 173 evaporation, 51 evapotranspiration, 52, 253 fertilizer, 56, 71, 191, 218, 220, 303, 306, 310, 318 FIA (Forest Inventory and Analysis database), 70, 207, 213 field capacity, 52 fit analysis, 123 fractal, forest, 71, 207, 213, 312, 323–324 GEM (General Ecological Model), 6, 43, 49, 73, 96, 151, 202 genetic algorithm, 82, 92, 95, 314, 320 GIS (Geographic Information System), 26–27, 30, 148, 200, 202, 209, 235, 252, 269, 275, 279, 283–284, 289, 316, 318 global optimization, 315 goal function (see also performance criterion, objective function), 309, 310, 324 GRASS (Geographic Resources Analysis Support System), 202, 235, 237, 252, 283 grid linking cells, 38 in SME, 30 vs lumped approach, 17, 303 habitat, 24, 48, 51, 53, 67, 70–71, 119, 124–125, 136, 146, 148, 200, 240, 243, 289, 310 distribution, 123, 127–128, 133–134, 138, 142, 151, 165, 175 quality, 238 switching (change), 122, 132, 134, 137, 143, 167, 173, 188, 199, 218 harvest, 65 hibernation, 256–257 hierarchy, 8, 177, 184 HSPF (Hydrological Simulation Program—FORTRAN), 209 hunting, 294 Hunting Creek, 70, 227, 302 hydrology, 49, 204, 207 spatial, 54, 157 impact disease, 280, 292–293 ecological, 121, 169 military training, 242, 249–251, 254–257, 261–262, 265–270 nutrient enrichment, 144, 165, 168–169 parasitism, 233, 243 infiltration, 50, 53, 253, 318 landscape, 1, 266 landscape model, 3, 269 land use, 215, 318 change, 205–206 LCTA (Land Condition Trend Analysis), 253 leaf area index, (LAI) 51–52 growth, 177 length, 181–182, 188–189, 191–192 light, 62, 182 litterfall, 65, 187–188 MADONNA, 48 management, 121, 143, 191, 217, 219, 244, 270, 296 of hydrology, 131, 135, 144 practice, best, 73, 191 MATLAB, 92 Michaelis–Menten, 63, 183 modularity, 44, 74, 203 in SME, 25 Monte-Carlo, 82, 90, 313, 317 MPI (Model Performance Index), 84, 86 tests included, 86 NDVI (Normalized-Difference Vegetation Index), 202, 212–213 neural network, 252, 254, 258–260, 267 NOAA (National Oceanic and Atmospheric Administration), 195, 253 NPP (Net Primary Production), 63, 178, 207, 220 nutrients, 55, 57, 190 dissolved, 183 enrichment, 144, 187 loading, 168, 173 management, 218 vertical movement, 59 Index objective function, 81, 85 (see also goal function) optimization global, 315 local, 316 methods, 82, 311 parasitism, 238, 239 patch, 266 Penman–Monteith, 52 percolation, 53 performance criterion (also see objective function), 309 plant growth, 57, 71, 164, 178 PLM, 67, 96, 103, 197 population density, 249, 257–266 dynamics, 255, 282, 290 porosity, 53 proximity, 205 PVA, 233 rainfall, 50, 306 RAMAS, 241 reproduction process, 63 riparian buffer, 324 root, 50, 56, 59 eelgrass, 189 zone, 51–52, 58, 63 runoff, 32, 253 satellite imagery, 202, 252, 254, 259 scale, and hierarchy, mulitscale dynamics, spatial (see spatial resolution) temporal (see time scale) scenario, 126, 134, 150, 161, 185, 190, 213, 258, 302, 309 vs optimization, 312 in SME, 23, 37 sea-level rise (SLR), 119, 123, 126–127, 130, 134–135, 138 sediment, 33, 120 septic, 306 sewage, 56, 219, 223, 306 SFWMM (South Florids Water Management Model), 144–145, 148–150, 167 comparison with ELM, 157–161 329 shellfish, 175 SME (Spatial Modeling Environment), 45, 47, 148, 277, 298 build, 29 classes, 26 components, 22 configuration, 23, 27, 37 event, 26 framelink, 38 import, 25, 36 portal, 31 run, 29, 38 user code, 36, 48 viewserver, 38 solar radiation, 49, 62 soil, 70, 122, 147, 169 compaction, 254, 257 loss/gain, 147, 155 moisture, 50, 168, 252–255 nutrients, 73, 150–152, 154, 161, 165–168 type, 32–33, 37, 48, 123, 167, 202, 212, 236, 240–241, 254, 306, 310, 318 spatial resolution, 30, 122, 146, 184, 202, 208, 236, 251, 289 sprawl, 226 STA (Stormwater Treatment Areas), 144–145, 149–150, 161, 168–169 stakeholders, 228 STELLA, 24, 34, 40, 45, 47, 73, 89, 178, 251, 278, 285 succession, 154, 243, 254 suitability, 236, 261, 289 TEM, 200 temperature, 60, 234, 250, 252–253, 256 limitation of decomposition, 67, 255 of water, 182 time biological, 60, 64–65 of crop rotation, 66–67 incubation, 281 in hydrology, 4, 14, 50–51, 54, 70, 148, 184, 190–191, 205, 209 in SME, 26 persistence, 266 retention, 133 scale, 4, 8, 16, 119, 167–168, 206, 252 330 Index time (cont.) series, 32, 45, 47, 56, 80, 87, 96, 102–103, 108, 189, 202, 207–208, 212, 306 step, 5–7, 14–16, 26, 122, 185, 200–201, 236, 251, 256–259, 276, 284, 298 translocation, 64, 178 transpiration, 52 uncertainty, 79, 144, 167, 169 unit model, 178 USGS (United States Geological Survey), 70, 209–210, 306, 308 validation, 125 value, 324 vegetation growth, 254–255 transition (see succession) verification, 77, 184 wrack, 178, 189 ... resolution and predictability in landscape modeling to allow a more “optimal” resolution to be chosen for specific modeling problems Our experience with spatial modeling indicates that complexity... Congress Cataloging-in-Publication Data Landscape simulation modeling: a spatially explicit, dynamic approach / editors, Robert Costanza, Alexey Voinov p cm.— (Modeling dynamic systems) Includes bibliographical... affected by the landscape and by how it evolves in space and in time This book puts you at the cutting edge of the science of spatial dynamic simulation modeling The Spatial Modeling Environment