Thomas Maxwell

Progress in integrated assessment and modelling

Environmental processes have been modelled for decades. However. the need for integrated assessment and modeling (IAM) has,town as the extent and severity of environmental problems in the 21st Century worsens. The scale of IAM is not restricted to the global level as in climate change models, but includes local and regional models of environmental problems. This paper discusses various definitions of IAM and identifies five different types of integration that Lire needed for the effective solution of environmental problems. The future is then depicted in the form of two brief scenarios: one optimistic and one pessimistic. The current state of IAM is then briefly reviewed. The issues of complexity and validation in IAM are recognised as more complex than in traditional disciplinary approaches. Communication is identified as a central issue both internally among team members and externally with decision-makers. stakeholders and other scientists. Finally it is concluded that the process of integrated assessment and modelling is considered as important as the product for any particular project. By learning to work together and recognise the contribution of all team members and participants, it is believed that we will have a strong scientific and social basis to address the environmental problems of the 21st Century.

Patuxent landscape model: integrated ecological economic modeling of a watershed

The Patuxent Landscape Model (PLM) is designed to simulate fundamental ecological processes on the watershed scale, in interaction with an economic component that predicts the land use patterns. The paper focuses on the ecological component of the PLM and describes how the spatial and structural rescaling can be instrumental for calibration of complex spatially distributed models. The PLM is based on a modified General Ecosystem Model (GEM) that is replicated across a grid of cells that compose the rasterized landscape. Different habitats and land use types translate into different parameter sets to be fed into GEM. Cells are linked by horizontal fluxes of material and information, driven mostly by the hydrologic flows. This approach provides additional flexibility in scaling up and down over a range of spatial resolutions and is essential to track the land use change patterns generated by the economic component. Structural modularity is another important feature that is implemented in the general purpose software packages (Spatial Modeling Environment and Collaborative Modeling Environment), that the PLM employs.

Development of a general ecosystem model for a range of scales and ecosystems

Abstract We have developed a General Ecosystem Model (GEM) that is designed to simulate a variety of ecosystem types using a fixed model structure. Driven largely by hydrologic algorithms for upland, wetland and shallow-water habitats, the model captures the response of macrophyte and algal communities to simulated levels of nutrients, water, and environmental inputs. It explicitly incorporates ecological processes that determine water levels, plant production, nutrient cycling associated with organic matter decomposition, consumer dynamics, and fire. While the model may be used to simulate ecosystem dynamics for a single homogenous habitat, our primary objective is to replicate it as a “unit” model in heterogeneous, grid-based dynamic spatial models using different parameter sets for each habitat. Thus, we constrained the process (i.e., computational) complexity, yet targeted a level of disaggregation that would effectively capture the feedbacks among important ecosystem processes. A basic version was used to simulate the response of sedge and hardwood communities to varying hydrologic regimes and associated water quality. Sensitivity analyses provided examples of the model dynamics, showing the varying response of macrophyte production to different nutrient requirements, with subsequent changes in the sediment water nutrient concentrations and total water head. Changes in the macrophyte canopy structure resulted in differences in transpiration, and thus the total water levels and macrophyte production. The GEMs modular design facilitates understanding the model structure and objectives, inviting variants of the basic version for other research goals. Importantly, we hope that the generic nature of the model will help alleviate the “reinventing-the-wheel” syndrome of model development, and we are implementing it in a variety of systems to help understand their basic dynamics.

Integrated ecological economic modeling of the patuxent river watershed, Maryland

Understanding the way regional landscapes operate, evolve, and change is a key area of research for ecosystem science. It is also essential to support the “place-based” management approach being advocated by the U.S. Environmental Protection Agency and other management agencies. We developed a spatially explicit, process-based model of the 2352 km2 Patuxent River watershed in Maryland to integrate data and knowledge over several spatial, temporal, and complexity scales, and to serve as an aid to regional management. In particular, the model addresses the effects of both the magnitude and spatial patterns of human settlements and agricultural practices on hydrology, plant productivity, and nutrient cycling in the landscape. The spatial resolution is variable, with a maximum of 200 × 200 m to allow adequate depiction of the pattern of ecosystems and human settlement on the landscape. The temporal resolution is different for various components of the model, ranging from hourly time steps in the hydrologic sector to yearly time steps in the economic land-use transition module. We used a modular, multiscale approach to calibrate and test the model. Model results show good agreement with data for several components of the model at several scales. A range of scenarios with the calibrated model shows the implications of past and alternative future land-use patterns and policies. We analyzed 18 scenarios including: (1) historical land-use in 1650, 1850, 1950, 1972, 1990, and 1997; (2) a “buildout” scenario based on fully developing all the land currently zoned for development; (3) four future development patterns based on an empirical economic land-use conversion model; (4) agricultural “best management practices” that lower fertilizer application; (5) four “replacement” scenarios of land-use change to analyze the relative contributions of agriculture and urban land uses; and (6) two “clustering” scenarios with significantly more and less clustered residential development than the current pattern. Results indicate the complex nature of the landscape response and the need for spatially explicit modeling.

Resolution and predictability: An approach to the scaling problem

We analyzed the relationship between resolution and predictability and found that while increasing resolution provides more descriptive information about the patterns in data, it also increases the difficulty of accurately modeling those patterns. There are limits to the predictability of natural phenomenon at particular resolutions, and “fractal-like” rules determine how both “data” and “model” predictability change with resolution.We analyzed land use data by resampling map data sets at several different spatial resolutions and measuring predictability at each. Spatial auto-predictability (Pa) is the reduction in uncertainty about the state of a pixel in a scene given knowledge of the state of adjacent pixels in that scene, and spatial cross-predictability (Pc) is the reduction in uncertainty about the state of a pixel in a scene given knowledge of the state of corresponding pixels in other scenes. Pa is a measure of the internal pattern in the data while Pc is a measure of the ability of some other “model” to represent that pattern. We found a strong linear relationship between the log of Pa and the log of resolution (measured as the number of pixels per square kilometer). This fractal-like characteristic of “self-similarity” with decreasing resolution implies that predictability may be best described using a unitless dimension that summarizes how it changes with resolution. While Pa generally increases with increasing resolution (because more information is being included), Pc generally falls or remains stable (because it is easier to model aggregate results than fine grain ones). Thus one can define an “optimal” resolution for a particular modeling problem that balances the benefit in terms of increasing data predictability (Pa) as one increases resolution, with the cost of decreasing model predictability (Pc).

Spatial ecosystem modelling using parallel processors

Abstract We have developed a spatial modeling workstation, that consists of a combination of hardware and software tools that allow development, implementation and testing of spatial ecosystem models in a convenient desktop environment. The system links commercially available Geographic Information Systems (GIS) for managing spatial data with a commercially available general dynamic simulation system for developing unit models (STELLATM) and a Spatial Modeling Package (SMP) that we developed for linking the unit models into a spatial array, handling horizontal exchanges, and running the array as a spatial model. The spatial model code is executed on either: (1) transputers (parallel processors) resident in a desktop microcomputer; or (2) on a remote Connection Machine parallel mainframe computer with 64 K processors. Resulting time series maps are readable by the GIS system for further display and analysis. In this paper we: (1) describe the hardware and software system; (2) describe a hypothetical model of simple diffusion over a landscape that we developed and tested using the system; and (3) describe a practical application of the system to spatial modeling of long-term habitat succession in the coastal Louisiana region. We find that a system using eight transputers on a Macintosh IIci can run spatial ecosystem models in about the same time asa CRAY X/MP.

A paris-model approach to modular simulation

Abstract The development of complex models can be greatly facilitated by the utilization of libraries of reusable model components. In this paper we describe an object-oriented module specification formalism (MSF) for implementing archivable modules in support of continuous spatial modeling. This declarative formalism provides the high level of abstraction necessary for maximum generality, provides enough detail to allow a dynamic simulation to be generated automatically, and avoids the “hard-coded” implementation of space-time dynamics that makes procedural specifications of limited usefulness for specifying archivable modules. A set of these MSF modules can be hierarchically linked to create a parsimonious model specification, or “parsi-model”. The parsi-model exists within the context of a modeling environment (an integrated set of software tools which provide the computer services necessary for simulation development and execution), which can offer simulation services that are not possible in a loosely-coupled “federated” environment, such as graphical module development and configuration, automatic differentiation of model equations, run-time visualization of the data and dynamics of any variable in the simulation, transparent distributed computing within each module, and fully configurable space-time representations. We believe this approach has great potential for bringing the power of modular model development into the collaborative simulation arena.

Development and application of the Everglades landscape model

Water management infrastructure and operations have fragmented the greater Everglades into separate, impounded basins, altering flows and hydropatterns in these internationally recognized wetlands. A significant area of this managed system has experienced anthropogenic eutrophication. This combination of altered hydrology and water quality has interacted to degrade vegetative habitats and other ecological characteristics of the Everglades. As part of a massive plan to “restore” the Everglades, simulation models are being applied to better understand the system’s hydrologic and ecological dynamics, evaluating options for restoration plans. One such tool is the Everglades Landscape Model (ELM), a process-based, spatially explicit simulation of ecosystem dynamics across a heterogeneous, 10,000 km region. The model development has proceeded in tandem with advances in Everglades research, improving its algorithms and calibration to best capture dynamics of key landscape attributes. The first spatial application of the model was in an intensively studied subregion along an anthropogenic nutrient gradient. The model captured the spatiotemporal dynamics of hydrology, surface and ground water phosphorus, periphyton biomass and community type, macrophyte biomass and habitat type, and peat accumulation. Refinements to the model have improved its hydrologic and ecological performance, with good calibrations of long term hydrologic and surface water quality dynamics across most of the Everglades landscape. Using this updated version, we evaluated phosphorus loading throughout the Everglades system under two base scenarios. The 1995 base case assumed current management operations, with phosphorus inflow concentrations fixed at their long term, historical average. The 2050 base case assumed future modifications in water management, with all managed inflows to the Everglades having reduced phosphorus concentrations (due to filtering by constructed wetlands). In an example “indicator” subregion that currently is highly eutrophic, the 31-yr simulations predicted that desirable periphyton and macrophyte communities were maintained under the 2050 base case, whereas in the 1995 base case, periphyton biomass and production decreased to negligible levels and macrophytes became extremely dense. The negative periphyton response in the 1995 base case was due to high phosphorus loads and rapid macrophyte growth that shaded this algal community. Along an existing 11 km eutrophication gradient, the model indicated that the 2050 base case had ecologically significant reductions in phosphorus accumulation compared to the 1995 base case. Indicator regions (in Everglades National Park) distant from phosphorus inflow points also exhibited reductions in phosphorus accumulation under the 2050 base case, albeit to a lesser extent due to its distance from phosphorus inflows. The ELM fills a critical information need in Everglades management, and has become an accepted tool in evaluating scenarios of potential restoration of the natural system. Refinements to the model will enable us to evaluate the full suite of ecological responses to management scenarios throughout the greater Everglades.

Non - spatial calibrations of a general unit model for ecosystem simulations

General Unit Models simulate system interactions aggregated within one spatial unit of resolution. For unit models to be applicable to spatial computer simulations, they must be formulated generally enough to simulate all habitat elements within the landscape. We present the development and testing of a unit model for the Patuxent River landscape in the state of Maryland, USA. The Patuxent Landscape Model (PLM) is designed to simulate the interactions among physical and biological dynamics in the context of regional socioeconomic behavior. The PLM is a tool for evaluating landscape change within the Patuxent watershed through simulation of ecological systems. A companion economic model estimates land development patterns and effects on human decisions from site characteristics, ecosystem properties, and regulatory paradigms. Landscape elements that are linked within the PLM are forest, agriculture and open water systems, and three levels of urban development. Urban developments are low and medium density residential areas (14.07% of the total watershed), and commercial, industrial and institutional area (5.7%). Forests are mixed populations of deciduous and evergreen species (45.11%). Agricultural areas (28.02%) are simulated through rotating crops of corn, winter wheat and soybeans within a cycle of two years. Open water (6.84%) represents the ecosystems within the rivers and streams where phytoplankton are the primary producers. In this paper we illustrate, how we gathered and formalized working models used within the Patuxent watershed for forests, agriculture urban settings and wetlands. Further, we show how we tested and merged the variety of models employed by scientific disciplines and created a general unit model to be used in the Patuxent Landscape Model (Pat – GEM). The Patuxent Landscape Model is built under the Spatial Modeling Environment.

An approach to modelling the dynamics of evolutionary self-organization

We have simulated the dynamics of species evolution in a systems context on a parallel supercomputer. Population dynamics are represented as generalized Lotka-Volterra systems defined as points in a generalized phenotype or character space T. Populations which are closest in T compete most strongly for resources. A variety of systems with varying assumptions, resource distributions, and number of trophic levels were simulated. Starting with a random initial seed proceeding through a complex temporal sequence, most cases converged to essentially the same configuration. The final equilibrium state consisted of a gridwork of localized population clusters in T, representing individual species. The intercluster spacing was roughly equal to the standard deviation of the resource utilization function. Thus the systems self-organize to an array of niches which maximally fills the available volume of resource space while minimizing the overlap of resource utilization functions. The simulations were performed on a Connection Machine (a massively parallel supercomputer) which allowed up to 32 000 distinct points in character space to be modelled in parallel. Simulation allows a more realistic treatment of evolutionary dynamics and greater flexibility in experimental manipulation than previous analytical approaches. We experimented with temporal variations in the resource base. In most cases the niche structure was not affected; species prospered or declined as a function of local resource availability but the niche pattern remained invariant. However, in the case in which each species depends on only one or two resources, increasing randomness in the resource base resulted in a decrease in the number of species.