The objective of this research was to create and operationalize a framework for analyzing alternative wetlands conservation strategies. The framework consists of a set of functions describing the expected environmental benefits and economic costs of wetlands restoration activities integrated into a spatial optimization model that can determine the configuration of restoration that maximizes some (weighted combination of) environmental benefits subject to a budget constraint. The model was developed and applied specifically to wetlands restoration in the Central Valley of California, but the methods, and to some degree the models themselves, could be transferred to different areas and policy contexts.
The research was designed to: (1) investigate the feasibility and utility of applying numerical optimization techniques to the problem of prioritizing sites for wetlands restoration; (2) investigate the importance of spatial effects on the provision of ecosystem services from wetlands; and (3) assess tradeoffs between different categories of environmental benefits from wetlands.
Optimizing Conservation Activities
Decisionmakers could address the problem of prioritizing sites for conservation activities (for preservation, enhancement, or restoration) by characterizing the expected benefits and costs of each conservation option, and then prioritizing the options in decreasing order of their benefit-cost ratio. However, in the case of wetlands conservation (and likely many other environmental land use policies), the benefits will be a function of the level of ecosystem services expected from the managed systems, which are partly a function of where wetlands are located; with respect to other wetlands and other land use types. Thus, the benefits of restoring a particular wetland can depend on whether or not other wetlands are restored nearby. Taking account of the spatial relationships that affect the provision of ecosystem services from wetlands requires more sophisticated optimization techniques, and for many real-world problems it may be impossible to guarantee a globally optimal solution when spatial effects are important.
To address the potential importance of spatial effects for optimal wetlands conservation, a numerical optimization model was developed based on a stylized description of a landscape with urban, agriculture, and wetland land use types. The hypothetical landscape consisted of a grid of 625 square cells (25 x 25), and was generated randomly, though subject to constraints that resulted in clumpy urban areas in a matrix of agriculture, with some remnant wetlands scattered throughout. Using two standard models of ecosystem processes from the literature-one describing nonpoint source runoff, and one describing habitat quality for a species that exhibits random radial dispersal-the results of three site selection algorithms were compared. The objective was to maximize the provision of ecosystem services (nonpoint source pollution reduction, provision of habitat, or some weighted combination of the two), subject to a budget constraint. The first was a simple heuristic based on maximizing the area of restored wetlands; the second was a greedy algorithm, which chose sites in an iterative fashion, selecting the best remaining site at each iteration; and the third was an optimizing algorithm, which checked all feasible combinations of sites. The results demonstrated that when spatial effects are strong enough, the simple heuristic based on maximizing wetland area will perform significantly worse than methods that account explicitly for the spatial relationships that affect the provision of ecosystem services. The results also demonstrated that while iterative heuristics may perform worse than optimizing algorithms, they may nevertheless perform nearly as well. This has important implications because many real-world problems will be too large to apply optimizing algorithms when spatial effects are important, which means that heuristics will have to be employed. The results of this and other research in the field suggest that appropriate heuristics are sufficient for the task.
Spatial Effects and Wetland Ecosystem Services
This research focused on two wetland ecosystem services: the provision of habitat for birds (with a focus on mallards), and the attenuation of nutrients in nonpoint source runoff from agriculture. Standard count regression techniques were used to relate the abundance of mallards in the Central Valley (as recorded in the North American Breeding Bird Survey) to a suite of landscape variables, including the percent of nine different land use types within 400 meters of each survey location. The regression models explained between 30 and 60 percent of the variation in mallard abundances, and the results indicated that mallards prefer a mix of wet and dry land use types in the breeding season. This means that wetlands restoration efforts could be targeted spatially to take advantage of these habitat preferences. An optimization analysis based on the regression results suggested that the potential gains from a spatially targeted approach, over a non-targeted approach, could be substantial.
A spatially distributed hydrologic simulation model was developed and applied to the Central Valley to estimate the amount of nutrients in nonpoint source runoff that could be attenuated in wetlands before reaching downstream water bodies. The model is based on a set of water and mass balance equations, which are applied to each of 1.4 million square cells (200 meters on each side) that make up the study area. Each cell is characterized by its land use type and soil type, which determine the amount of irrigation water and nutrients applied (for agriculture cells) or stormwater runoff and nutrient concentrations therein (for urban cells). Runoff from each cell flows in the direction of the shortest distance to the nearest surface water body, and along the way runoff may be intercepted by wetland cells, in which case nutrients are attenuated according to a first-order removal rate equation (using standard parameter values from the literature). The model can estimate the portion of nitrogen and phosphorus applied to agriculture that is taken up by crops, infiltrates to the groundwater, is attenuated in wetlands, or flows in runoff to surface waters. The baseline results indicated that of approximately 350 million kg/yr of nitrogen inputs to the Central Valley, about 90 percent is taken up by crops for growth, 8.5 percent leaches to the groundwater, and 1.5 percent enters rivers by way of nonpoint source pollution in surface runoff. The amount of nitrogen attenuated in existing wetlands is approximately 9.1 percent of the total load to rivers and streams from surface runoff. The model also can calculate the expected nutrient attenuation rates in restored wetlands, and was used in this capacity, along with the mallard model described above, as part of a numerical optimization model to investigate tradeoffs between habitat and water quality benefits from wetlands restoration activities in the region.
Tradeoffs Between Different Environmental Objectives
The mallard habitat model and the hydrologic simulation model, along with estimates of land values and wetlands restoration costs in the region, were integrated into a numerical spatial optimization model. The integrated optimization model was designed to investigate the potential tradeoffs between habitat and water quality benefits from wetlands restoration activities in the Central Valley. First, the model was used to trace out a "production possibilities frontier" (PPF) for the two ecosystem services in four small watersheds in the Central Valley. The PPF is a curve (or collection of points) that indicates the maximum possible levels of ecosystem services (either habitat quality or water quality or some weighted combination of the two) attainable given the restoration budget. Each point on the PPF is defined by a different level of habitat improvement (expected increase in breeding mallard population size) and water quality improvement (expected decrease in mass of nitrogen delivered to rivers and streams), and is associated with a different spatial configuration of restoration activities. The results of this analysis indicated that tradeoffs between habitat and water quality benefits could be severe. The set of restoration activities that maximized habitat improvement yielded very little water quality improvement, and vice versa. This result came about because of the different spatial relationships embodied in the production functions for the two ecosystem services. Habitat benefits were maximized by wetlands interspersed with uplands, while water quality benefits were maximized by wetlands along the rivers' edges and downstream of drainage areas dominated by agriculture.
The integrated optimization model also was used in a simulated Wetlands Reserve Program (WRP) decision scenario, based on data for landowner offerings to the program for the year 2000 in California. In that year, 84 parcels of land were offered for enrollment in the WRP in the Central Valley. Precise information on the location of the parcels was not available, but their sizes and the number offered in each county were known. This information served as the basis for simulating a set of offered sites: contiguous cells of agriculture were selected randomly, such that their sizes fell within the range of sizes of parcels offered, and the number of simulated sites matched the number of offered parcels in each county. The integrated optimization model was then applied to the set of simulated sites, choosing the subset that maximized habitat quality and then water quality, subject to a budget constraint of $10 million, which is approximately equal to the California WRP budget in an average year. The exercise was repeated multiple times, each time based on a different set of randomly generated sites, to produce a distribution intended to span the range of likely outcomes from the WRP in California in an average year. The results of this analysis showed that tradeoffs between habitat and water quality improvements could be substantial, but they were less severe than in the watersheds case described above. This was because the hypothetical manager in the WRP case was limited to selecting sites from those offered; in the watersheds case the manager could select any set of cells for restoration. Nevertheless, in the WRP case, the water quality benefits resulting from selecting sites to maximize habitat benefits were less than one-third of the maximum water quality benefits, and the habitat benefits resulting from selecting sites to maximize water quality benefits were less than one-half of the maximum habitat benefits. The results of this analysis also showed that a simple heuristic based on maximizing restored wetland area could deliver nearly two-thirds of both maximum habitat and water quality benefits.
In this project, an integrated optimization model was developed that can provide a useful framework for analyzing and prioritizing wetlands restoration activities. This research demonstrated the utility of an optimization approach to prioritizing wetlands conservation activities, as well as the difficulties involved in accounting explicitly for ecosystem services when analyzing environmental policy options. Only two ecosystem services were modeled here-the provision of habitat for mallards in the breeding season, and the attenuation of nutrients in nonpoint source runoff-but others also may be important. For example, another often cited category of ecosystem services from wetlands is the potential flood control benefits they can provide. Future work stemming from this research will include an extension of the hydrologic simulation model to allow estimates of the expected reductions in downstream flooding from restored wetlands.
The results of this research showed that merely maximizing wetland area likely will lead to suboptimal provision of ecosystem services from wetlands. This is because a wetland's size is only one of the factors that determines the level of ecosystem services it provides; the wetland's location-with respect to other wetlands, other land use types, and downstream water bodies-also is important. This research also showed that as a result of these spatial effects, the tradeoffs between different ecosystem services from wetlands restoration could be substantial.
Direct extensions of this research should focus on improving our capacity to: (1) predict the environmental impacts of alternative wetlands conservation options; and (2) apply numerical optimization techniques to real-world problems, which will include substantial nonlinearities and many discrete decision variables. Advances along these lines will allow decisionmakers to identify more cost-effective wetlands management options, thereby increasing the level of environmental benefits delivered per dollar spent towards meeting the nation's stated goal of "no net loss" of wetlands.