The goal of my dissertation was to explore how scale influences stream restoration prioritization strategies for an anadromous species and identify influential uncertainties that exist at different scales. My objectives were to (1) produce a comprehensive review of the Chinook salmon management challenges in California’s Central Valley and identify the those related to scale of the management, (2) apply a structured decision making (SDM) to a large scale, spatially implicit stream restoration decision problem and derive an optimal stream restoration strategy, (3) apply a SDM to a small scale, spatially explicit stream restoration decision problem and use dynamic programming to derive an optimal stream restoration strategy, and (4) develop an approach to derive optimal policies for multi-scale stream restoration decision problem with multiple decision makers working in a hierarchy.
Issues of scale can create distinct problems in natural resource management. I used salmon management in the Central Valley of California an example where scale mismatches that have hindered conservation goals. Salmon stocks in California’s Central Valley have been declining steadily over the last century, which has resulted the congressional establishment of the Central Valley Plan Improvement Act (CVPIA) in 1992 to address the declines. Despite the oversight of the CVPIA, fisheries management in the basin has remained largely uncoordinated and unstructured with management actions often occurring simultaneously and at potentially conflicting scales. Having such large differences in spatial scales meant that any reduction of system uncertainties wasn’t necessarily transferable to other populations of anadromous fish within the Central Valley. The hierarchical structure of the CVPIA and the entities that implement CVPIA related actions provide an opportunity to evaluate how scale may influence restoration decision making.
I developed a large scale, spatially implicit decision model to evaluate the effects of potential habitat restoration projects on populations of fall-run Chinook salmon in California’s Central Valley. The extent of the model was the entire Central Valley and the grain was an individual watershed, 25 in total. Large scale natural resource management problems require special considerations relative to smaller scale problems due in part to the fact that uncertainties tend to increase as spatial scale increases. The model was primarily parameterized with expert judgement due to a lack of available empirical data at the watershed scale. This model and the decision alternatives were formatted as a Markov decision model (MDP) that I solved using dynamic programming and policy iteration. The results of the policy optimization suggest that focusing multiple restoration efforts on a small set of watersheds is the most effective habitat restoration strategy.
Most stream restoration efforts occur on a small spatial scales, often on reaches less than 1 km. I developed a fine scale, spatially explicit structured decision model to derive a state-specific stream restoration strategy for a population of Chinook salmon from a stream in lower American River. The decision problem was represented as Markov decision problem and I used dynamic programming to derive a state-specific, optimal policy for individual reaches within the study stream. The optimal policies depended on four pieces of observable information in a given reach: the amount of spawning habitat, the amount of juvenile rearing habitat, the average number of redds counted over a 5 year period, and the temperature suitability of the reach. Implementing the optimal policy during a 100 year simulation resulted in significant increases in natural production compared to a scenario where no actions were taken over the same time horizon.
Decision problems in natural resource management often involve several, interconnected decision makers, usually working at different temporal and spatial scales. Multitime-scale Markov decision processes (MMDPs) provide a framework to derive optimal decisions from hierarchically structured sequential decision making processes. The work in this study bridges the gap between large and small-scale decision models in natural resource management by applying a MMDP to a Chinook salmon management problem in CVPIA streams with two tiers of decision makers. The fundamental objective of each tier of decision makers was to maximize the production of natural origin Chinook salmon. The decision problem was structured with an upper tier decision maker (large scale) allocating funds to lower tier decision makers (fine scale) who actually implement on the ground restoration projects. The upper tier optimal policy identified optimal resource allocation strategies that favored providing funds to watersheds with high juvenile survival despite high costs.