Forecasting Wetland Habitat to Support Multi-Species Management Decisions in the Central Valley of California

In arid and semi-arid regions, landscape scale wetland conservation requires understanding how biodiversity responds to dynamic fresh water availability and human actions to enhance wetland habitat. Despite losing 90% of the naturally occurring wetlands, California's Central Valley retains critical wetlands for migratory waterbirds and wetland dependent species. Balancing diverse habitat needs with water management across the Central Valley requires frequent, landscape-scale environmental data; data that is made possible by satellite-based observations. To support wetland management decision-making, we used Landsat satellite time series (2000-2018) and machine learning to build habitat suitability models for waterbirds and snakes as a function of dynamic freshwater supply and land cover types. Our models highlighted that the availability of open water and wetlands are essential drivers of species distribution and suitability. Comparing suitability between wet and dry years across multiple avian species, we quantify where habitat has historically declined in dry years and which land cover types are most vulnerable to annual variation in water availability. While models of historic patterns are instructive, decisions about where to focus water resources for conservation are being made two to six months in advance of implementation. Building forecast models of wetland habitat availability using within-year indicators (e.g. projected annual run-off; water year type) and the Landsat time series, we applied our species models to develop habitat suitability maps for species 6-9 months into the future of a given water management year. These forecasts are being integrated through a spatially-explicit conservation prioritization framework to support improved, coordinated, landscape-scale dynamic conservation decisions across multiple benefits.