Forecasting Wetland Habitat to Support Dynamic Multi-Benefit Water Management Decisions in the Central Valley of California

Abstract:

In arid and semi-arid regions, effective landscape scale wetland conservation requires understanding how biodiversity responds to dynamic fresh water availability. Despite losing 90% of its naturally occurring wetlands primarily to agriculture, California’s Central Valley retains critical wetlands for migratory waterbirds and wetland-dependent species. Balancing water management across the Central Valley for diverse human and habitat needs requires frequent, landscape-scale environmental data that includes satellite-based observations. To support wetland management decisions, we used Landsat satellite data and machine learning to build habitat suitability models for waterbirds and an imperiled snake as a function of dynamic freshwater supply and land cover types. Our models confirmed that open water and wetlands are essential drivers of landscape-scale habitat suitability and species distribution. Comparing habitat 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. Hence, 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 that solve for multiple benefits.