Fig. 1: Ridge to reef modeling framework.

a Land cover maps showing the areas where watershed interventions are feasible given biophysical and zoning opportunities and constraints, including projected native forest protection within existing and planned fenced areas (forest green = native forest within protected areas) and potential agroforest transition areas (yellow = dry multi-strata agroforest; blue = mesic multi-strata agroforest; pink = wet multi strata agroforest; orange = koa silvopasture) coupled with the modeled coral reef habitat area (grey) with survey locations used to parameterize the blue food model. The light grey represents the land areas that were not modeled due to no land use cover change (Kahoʻolawe) and minimal surface runoff (West Hawaiʻi). b Sediment retention model, which incorporates rainfall, Digital Elevation Model (DEM), and soil characteristics, and land cover information to compute the annual sediment loads retained by vegetation and reaching the ocean (t/year). c A water quality model diffuses the change in annual sediment loads from the sediment retention model into the marine environment (Total Suspended Sediment [TSS]) as a function of wave energy, currents, and distance from river mouths. d The reef fish surveys are summarized by habitat type at the island scale to obtain the fish standing stock (g/m2) as a function of change in TSS. e The average fish standing stock by island is linked to the Standardized CPUE using a regression model (Supplementary Fig. 3). f The regression model is then used to predict CPUE at the reef pixel level and multiplied with total fishing effort to obtain total reef fish catch (kg by island) and number of seafood meals under different levels of watershed management.