These rasters represent landscape-scale (i.e., 3.6 x 3.6 km window) predictions of unprotected grassland loss (i.e., grass, shrub, herbaceous wetland cover) from 2019-2038 at 270 m resolution for the contiguous US. Each 270 m pixel value is a grassland loss prediction for the 3.6 x 3.6 km area around the pixel. These were generated by training random forest regression models by relating unprotected grassland loss from 2001-2019 to a suite of covariates from ~2001, and then applying these models to updated surface data ~2019 to make loss predictions for the period of 2019-2038. We developed models for two loss metrics:

1- Total loss – predictions represent ha unprotected grass loss per 3.6 x 3.6 km area around each pixel

2- Proportion loss – predictions represent the amount of unprotected grass lost out of what was available to be lost per 3.6 x 3.6 km area around each pixel

See the included README document for more information.