Chaparral shrublands are the dominant wildland vegetation type in Southern California and the most extensive ecosystem in the state. Disturbance by wildfire and climate change have created a dynamic landscape in which biomass mapping is key in tracking the ability of chaparral shrublands to sequester carbon. Despite this importance, most national and regional scale estimates do not account for shrubland biomass. Employing plot data from several sources, we built a random forest model to predict aboveground live biomass in Southern California using remote sensing data (Landsat Normalized Difference Vegetation Index (NDVI)) and a suite of geophysical variables. By substituting the NDVI and precipitation predictors for any given year, we were able to apply the model to each year from 2000 to 2019. Using a total of 980 field plots, our model had a k-fold cross-validation R2 of 0.51 and an RMSE of 3.9. Validation by vegetation type ranged from R2 = 0.17 (RMSE = 9.7) for Sierran mixed-conifer to R2 = 0.91 (RMSE = 2.3) for sagebrush. Our estimates showed an improvement in accuracy over two other biomass estimates that included shrublands, with an R2 = 0.82 (RMSE = 4.7) compared to R2 = 0.068 (RMSE = 6.7) for a global biomass estimate and R2 = 0.29 (RMSE = 5.9) for a regional biomass estimate. Given the importance of accurate biomass estimates for resource managers, we calculated the mean year 2010 shrubland biomasses for the four national forests that ranged from 3.5 kg/m2 (Los Padres) to 2.3 kg/m2 (Angeles and Cleveland). Finally, we compared our estimates to field-measured biomasses from the literature summarized by shrubland vegetation type and age class. Our model provides a transparent and repeatable method to generate biomass measurements in any year, thereby providing data to track biomass recovery after management actions or disturbances such as fire.
New Biomass Estimates for Chaparral-Dominated Southern California Landscapes