Biomass estimates for shrub-dominated ecosystems in southern California have been generated at national and statewide extents. However, existing data tend to underestimate biomass in shrub vegetation types, are limited to one point in time, or estimate aboveground live biomass only. In this study, we extended our previously developed estimates of aboveground live biomass (AGLBM) based on the empirical relationship of plot-based field biomass measurements to Landsat NDVI and multiple environmental factors to include other vegetative pools of biomass. AGLBM estimates were made by extracting plot values from elevation, solar radiation, aspect, slope, soil type, landform, climatic water deficit, evapotranspiration, and precipitation rasters and then using a Random Forest model to estimate per-pixel AGLBM across our southern California study area. By substituting year-specific Landsat NDVI and precipitation data, we created a stack of annual above ground live biomass raster layers for each year from 2001 - 2021. Using these AGLBM data as a foundation, we developed decision rules to estimate belowground, standing dead, and litter biomass pools. These rules were based on relationships between AGLBM and the biomass of the other vegetative pools derived primarily from peer reviewed literature and an existing spatial dataset. For shrub vegetation types (our primary focus), rules were based on literature estimates by the post-fire regeneration strategy of each species (obligate seeder, facultative seeder, obligate resprouter). Similarly, for non-shrub vegetation types (grasslands, woodlands) we used literature and existing spatial datasets specific to each vegetation type to define rules to estimate the other pools from AGLBM. Using a Python language script that accessed ESRI raster GIS utilities, we applied decision rules to create raster layers for each of the non-AGLBM pools for the years 2001-2021. The resulting spatial data archive contains a zipped file for each year; each of these files contains four 32-bit tiff files for each of the four biomass pools (AGLBM, standing dead, litter, and below ground). The biomass units are g/m2 . We estimated the uncertainty of our biomass data by conducting a Monte Carlo analysis of the inputs used to generate the data. Our Monte Carlo technique used randomly generated values for each of the literature-based and spatial inputs based on their expected distribution. We conducted 200 Monte Carlo iterations, which produced percent uncertainty values for each of the biomass pools. Results showed, using 2010 as an example, mean biomass for the study area and percent uncertainty for each of the pools as: AGLBM (905.4 g/m2 , 14.4%); standing dead (644.9 g/m2 , 1.3 %); litter (731.2 g/m2 , 1.2 %); and belowground (776.2 g/m2 , 17.2 %). Because our methods are consistently applied across each year, the data produced can be used to inform changes in biomass pools due to disturbance and subsequent recovery. As such, these data provide an important contribution to supporting the management of shrub-dominated ecosystems for monitoring trends in carbon storage and assessing the impacts of wildfire and management activities, such as fuel management and restoration. There are no copyright restrictions on the data set; please cite this paper and the data package when using these data. This article is protected by copyright. All rights reserved.
Read full abstract