RCCZO -- Land Cover, LiDAR, Vegetation -- Biomass Estimate of Sagebrush -- Reynolds Creek Experimental Watershed -- (2012-2012)

Vegetation biomass estimates across drylands at regional scales are critical for ecological modeling, yet the low-lying and sparse plant communities characterizing these ecosystems are challenging to accurately quantify and measure their variability using spectral-based aerial and satellite remote sensing. To overcome these challenges, multi-scale data including field-measured biomass, terrestrial laser scanning (TLS) and airborne laser scanning (ALS) data, were combined in a hierarchical modeling framework. Data derived at each scale were used to validate an increasingly broader index of sagebrush (Artemisia tridentata) aboveground biomass. First, two automatic crown delineation methods were used to delineate individual shrubs across the TLS plots. Second, three models to derive shrub volumes were utilized with TLS data and regressed against destructively-sampled individual shrub biomass measurements. Third, TLS-derived biomass estimates at 5 m were used to calibrate a biomass prediction model with a linear regression of ALS-derived percent vegetation cover (adjusted R2 = 0.87, p < 0.001, RMSE = 3.59 kg). The ALS prediction model was applied to the study watershed and evaluated with independent TLS plots (adjusted R2 = 0.55, RMSE = 4.01 kg, normalized RMSE = 35%). The biomass estimates at the scale of 5 m is sufficient for capturing the variability of biomass needed to initialize models to estimate ecosystem fluxes, and the contiguous estimates across the watershed support analyzing patterns and connectivity of these dynamics. Our model is currently optimized for the sagebrush-steppe environment at the watershed scale and may be readily applied to other shrub-dominated drylands, and especially the Great Basin, U.S., which extends across five western states. Improved derived metrics from ALS data and collection of additional TLS data to refine the relationship between TLS-derived biomass estimates and ALS-derived models of vegetation structure, will strengthen the predictive power of our model and extend its range to similar shrubland ecosystems.