Improved Radiometric Capabilities of Landsat 8, Coupled With Lidar, Estimate Semi-arid Rangeland Biomass and Cover

Abstract:

Remote sensing based quantification of semi-arid rangeland vegetation provides the large scale observations necessary for monitoring native plants distribution, estimating fuel loads and measuring carbon storage. Improved signal to noise ratio and radiometric resolution of recent satellite imagery and fine scale 3-dimensional information from lidar provides opportunities for refined measurements of vegetation structure. We leverage a large number of Landsat 8 and lidar-based metrics for prediction of biomass and cover of shrubs in the semi-arid rangeland of the western United States. Time-series Landsat 8 images were used to develop 20 ratio based vegetation indices. Similarly, 35 vegetation metrics, including metrics based on numerical values (e.g. elevation, canopy height) and on density of points (e.g. canopy density) were developed from airborne lidar point clouds. These vegetation indices and metrics were trained and linked to insitu measurements (n=141) with the Random Forest regression to impute biomass and cover map across a large scale. We also validate our model with an independent data-set (n=44), explaining up to 63% of variation in cover and 53% in biomass. The preliminary results suggest that Landsat performs better in estimating vegetation cover whereas lidar performs better in estimating biomass with no significant relationship to topographic variables (e.g. slope, aspect and elevation). We further compare our results with historical fire data to show that both biomass and cover decreases with the increase of fire frequency in the study site. This study demonstrates the new opportunities of using Landsat 8 with established lidar approaches to better quantify vegetation in semiarid ecosystems.