Date of Final Oral Examination (Defense)
Type of Culminating Activity
Master of Science in Hydrologic Sciences
Nancy F. Glenn, Ph.D.
Alejandro N. Flores, Ph.D.
Douglas J. Shinneman, Ph.D.
Aihua Li, Ph.D.
Remote sensing based quantification of semiarid rangeland vegetation provides the large scale observations required for monitoring native plant distribution, estimating fuel loads, modeling climate and hydrological dynamics, and measuring carbon storage. Fine scale 3-dimensional vertical structural information from airborne lidar and improved signal to noise ratio and radiometric resolution of recent satellite imagery provide opportunities for refined measurements of vegetation structure.
In this study, we leverage a large number of time series Landsat 8 vegetation indices and lidar point cloud - based vegetation metrics with ground validation for scaling aboveground shrub and herb biomass and cover from small scale plot to large, regional scales in the Morley Nelson Snake River Birds of Prey National Conservation Area (NCA), Idaho. The Landsat vegetation indices were trained and linked to in-situ measurements (n = 141) with the random forest regression to impute vegetation biomass and cover across the NCA. We also validated our model with an independent dataset (n = 44), explaining up to 63% and 53% of variation in shrub cover and biomass, respectively. Forty six of the in-situ plots were used in a model to compare the performance of lidar and Landsat data in estimating vegetation characteristics. Our results demonstrate that Landsat performs better in estimating both herb (R2 ~ 0.60) and shrub cover (R2 ~ 0.75) whereas lidar performs better in estimating shrub and total biomass (R2 ~ 0.75 and 0.68, respectively). Using the lidar only model, we demonstrate that lidar metrics based on shrub height have a strong correlation with field-measured shrub biomass (R2 ~ 0.76). We also compare processing the lidar data with raster-based and point cloud-based approaches. The results are scale-dependent, with improved results of biomass estimation at coarser scales with point cloud processing. Overall, the results of this study indicate that Landsat and lidar can be efficiently utilized independently and together to estimate biomass and cover of vegetation in this semi-arid rangeland environment.
Dhakal, Shital, "Assessing the Limitations and Capabilities of Lidar and Landsat 8 to Estimate the Aboveground Vegetation Biomass and Cover in a Rangeland Ecosystem Using a Machine Learning Algorithm" (2016). Boise State University Theses and Dissertations. 1108.