This Is AuburnElectronic Theses and Dissertations

Regional-scale Mapping of Canopy Height and Aboveground Biomass using ICESat-2 and Landsat-8 Data

Date

2022-12-02

Author

Tiwari, Kasip

Type of Degree

Master's Thesis

Department

Forestry and Wildlife Science

Restriction Status

EMBARGOED

Restriction Type

Full

Date Available

12-02-2024

Abstract

The Ice, Cloud, and land Elevation Satellite- 2 (ICESat-2), launched in September 2018, is a follow-on mission to ICESat, which operated from 2003 to 2009. While ICESat-2 observations do not provide full coverage, integrating these observations with other wall-to-wall data, like Landsat imagery, may be used to achieve spatially comprehensive information. The availability of canopy height information from ICESat-2’s land and vegetation product, or ATL08, presents opportunities for developing full coverage products over broad spatial scales. The primary goal of this study was to develop wall-to-wall 30-meter canopy height and aboveground biomass (AGB) products over the southeastern US, for the Southeastern Plains ecoregion and Middle Atlantic Coastal Plain ecoregion. First, Landsat-8 bands and derived vegetation indices (normalized difference vegetation index (NDVI), enhanced vegetation index (EVI), and modified soil-adjusted vegetation index (MSAVI)) along with National Land Cover Database’s canopy cover. Digital elevation models (DEMs) were used to extrapolate ICESat-2 canopy height from tracks to the regional level. Statistical values were calculated to quantify the accuracy of ICESat-2 canopy heights and to compare results derived from two modeling approaches: random forest (RF) and regression kriging (RK). The best model (RK) estimated canopy heights with an R2 value of 0.69 and RMSE of 3.49m for independent validation and R2 value of 0.47 and RMSE value of 5.96m when compared with reference airborne lidar-derived canopy heights. The extrapolated canopy height from ICESat-2 using the RK model, along with the Landsat variables, DEMs, and canopy cover, were then used as predictor variables to map AGB using machine learning and geostatistical approaches. Four models were used in this study, i.e., machine learning (RF and support vector machine (SVM)) and geostatistical (random forest regression kriging (RFRK) and support vector machine regression kriging (SVMRK)) to model and extrapolate AGB across the study area. Resulting R2 values ranged from 0.34 to 0.61, and RMSEs were between 22 Mg/ha and 31 Mg/ha. AGB estimated using the SVMRK model was significantly better than any other model, showing SVMRK's great potential for mapping AGB regionally. The results suggest feasibility for the implementation of RK methods over a larger spatial extent and potential for combining other remote sensing and satellite data for future monitoring of canopy height and AGB dynamics.