Date of Award

Fall 12-15-2023

Level of Access Assigned by Author

Open-Access Thesis

Degree Name

Master of Science (MS)

Department

Ecology and Environmental Sciences

Advisor

Parinaz Rahimzadeh-Bajgiran

Second Committee Member

Aaron Weiskittel

Third Committee Member

Michael Premer

Abstract

Growing stock volume (GSV) is an important metric for determining economic yield, carbon sequestration and other ecosystem services. GSV has traditionally been estimated in situ by measuring individual trees in a stand. This process is slow and expensive, and, as a result, is not a viable means to estimate GSV on a large scale. It is also not feasible in places that are difficult to access and in places that do not have reliable management records. Multispectral optical sensors mounted on satellites are an important technology for monitoring forest resources because they offer the possibility of measuring forest resources quickly and over large areas. In this study, forest potential productivity was estimated by evaluating 65 variables including several remotely sensed optical variables and site and climate data. Optical variables were Sentinel-2 band 3, band 8a, the Normalized Difference Vegetation Index using bands 4 and 5 (NDVI45) and the Sentinel-2 red-edge position index (S2REP). The variables were used as inputs in a random forest machine learning algorithm. The response variable was constructed using the tree height differences estimated using the National Agricultural Imagery Program (NAIP) orthographic imagery data derived from the NAIP 2018 and NAIP 2021 (ΔNAIP) data. This study was conducted in Maine, USA, where 89% of the land is covered by forests and forest product industry is a significant contributor to the state economy. The best-performing final model to estimate forest productivity (growth), which incorporated Sentinel-2 band 3, the NDVI45, and the S2REP as well as seven site variables, achieved an R² value of approximately 0.56.

Download

Share