University of Twente Student Theses

Login

Tropical Forest Height Estimation Using L-and P-Band Polarimetric Synthetic Aperture Radar Backscatters with Machine Learning Models

Bhuiyan, Rezaul Hasan (2024) Tropical Forest Height Estimation Using L-and P-Band Polarimetric Synthetic Aperture Radar Backscatters with Machine Learning Models.

[img] PDF
2MB
Abstract:Forest Canopy Height (CH) is an important biophysical parameter for effective forest management and conservation efforts. This study presents a Fully Convolutional Network (FCN)-based approach for estimating CH using L- and P-band polarimetric synthetic aperture radar (PolSAR) backscatters and their combinations. Specifically, a customized UNet architecture, tailored to the unique characteristics of SAR data, is employed to estimate CH in both heterogeneous and homogeneous forest sites, Lope and Mabounie respectively, located in Gabon. Results indicated that combinations of L- and P-band polarimetric backscatters led to CH predictions that were more accurate compared to single-band retrievals, with dual-band combinations producing Root Mean Square Error (RMSE) values of 4.03 m for Lope and 3.78 m for Mabounie. The estimation accuracies from the combinations of Synthetic Aperture Radar (SAR) bands were consistent across the two study areas, whereas the retrieval performance varied for individual bands. P-band-based retrievals were more accurate than L-band for the homogeneous Mabounie site (RMSE of 4.26 m vs. 4.63 m). However, for the heterogeneous Lope site, no significant RMSE difference was found between L- and P-band models. Upon comparison with other machine learning models, it was observed that the customized UNet model produced RMSE values three times lower than those of Random Forest (RF) and Light Gradient Boosting Machine (LGBM). These results are relevant in the context of upcoming long-wavelength SAR missions, such as the European Space Agency (ESA) BIOMASS and NASA-ISRO Synthetic Aperture Radar (NISAR), which could potentially be used for global forest canopy height mapping.
Item Type:Essay (Master)
Faculty:ITC: Faculty of Geo-information Science and Earth Observation
Subject:43 environmental science
Programme:Geoinformation Science and Earth Observation MSc (75014)
Link to this item:https://purl.utwente.nl/essays/102965
Export this item as:BibTeX
EndNote
HTML Citation
Reference Manager

 

Repository Staff Only: item control page