Version 1
: Received: 24 September 2024 / Approved: 24 September 2024 / Online: 25 September 2024 (17:52:01 CEST)
How to cite:
Sharma, S.; Khanal, P. Predicting Above-Ground Biomass of Forest in South Carolina: Integrating Remote Sensing, Machine Learning, and Interpolation Techniques. Preprints2024, 2024091867. https://doi.org/10.20944/preprints202409.1867.v1
Sharma, S.; Khanal, P. Predicting Above-Ground Biomass of Forest in South Carolina: Integrating Remote Sensing, Machine Learning, and Interpolation Techniques. Preprints 2024, 2024091867. https://doi.org/10.20944/preprints202409.1867.v1
Sharma, S.; Khanal, P. Predicting Above-Ground Biomass of Forest in South Carolina: Integrating Remote Sensing, Machine Learning, and Interpolation Techniques. Preprints2024, 2024091867. https://doi.org/10.20944/preprints202409.1867.v1
APA Style
Sharma, S., & Khanal, P. (2024). Predicting Above-Ground Biomass of Forest in South Carolina: Integrating Remote Sensing, Machine Learning, and Interpolation Techniques. Preprints. https://doi.org/10.20944/preprints202409.1867.v1
Chicago/Turabian Style
Sharma, S. and Puskar Khanal. 2024 "Predicting Above-Ground Biomass of Forest in South Carolina: Integrating Remote Sensing, Machine Learning, and Interpolation Techniques" Preprints. https://doi.org/10.20944/preprints202409.1867.v1
Abstract
This study evaluates the effectiveness of a Random Forest model for predicting above-ground biomass in South Carolina (SC), utilizing diverse remote sensing and climatic data sources. SC, with its humid subtropical climate and varied geography, including the Atlantic coastal plain, Piedmont, and Blue Ridge Mountains, poses unique challenges for biomass estimation. We integrated global biomass datasets for 2010, MODIS vegetation indices (NDVI and EVI), Leaf Area Index (LAI) from MOD15A2H, and climate data from TerraClimate. The model was trained using 2010 data and applied to 2022 datasets to assess biomass changes. To validate the model, plot-level biomass estimates from the 2023 FIA data were interpolated using Inverse Distance Weighting (IDW). Performance evaluation showed a strong positive correlation between predicted and observed biomass, with a correlation coefficient of 0.77 and an R² value of 0.62, indicating that the model explains 62% of the variability in biomass. Comparison with IDW-interpolated biomass data resulted in a correlation coefficient of 0.64, confirming the model's validity. Although the Random Forest model demonstrated reliable predictions, the study suggests potential improvements by incorporating additional data sources and advanced modeling techniques. The findings emphasize the value of integrating remote sensing data, machine learning, and interpolation methods to enhance biomass estimation accuracy. This research provides crucial insights into biomass distribution in SC and establishes a basis for future studies on forest monitoring and carbon accounting, highlighting the importance of combining various data sources for comprehensive environmental analysis.
Keywords
Machine learning; random forest; IDW; FVS; GEE
Subject
Biology and Life Sciences, Forestry
Copyright:
This is an open access article distributed under the Creative Commons Attribution License which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
