Predicting Forest Canopy Height Using GEDI LiDAR Based Machine Learning Technique Over Similipal Biosphere, India

One of the elements of an ecosystem's functioning that is thought to be most important is its forest ecosystem. It offers a range of services, including climate regulation, watershed preservation, and the provision of food and materials. With the use of Sentinel 1A, Sentinel 2B, the SRTM digital elevation model, and GEDI LiDAR data, the final forest canopy height (FCH) map was created in Similipal Biosphere Reserve, Odisha, India. Fourteen parameters of varied relevance are predicted using three machine learning models (Random Forest (RF), eXtreme Gradient Boosting (XGB), and Classification and Regression Tree (CART) for FCH estimation. The XGB tree model (R2 = 0.524, RMSE = 6.553) and the CART method (R2 = 0.891, RMSE = 2.293) had the next highest goodness of fit, behind the RF model (R2 = 0.745, RMSE = 4.441). Another crucial element that may be utilized to track a forest's biomass and aboveground carbon storage is the height of the canopy. These findings imply that incorporating various variables and data sources enhances canopy cover and height prediction accuracy when compared to using a single data source. The findings of this study may be useful in developing forest management plans that promote the sustainable use of forest resources in near real time basis. Future research will include additional parameters, such as meteorological, vegetation biophysical, geo-environmental, and tree species data. Advanced methods, like hybrid machine learning and deep learning, as well as ground-based UAVs and ICESat platforms, will be helpful in estimating FCH quickly.