研究目的
To test the ability of SAR, optical and textural data to estimate forest parameters (biomass, height, diameter and density) and to evaluate the improvement of combining these remote sensing data.
研究成果
L-band SAR is best for AGB retrieval, C-band SAR for height, DBH, and density. Combining SAR and textural data improves estimations, but adding C- and L-band together does not show significant improvement. The approach is promising for forest monitoring with freely available remote sensing data.
研究不足
The study is limited to monospecific pine forest stands; further work is needed for other forest structures and species. Temporal information from Sentinel-1 time series was not analyzed, and Random Forest Regression should be tested. The decrease in L-band HV backscatter for high DBH values is not explained.
1：Experimental Design and Method Selection:
The study used a regression approach with Support Vector Regression (SVR) to combine multisensor remote sensing data for estimating forest parameters.
2：Sample Selection and Data Sources:
Ground data were collected from monospecific pine forest stands in the Landes de Gascogne forest in France, including dendrometric parameters like tree density, DBH, height, and calculated aboveground biomass.
3：List of Experimental Equipment and Materials:
Remote sensing data from Sentinel-1 (C-band SAR), Sentinel-2 (optical), ALOS-PALSAR (L-band SAR), and SPOT-6/7 (optical for texture) were used.
4：Experimental Procedures and Operational Workflow:
Data were processed (e.g., calibration, orthorectification, filtering), features were extracted (e.g., backscatter coefficients, spectral indices, texture indices), and SVR was applied with leave-one-out validation.
5：Data Analysis Methods:
Root mean square error (rmse), relative error (rrmse), and coefficient of determination (r2) were used to evaluate estimations.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
