The High-Resolution Global Land Surface Satellite (Hi-Glass) Products Suite

Published: Jan. 1, 2024

Language: Английский

Research on Leaf Area Index Inversion Based on LESS 3D Radiative Transfer Model and Machine Learning Algorithms

Remote Sensing, Journal Year: 2024, Volume and Issue: 16(19), P. 3627 - 3627

Published: Sept. 28, 2024

The Leaf Area Index (LAI) is a critical parameter that sheds light on the composition and function of forest ecosystems. Its efficient rapid measurement essential for simulating estimating ecological activities such as vegetation productivity, water cycle, carbon balance. In this study, we propose to combine high-resolution GF-6 2 m satellite images with LESS three-dimensional RTM employ different machine learning algorithms, including Random Forest, BP Neural Network, XGBoost, achieve LAI inversion stands. By reconstructing real stand scenarios in model, simulated reflectance data blue, green, red, near-infrared bands, well data, fused some inputs train models. Subsequently, used remaining measured validation prediction inversion. Among three Forest gave highest performance, an R2 0.6164 RMSE 0.4109, while Network performed inefficiently (R2 = 0.4022, 0.5407). Therefore, ultimately employed algorithm perform generated spatial distribution maps, achieving innovative, efficient, reliable method

Language: Английский

Optimizing GEDI Canopy Height Estimation and Analyzing Error Impact Factors Under Highly Complex Terrain and High-Density Vegetation Conditions

Forests, Journal Year: 2024, Volume and Issue: 15(11), P. 2024 - 2024

Published: Nov. 17, 2024

The Global Ecosystem Dynamics Investigation (GEDI) system provides essential data for estimating forest canopy height on a global scale. However, factors such as complex topography and dense can significantly reduce the accuracy of GEDI estimations. We selected South Taihang region Henan Province, China, our study area proposed an optimization framework to improve estimation accuracy. This includes correcting geolocation errors in footprints, screening analyzing features that affect errors, combining two regression models with feature selection methods. Our findings reveal error 4 6 m footprints at orbital scale, along overestimation region. Relative (RH), waveform characteristics, topographic features, cover influenced error. Some studies have suggested estimates areas high lead underestimation, found increased higher terrain vegetation. model’s performance improved after incorporating parameter into model. Overall, R2 best-optimized model was from 0.06 0.61, RMSE decreased 8.73 2.23 m, rRMSE 65% 17%, resulting improvement 74.45%. In general, this reveals affecting vegetation cover, premise minimizing errors. Employing enhanced estimates. also highlighted crucial role improving precision estimation, providing effective approach monitoring regions conditions. Future should further classification tree species expand diversity sample test estimated by different structures, consider distortion optical remote sensing images caused rugged terrain, mine information waveforms so enhance applicability more diverse environments.

Language: Английский