An efficient retrieval method on Google Earth Engine and comparison with hybrid methods: a case study of leaf area index retrieval

文献类型： 外文期刊

第一作者： Li, Sijia

作者： Li, Sijia;Tang, Zhiguang;Ma, Kaisen;Wang, Zhenyi;Li, Wenjuan

作者机构：

关键词： Google Earth Engine; remote sensing retrieval; look-up table (LUT); hybrid methods; LAI retrieval using LUT

期刊名称：INTERNATIONAL JOURNAL OF DIGITAL EARTH （ 影响因子：4.9； 五年影响因子：4.7 ）

ISSN： 1753-8947

年卷期： 2025 年 18 卷 1 期

页码：

收录情况： SCI

摘要： Google Earth Engine (GEE) has emerged as a powerful tool offering extensive resources and algorithms for large-scale retrieval, facilitating various applications. However, these algorithms cannot quantify the uncertainty of estimates. This study proposes an efficient retrieval method for GEE based on the look-up table (LUT) that can generate uncertainty and take leaf area index estimation as a case. The performances of LUT and hybrid methods, including random forest (RF), gradient boosting regression tree (GBRT), classification and regression tree (CART), support vector regression (SVR), and Gaussian process regression (GPR), were evaluated on GEE by simulation experiments. LUT and CART have superior efficiency. RF, GBRT, and SVR exhibit higher accuracies, with R2 of 0.68, 0.64, and 0.63, than LUT, CART, and GPR (0.60, 0.59, and 0.53). LUT accuracy improves as RF by introducing nonlinear stretching in LUT establishment, which considers the nonlinearity between target variable and features. LUT uncertainty is less than GPR's, and their influencing factors differ. LUT demonstrates a smaller runtime increase with increasing feature numbers. LUT, GPR, and CART exhibit high parallel capabilities. This study offers GEE a method balancing uncertainty quantification, efficiency, and accuracy. Further studies on other retrieval tasks and evaluations with ground truth data are needed.

分类号：

  • 相关文献
作者其他论文 更多>>

微信小程序