Regional-Scale Topsoil Organic Matter Estimation Based on a Geographic Detector Model Using Landsat Data, Pingtan Island, Fujian, China

被引：0

作者：

Fang, Junjun ^{[1
,2
,3
]}

Li, Xiaomei ^{[4
]}

Sha, Jinming ^{[1
]}

Dong, Taifeng ^{[5
]}

Shang, Jiali ^{[5
]}

Shifaw, Eshetu ^{[1
,6
]}

Su, Yung-Chih ^{[1
]}

Wang, Jinliang ^{[7
]}

机构：

[1] Fujian Normal Univ, Coll Geog Sci, Fuzhou 350117, Peoples R China

[2] Chinese Acad Sci, Inst Geog Sci & Nat Resources Res, State Key Lab Resource & Environm Informat Syst, Beijing 100101, Peoples R China

[3] Univ Chinese Acad Sci, Beijing 100049, Peoples R China

[4] Fujian Normal Univ, Coll Environm Sci & Engn, Fuzhou 350117, Peoples R China

[5] Agr & Agrifood Canada, Ottawa Res & Dev Ctr, 960 Carling Ave, Ottawa, ON K1A 0C6, Canada

[6] Wollo Univ, Geog & Environm Studies, POB 1145, Dessie, Ethiopia

[7] Yunnan Normal Univ, Fac Geog, Kunming 650500, Peoples R China

来源：

SUSTAINABILITY | 2023年 / 15卷 / 11期

关键词：

digital soil organic matter mapping; influencing factors; geostatistics; geodetector; remote sensing; arenosols; SOIL PROPERTIES; CARBON; REFLECTANCE; PREDICTION; VEGETATION; VARIABILITY; COVER;

D O I：

10.3390/su15118511

中图分类号：

X [环境科学、安全科学];

学科分类号：

08 ; 0830 ;

摘要：

Understanding the spatial distribution of soil organic matter (SOM) is important for land use management, but conventional sampling methods require significant human and financial resources. How to map SOM and monitor its changes using a limited number of sample points combined with remote sensing techniques that provide long-time series data is crucial. This study aimed to generate a regional-scale near-surface SOM map using 70 soil samples and covariate environmental factors extracted mainly from Landsat 8 OLI. Firstly, the sensitivity of each environmental factor to SOM was tested using a geographic detector model (GDM). Secondly, the tested factors were selected for modeling and mapping by ordinary least squares (OLS) and geographically weighted regression kriging (GWRK). The performance of these two models was compared. Finally, the mapping results of the better model (GWRK) were compared and analyzed with the traditional interpolation results based solely on sampling points to verify the rationality of the proposed method. The results show that three environmental factors, ratio vegetation index (RVI), differential vegetation index (DVI), and terrain roughness (TR), have a strong influence on the spatial variability of SOM. Using these three factors in combination with the GWRK method, a more accurate and refined spatial distribution map of SOM can be obtained. Comparing the SOM maps of GWRK and the traditional interpolation method, the results show that the accuracy of GWRK (R-2 = 0.405; mean absolute error = 0.637, and root mean square error = 0.813) is higher than that of traditional interpolation methods (R-2 = 0.291, MAE = 0.609, and RMSE = 0.863). The spatial recognition rate (fineness) of SOM patches at all levels using the GWRK method increased by more than 73 times compared to the traditional kriging. We conclude that the combination of limited SOM samples, environmental variables, GDM, and GWRK is a pragmatic approach for estimating regional-scale SOM.

引用

页数：18