Open AccessAnalyzing Variation of Soil Salinity Content in the Agricultural Areas: A Factorial Analysis Based Random Forest Estimation Method
Author(s) -
H Wang,
Y P Li
Publication year - 2021
Publication title -
iop conference series. earth and environmental science
Language(s) - English
Resource type - Journals
eISSN - 1755-1307
pISSN - 1755-1315
DOI - 10.1088/1755-1315/793/1/012032
Subject(s) - random forest , soil salinity , environmental science , support vector machine , salinity , irrigation , soil science , statistics , agricultural engineering , computer science , mathematics , soil water , machine learning , agronomy , engineering , ecology , biology
In this study, to identify the complex relationship between soil salinity content (SSC) and factors, and efficiently quantify the individual effects of factors on SSC, a random-forest-based factorial analysis (RF-FA) method by incorporating random forest (RF) and factorial analysis (FA) is developed. The RF-FA is applied in a case study. Results reveal that: (i) Compared with Back Propagation Neural Network (BPNN) and Support Vector Regression (SVR), RF is a more robust model for SSC simulation; (ii) The contributions of factors to the variation of SSC follow the order: Elevation (14.68%) > LST (11.69%) > Albedo_N (11.41%) > SI_2 (10.71%) >SAVI (10.21%), indicating topography and temperature are the macro factors controlling SSC; hence, providing sufficient irrigation water is necessary to mitigate soil salinization. The findings can help make effective strategies to relieve the soil salinization of the farmland and support the sustainable development of agriculture.