Application of decision tree based on multisource remote sensing data in land use/cover classification in the Shiyang River Basin

Based on multi-source remote sensing data, combining land use maps and time-series NDVI characteristic spectrum lines of different landform types, drawing support from elevation information, the Land use/cover (LUC) classification decision tree of the Shiyang River Basin (SYRB) was constructed. The LUC of SYRB was divided into eight types named evergreen forest (EF), deciduous forest (DF), farmland (FL), desert area (DA), oasis-desert transition (ODT), high-coverage grassland (HG), low-coverage grassland (LG) and ice-snow (IS) areas. The statistical characteristics of Land use/cover change (LUCC) from 2001 to 2020 are obtained. The conclusions are as follows: 1) The classification method of decision tree based on multi-source remote sensing data resulted in a satisfactory precision, gained a high accuracy of 81%. 2) From 2001 to 2020, the area of FL, EF, DF, HG and ODT showed increasing trend, and the area of LG, IS and DA decreased to a certain extent. The variation range of FL and IS was small with 0.57×104 hm2 /10a and -0.16×104 hm2 /10a, respectively; while the HG and DA was large with 6.08×104 hm2 /10a and -10.18×104 hm2 /10a, respectively.