Semi-automated object-based landform classification modelling in a part of the Deccan Plateau of central India

Abstract

ABSTRACTLandform mapping holds significance in governing boundary conditions for the underlying processes operative in the fields of natural resource management, yet the automation in recognizing landform occurrence remains difficult. Geospatial object-based image analysis (GEOBIA) technique has evolved as a promising tool for addressing the issue. Majority of the GEOBIA-based landform classification studies document generic approach. The present study undertaken in Katol Tehsil of Nagpur District, a part of Deccan Plateau of central India aims at knowledge-based modelling through a multi-scale mapping workflow comprising multi-resolution segmentation (input raster dataset of IRS-P6 LISS-IV image and Cartosat-1 digital terrain model), knowledge-based classification, and accuracy assessment against a reference landform map. Contour- and drainage-based relative topographic position zone is derived in a novel attempt. Finally, knowledge-based rules are framed using the primary terrain parameters of elevation, slope, profile curvature, and drainage for deriving final output. The results of landform classification indicate the dominance of erosive landform over depositional one; maximum area of 6244 ha being under pediment. An accuracy assessment exercise is carried out in a watershed occurring in the study area, which shows very good statistical agreements between modelled and reference landforms including partial detection. The key constraint of this knowledge-based modelling is its limited adaptability to only localized conditions. However, such kind of object-based and knowledge-based analyses have immense potential with the increasing availability of finer resolution remote-sensing data products that demand the alternative paths of deriving objects that are made up of several pixels.