Combining ALOS PALSAR and AVNIR-2 data for effective land use/land cover classification in Jharia coalfields region

There are various classification techniques available which produce desired results. However, some of the land use/land cover (LU/LC) classes are not discernible in such classifications. The present study attempts for improving LU/LC classification accuracy by applying data fusion techniques. For this, we considered combinations of: (1) Synthetic Aperture Radar (SAR) multi-looked intensity and optical, (2) backscatter with optical and (3) terrain corrected backscatter with optical data. The fusion of terrain corrected backscatter with optical has been considered in this study to negate the effect of topographic undulations on backscatter. The classification accuracy for combinations of cross-polarised terrain corrected backscatter data with Advanced Visible and Near-Infrared Radiometer type 2 (AVNIR-2) (90.33%), co-polarised terrain corrected backscatter data with AVNIR-2 (89.66%), cross-polarised backscatter data with AVNIR-2 (89.0%) and cross-polarised multi-look intensity with AVNIR-2 (87.0%) were found to be better than classified outputs of AVNIR-2 data alone (84.6%), combinations of co-polarised backscatter and AVNIR-2 data (82.7%) and co-polarised multi-look intensity with AVNIR-2 data (80.1%), and combinations of multi-date terrain corrected backscatter (80.66%), multi-date co-polarised backscatter (80.0%) and multi-date co-polarised multi-look intensity (79.0%). The highest accuracy achieved in LU/LC classification is with cross-polarised terrain corrected backscatter with AVNIR-2 (90.33%) data. Data fusion techniques can be an alternative for LU/LC classification. © 2016 Informa UK Limited, trading as Taylor & Francis Group.