Tracing the spatio-temporal evolution of the Merapi 2010 erupted deposits based on object-oriented classification and object-based image analysis of multi-temporal VHR optical and ALOS radar imagery

Abstract:

We compare the extent to which VHR optical and radar images delineate the eruption impacts and trace the evolution of erupted deposits on active volcanoes. We could identify about 75% of the 2010 Merapi erupted deposits recognized in traditional geological mapping using object-oriented classification and spectral indices on sub-metric GeoEye and Pléiades images. We recognized sixteen PDC depositional units including high-energy surge deposits on the upper south flank, valley-confined BAF deposits channeled in the Gendol River, and overbank BAF with ash-cloud surge deposits on valley margins. We used an innovative method to map PDC and tephra-fall deposits exploiting direct- and cross-polarized L-band SAR data from ALOS-PALSAR before and after the eruption and combining changes in amplitude of the radar signal with temporal decorrelation. Deposits were separated according to increase or decrease in ground backscattering in direct (HH) and cross (HV) polarizations. The maximum likelihood classification applied to ALOS images provided a result consistent with previous studies with 70% classification accuracy for deposits overall. Scatter diagrams of NDWI, NDVI and NDRSI from three VHR images and morphometric analysis of the initial drainage network enabled us to trace the spatio-temporal evolution (2010-2012) of impacted areas against re-vegetation and surficial erosion. In two years after the eruption, the drainage network was fully developed in the upper catchment devastated by high energy surges but far less developped on fans formed by overbank BAF deposits in the middle valley, suggesting the importance of slope gradient and the deposit grain size, permeability and thickness. We updated the Merapi hazard assessment using Pleiades images as the 2010 eruption changed the summit crater morphology and valley channels. Potential sites favorable to future lahar overbank were identified by computing three morphometric parameters of the river channels.