Looking for green grass in the desert: methods for land-cover classification in drylands


In recent years, applications of Earth Observation for archaeology have been boosted by data acquisition and by the increased spatial and temporal resolution of new products (e.g. Sentinel-2, WorldView series, Pléiades mission). Nowadays, archaeologists are looking for ways to effectively merge multi-spatial and multi-temporal imagery, integrating spectral and contextual information as well. In arid lands, the lack of adequate data on long-term vegetation dynamics is hampering our capacity of understanding human-environment interactions.

In this paper we present an ongoing research aimed at refine the land cover classification of the archaeological landscapes of Central Sahara and Gujarat (South Asia). Those areas where not subjected to major environmental changes throughout the late Holocene, and thus represent ideal case studies to apply remotely sensed based approaches. The understanding of the land cover dynamics in arid lands is key to refine our reconstruction of past cultural trajectories. In doing this, we are exploring a set of different approaches in land-cover classification (e.g. spectral ratios, LDA, image segmentation and OBIA) in open-source platforms (i.e. R, OTB) in order to get automatic and accurate results from multiple datasets.

Cite this Record

Looking for green grass in the desert: methods for land-cover classification in drylands. Francesc C. Conesa, Agustin Lobo, Stefano Biagetti, Marco Madella. Presented at The 81st Annual Meeting of the Society for American Archaeology, Vancouver, British Columbia. 2017 ( tDAR id: 431596)

Record Identifiers

Abstract Id(s): 15977