MULTI-TEMPORAL EARTH OBSERVATION TO MAP VEGETATION DYNAMICS IN KSA
GSD was tasked with providing remotely sensed analysis for a very large region of Saudi Arabia for this confidential client. The goals of this project were to provide historical vegetation analysis as well as an assessment of current land cover patterns.
This project contained a variety of challenges, from identifying vegetation cover in desert conditions, to mapping them coherently through time. To achieve these goals, the entire Landsat satellite archive was used, spanning 30 years of data collection.
To analyse this very large dataset, annual composites of the region were created to ensure cloud free images with minimal aerosol presence. From these, a variety of optical indices and band data were analysed in a pixel stack, allowing highlights in trend changes in vegetation productivity and extent to be extracted. This was then modelled to identify hotspots of vegetation change, high productive areas and trends were identified in order to predict future change.
From this analysis, both map-based and spatial dataset products were created that condensed the large amount of processed data into interpretable products with respect to planning and conservation applications.
To compliment this, a modern-day land cover map was also created using higher resolution Sentinel-2 data. Works here focused on the identification of ‘speculative’ land covers, with no training data available to train a specific model. Here, methods were developed to identify distinct terrestrial and marine extents using the spectral data, as well as elevation and other radar-derived products.