ISLAND HABITAT MAPPING UAE
GSD sourced and pre-processed very high resolution spatial data (2 m x 2 m pixel sizes) imagery for this detailed habitat mapping project. After atmospherically correcting the imagery using python scripting and open source libraries, training data was generated in close consultation with the client and was used to generate habitat models, which were then discussed iteratively with the client.
There were mangroves on site which were of high conservation value and as the client was developing a residential development with a focus on maintaining the in situ biodiversity, mapping the extent and condition of these was of significant interest to the client.
GSD’s work expanded to a focussed study, on the mangrove habitat, using spectral indices to explore the condition of the mangrove and its resilience on site. The aim of this second piece of work was to use satellite imagery to assess whether there is a relationship between mangrove health, using NDVI (a spectral index called Normalised Difference Vegetation Index) as a proxy, and percentage of time mangrove areas are inundated. The expected relationship tested was that healthy mangrove displayed higher NDVI values where inundation times were relatively high. This was investigated using data on a pixel by pixel basis.
Mangrove areas from the Jubail habitat map were used to extract the NDVI values of only the mangroves. The areas were converted to points and at each of the point locations the NDVI and the inundation value from the inundation dataset was extracted. Areas of mangroves that were smaller than 300 m2 were removed from the analysis; since they are most likely to have an influence from the soil reflectance. The resulting areas were converted to points, inundation and NDVI values were extracted, values were plotted and statistics were produced.
These larger areas were also analysed by extracting the minimum, maximum, median, modal, and mean NDVI and inundation values for each area rather than using each point within the area. This assessed correlation between the mangrove health and inundation time on a larger scale, thereby reducing influences of soil reflectance and the difference with the spatial resolution of the inundation model.