Abstract:
Monitoring coastal ecosystem resilience for climatic and/or anthropogenic vulnerabilities is
challenging with moderately resolution Landsat images. A simple, low-cost Kite Aerial Photograph
platform (KAP) was vital to obtain high-resolution images for a small area to develop coastal GIS models.
This study examines post-tsunami relief in two coastal shrub ecosystem and a mangrove ecosystem in
terms of vegetation bioshield mass and sea level rise perspectives. A KAP platform was created using two
light-weight automatic cameras with dual bandpass Red-NIR filters, a Picavet stabilizing rig, a GPS
tracker and a Parafoil Kite. The KAP images were processed to build mosaic images, orthorectified and
geo-referenced Digital Elevation Model (DEM) using structure-from-motion (SFM) and remote sensing
software (Agisoft PhotoScan and ENVI respectively). KAP has been utilised for coastal mapping under
three scenarios: (i) object-orient feature extraction for discriminate Prosopis juliflora, an invasive alien
species, and texture analysis for coastal shrub and herbaceous vegetation classification (ii) DEM for sea
level rise, and (iii) Normalized Difference Vegetation Index (NDVI) for mangrove bioshield mass
estimation. The image processing produced a point cloud with an average density of 35 points/m2; a DEM
with 17 cm resolution; and an orthophoto mosaic with an average resolution of 4.0 cm. The results
showed that object orient feature extraction can discriminate Prosopis juliflora from the coastal shrubs
with 62% accuracy, while supervised classification accuracy was 51%. Mangrove vegetation in Rekawa
was discriminated from grassland and other coastal shrub vegetation types at ≥4 NDVI threshold resulted
in 0.33 ha of mangroves (28% of 1.15 ha of the total area). The Kahandamodara beach coastal vegetation
was dominant by Ipomoea pes-capre with 26% coverage. In conclusion, KAP has a wide potential to
bridge science with high spatial/temporal resolution in-situ data for coastal habitat mapping, where the
researchers can utilize the data within a low-cost budget.
