Application of Hyperspectral Image Data for Species Detection and Biomass Estimation of Submerged Macrophytes in UK Chalk Streams

Expected improvements of spatial and spectral resolution of remote sensing data in the near future will finally enable their application for the monitoring of some of the UK’s most biodiverse ecosystems: lowland chalk streams. The possibility to remotely map cover extent and submergence depth of chalk stream macrophytes could improve biomass estimates and help understanding of macrophyte community dynamics. This study aims to improve Ranunculus (Water crowfoot) submergence depth estimates from a Specim Eagle hyperpsectral image taken of the River Frome in Dorset, UK by combining data from previous studies. NDVI values calculated from Ranunculus vegetation spectra measured with a GER1500 show a depth dependence, which corresponds well with modelled values. NDVI values extracted from the image data do not show a similar relationship. The results highlight the difficulty of obtaining accurate submergence depth information when vegetation cover and submergence depth vary at sub-pixel level.
Data quality issues also hamper image analysis at this level of detail. Blurring/smearing of the data will have affected the NDVI-submergence depth relationship derived from the image. An attempt was made to improve the data quality by estimating an empirical Point Spread Function (PSF) from the bank vegetation-river water interface and trialling different deconvolution algorithms and in put parameters. The application of this method was unsuccessful and specified some of the limitations of a technique that has been successfully applied in other situations.