Field spectrometry of papyrus vegetation (Cyperus papyrus L.) in swamp wetlands of St Lucia, South Africa

Abstract

Techniques for discriminating species in swamp wetlands are critical for rapid wetlands assessment and proactive management. In this study, we tested whether field spectrometry could discriminate between papyrus swamp and its co-existence species (Phragmites australis, Echinochloa pyramidalis, and Thelypteris interrupta). Canopy spectral measurements were taken from the species using Analytical Spectral Devices but later resampled to Hyperspectral Mapper (HYMAP) resolution. The random forest algorithm and a forward variable selection technique were used to identify key wavelengths for discriminating the species. This method yielded ten wavelengths (1409 nm, 710 nm, 437 nm, 464 nm, 452 nm, 1424 nm, 725 nm, 480 nm, 587 nm, and 603 nm) located in the visible and SWIR portions of the electromagnetic spectrum with lowest out-of-bag estimate error rate of 9.5 %. The use of random forest as a classification algorithm resulted in overall accuracy of 90.5 % and a KHAT value of 0.87 for all class pairs (n =3D 6), with individual class accuracies ranging from 93. 73 % to 100 %. The study also demonstrated the possibility to scale up the method to airborne sensors such as HYMAP for discriminating swamp wetland species.