Mapping Belize´s Coastal Mangrove Forests: experiments with conjunctive use of SPOT, L- and X-band SAR and SRTM data.

Abstract

Four experimental classifications of coastal mangroves in Belize were developed using Definiens software using various combinations of SPOT multispectral data and ALOS L-band PALSAR data combined with digital surface models obtained from airborne X- band SAR or SRTM data. Each classification was compared to reference data polygons digitised from a visual interpretation of IKONOS imagery and oblique aerial photography and confusion matrices were generated. Accuracy was also evaluated visually by comparing the classifications with the extents of three different mangrove classes depicted on the last edition of the National Mangrove Map of Belize. Elevation data from X-band and SRTM SAR was used to assess the reliability of the “medium” and “tall” mangrove classes defined in the pre-existing National Mangrove Map, which had been created before such DEM data was available in Belize. In most cases, vegetation in the two classes was revealed to actually be almost the same height (6-7 m). As a result, only two classes of mangroves, namely ‘dwarf’ (<3 m) and ‘medium-tall’ (>3 m) were classified in our experiments. The highest classification accuracy of 83% was achieved using a combination of SPOT, with elevation supplied by X-band airborne SAR data that was available only within the test area. There was also a reasonable spatial coincidence between the classes produced by this ‘best’ combination and the mangrove areas on the National Mangrove Map (60% medium-tall, 27% dwarf mangrove overestimation). A second classification using SPOT and globally available SRTM data provided a satisfactory result with an accuracy of 74% and a spatial coincidence of 60%, which might be sufficient to develop a classification of mangroves for the entire Belizean coastline.