A TECHNIQUE FOR MAPPING MANGROVES WITH LANDSAT TM SATELLITE DATA AND GEOGRAPHIC INFORMATION-SYSTEM

The mangroves in a 2845 km(2) area in the Southern Gulf of Carpentaria , Australia, were mapped from Landsat TM satellite data. The mangroves were mapped by selecting 10 training set areas in dense mangrove (100 % cover), and using the maximum and minimum training set values for gr een, red, near-infrared (NIR) and NIR/red to map the remaining mangrov es. The accuracy of the map was improved by using ecological informati on about mangroves-they are found in tidally inundated areas-to derive simple rules in a Geographic Information System, to subdivide the are as labelled 'mangrove' from image processing of satellite data on the basis of nearness to water (next to water and not adjoining water), gr ound elevation [higher and lower than 10 m above mean sea level (MSL)] and distance from water (> 2 and < 2 km). Each zone was cross-checked with 1:50 000 panchromatic aerial photographs. Zones that were still mixed vegetation after applying these simple rules were further subdiv ided by eye. This process resulted in a map with zones identified as e ither 100% mangrove or 0% mangrove. The areas that were identified as mangrove were also subdivided on the basis of the three main river sys tems in the study area. The Norman, Bynoe and Flinders Rivers had 40.8 6, 10.09 and 5.42 km(2) of mangroves, respectively. These areas combin ed with the 9.89 km(2) of coastal mangrove to give a total of 66.25 km (2) of mangrove in the study area. (C) 1996 Academic Press Limited