An efficient method for mapping flood extent in a coastal floodplain using Landsat TM and DEM data

Abstract

An efficient and economical method for mapping flooding extent in a coastal floodplain is described. This method was based on the reflectance features of water versus non-water targets on a pair of Landsat 7 Thematic Mapper (TM) images (before and during the flood event), as well as modelling inundation using Digital Elevation Model (DEM) data. Using limited ground observation, most flooded and non-flooded areas derived from this analysis were verified. Utilizing only TM data, the total flooded areas in Pitt County. North Carolina on 30 September 1999 was 237.9 km2 or 14.0% of the total county area. This number could be low due to the underestimation of the flooded areas beneath dense vegetation canopies. To further investigate this underestimation, a subset of the area covering the four central topographic quadrangles, the Greenville area, in Pitt County was selected. Through addition of the DEM data into the flood mapping analysis of the Greenville area revealed that the total flooded area was 98.6 km2 (out of a study area of 593.9 km2) or 16.5%. In the Greenville study area, the three landuse and landcover categories most affected by the flood were bottomland forest/hardwood swamps (32.7 km2), southern yellow pine (28.8 km2), and cultivated land (19.1 km2). Their total flooded areas were 80.6 km2 or 81.7% of the total flooded area within this study area. The DEM data helped greatly in identifying the flooding that occurred underneath forest canopies, especially within bottomland forest and hardwood swamps. The method was reliable and could be applied quickly in other coastal floodplain regions using data that are relatively easy to obtain and analyse, and at a reasonable cost. This method should also work well in areas of large spatial extent where topography is relative flat.