Bank condition is an important indicator of physical form for streams that is related to the environmental condition of riparian corridors. Our objective was to develop and test an approach for mapping bank condition from high spatial resolution airborne UltracamD image and LiDAR data in a temperate forested/urban environment. Field observations of bank condition scores were related to UltracamD and LiDAR derived metrics including: valley confinement; bank height; bank width; plant projective cover; grass cover; bank slope; bank top crenulation; stream sinuosity; and channel width and height variability. The highest correlation was achieved between field assessed bank condition scores and image derived grass cover (R2 = 0.43, n = 41), channel width/height ratio (R2 = 0.41, n = 41), average bank slope (R2 = 0.60, n = 41) and valley confinement (producer’s accuracy = 100%, n = 9). Based on these metrics, predictive models were produced to calculate five discrete bank condition scores, indicating: (score = 0) very poor bank condition with distinct erosive features; (score = 1) poor bank condition; (score = 2) average bank condition; (score = 3) good bank condition; and (score = 4) excellent bank condition. The predictive models produced bank condition scores that corresponded to those assessed in the field in 68% of cases. Confined valleys were first mapped in eCognition Developer 8, as these produced bank condition scores of 4 in all cases. Remaining bank areas were mapped using the predictive models and the image derived metrics of grass cover, bank slope, channel width, and channel height to automate the bank condition mapping approach through the use of a developed Cognition Network Language rule set for areas classified as stream banks.


  •  Victoria: Werribee River, Lerderderg River, Parwan Creek, Djerriwarrh Creek, Pyrites Creek, Goodman Creek


  • Kasper Johansen


  • University of Melbourne, Department of Sustainability and Environment

Timeframe: 2009-2010


  • Department of Sustainability and Environment