Abstract for presentation at The 13th Australasian Remote Sensing and Photogrammetry Conference

Landslide prediction using vegetation index, surface temperature and surface moisture content derived from Landsat data

  • Wing Yip Lau, School of Surveying and Spatial Information Systems, The University of New South Wales, Australia
  • Dr Xiuping Jia, School of Information Technology and Electrical Engineering, The University of New South Wales,Australian Force Defense Academy, Australia
  • Dr Linlin Ge, School of Surveying and Spatial Information Systems, The University of New South Wales, Australia

    • Limited studies were conducted to predict the future rainfall induced landslide locations using earth surface temperature anomaly derived from remotely sensed data. This approach might not be reliable when applied to heterogeneous regions and hilly terrain. This is because surface temperature can be affected by terrain elevation. Moreover, different vegetation covers can alter the local surface temperature. As results, using surface temperature measure alone for landslide prediction can be misleading.
    Generally speaking, a rainfall induced landslide happens when the shear stress of hill slope materials exceeding the strength of slope materials. The magnitude of the stress is a function of cohesion, pore pressure, internal friction and slope angle. However, the changes of the first three factors are dependent of soil moisture. For this reason, this study proposed to use moisture content together with surface temperature derived from Landsat TM images for rainfall induced landslide prediction. Elevation and Vegetation index were also taken into account for removing their effects on surface temperature. A three-layer multistage decision-making structure was developed, whereby the image was sequentially segmented by NDVI (Normalised Difference Vegetation Index), surface temperature and NDWI (Normalised Difference Water Index).
    A testing image acquired on 16th August 1995 was used to demonstrate the inverse relationship between surface temperature and terrain height, and positive relationship between NDVI and surface moisture content. An image acquired over Wollongong, Australia, on 30th July 1998, was used to implement the developed classification scheme. Since the image was acquired 17 days prior to the landslides, therefore, excellent prediction results were not expected. The overall accuracy of the prediction analysis was 33.3%, of which four out of sixteen actual landslides were predicted. Nevertheless, a framework of the improved landslide prediction algorithm was illustrated in details and the experiments indicated a good potential for further exploration.

    Conference Organiser - ICMS Pty Ltd