Knowledge Base Classification of Wetlands from Coarse Resolution Satellite Data

  • Agarwal R. Department of Geography, University of IOWA, IA-52246, USA
  • Garg J. K. University School of Environment Management, GGS IP University, Delhi 100006, India


Mapping of wetlands is carried out using remotely sensed data globally from various sensors data and techniques. Present paper describes a knowledge-based classifier to identify the wetlands in Gujarat state using Moderate Resolution Imaging Spectroradiometer (MODIS) sensor data. Spectral knowledge of BLUE, RED and NIR bands in conjunction with Land Surface Temperature (LST) and Normalized Difference Vegetation Index (NDVI) have been used to define the thresholding ranges for various wetland classes. A comparative study of the performance of this classifier in comparison to two traditional classifiers namely Minimum distance to mean, and Maximum Likelihood classifiers was carried out for a small area. Results of the classification indicates better performance of knowledge based classifier in terms of overall accuracy as well as kappa statistic which are 84% and 0.79 respectively. The results indicate the feasibility of the development of a generalized knowledge based classifier for automated extraction of wetlands at regional scale using coarse spatial resolution data.