Chapter 4 - A Reflection on Image Classifications for Forest Ecology Management: Towards Landscape Mapping and Monitoring

Abstract

Different remote sensing techniques act as an alternative to traditional fieldwork; thereby, providing an efficient and time-saving approach to extract land use land cover (LULC) information, especially in isolated and inaccessible regions. They serve as a vital source for generating information for land resource managers and forest ecosystem conservationists. However, the major issues that determine the accuracy of such spatially explicit information are the types of data used and the choice of an appropriate classification algorithm. There's a pre-emptive need to improve algorithm-benchmarking with consensus on both LULC definitions and reference maps for their effectiveness aimed at conservation and management efforts. In this chapter, we elaborate on the classification methods available for information extraction from satellite data. As an effort of this chapter, we explore the accuracy of different classification methods in a complex heterogeneous terrain of the central Himalayan region in Asia. We explore both parametric and non-parametric classifiers to produce maps with major forest types in the study region. We recommend more directed research efforts required to reduce uncertainties of mapping methodologies, while quantifying factors affecting forest classification, especially with the integration of ancillary information and multi-source data to improve classification accuracy in a remote sensing context.