Detecting and Mapping Snow Induced Damage to Commercial Forest Plantations Using Sparse Partial Least Squares – Discriminant Analysis

Abstract

Snow can bring about devastating consequences to commercial forestry in the KwaZulu-Natal midlands. These consequences may result in serious ecological and economic risk for commercial forestry in the region. With the use of remote sensing methods, the detection and mapping of snow induced damage to vegetation used for commercial forestry in the region can potentially provide useful information on both the spatial extent and severity of damage, which can feed into better informed and sustainable forest management. To our knowledge, no study has explored the ability of remote sensing in detecting and mapping snow damaged vegetation in South Africa. Therefore, this study integrates vegetation indices derived from multi-temporal Landsat 8 imagery with Sparse Partial Least Squares - Discriminant Analysis (SPLS-DA) to detect and map snow induced damage to vegetation used for commercial forestry. The results indicated the ability of SPLS-DA to detect and map snow damage with overall accuracies of 55%, 78%, and 77% achieved for the 01/09/2018, 10/09/2018 and 17/09/2018 dated images, respectively. The results of this study validate that the SPLS-DA model can be successfully used to detect and map snow induced damage. In addition, the use of multi-temporal imagery allowed for the successful validation of snow damage and illustrated the progression of damage overtime.