DETECTING AND MAPPING INVASIVE ALIEN WATTLE IN KWAZULU-NATAL, SOUTH AFRICA

摘要

Invasive alien wattle disturbs ecological and economic sectors in South Africa, which may lead to losses in biodiversity and ecosystem services. Remote sensing techniques utilized for early detection and mapping of invasive wattle are pivotal in the South African context, as the relevant decision-making processes need to be taken to eradicate and reduce invasion. In this study, we integrated image texture combinations computed from a SPOT-6 image with sparse partial least squares discriminant analysis (SPLS-DA) to detect invasive alien wattle and surrounding land cover classes. From the results, the texture combination model (OA = 74%; kappa statistic = 70) outcompeted the single band image texture model (OA = 68%; kappa statistic = 65) and vegetation indices (OA = 62%; kappa statistic = 59). The most significant texture parameters selected by the SPLS-DA model were correlation, second moment and homogeneity, which were predominantly computed from the red and N1R bands. The 5×5 moving window was the most frequently selected window for detecting and mapping invasive alien wattle. Overall, this study confirms the ability of image texture combinations integrated with SPLS-DA to detect and map the spatial distribution of invasive alien wattle.