Spatio-temporal Distribution Mapping of Invasive Weed Lantana camara Using Satellite Imageries in Chitwan Annapurna Landscape, Nepal

Abstract

One of the major environmental concerns in Nepal is spread of the invasive alien plants and its threat to biodiversity. The detection of invasive alien plant species (IAPS) at landscape level can aid in monitoring and managing their invasion in ecosystem. Remote sensing has been an important tool for large scale ecological studies of IAPS. In the present study knowledge-based classification using Landsat images was employed to determine the distribution of Lantana camara in Chitwan Annapurna Landscape (CHAL), Nepal. Maximum Likelihood Classification (MLC) technique was used for the extraction of land use/land cover types from remotely sensed data. Variables like elevation, aspect, slope, land use/land cover, temperature, rainfall and normalized difference vegetation index (NDVI) were used for the knowledge based classification approach. For the comparison of satellite data, World View-2 (WV2) of fine spatial resolution (2x2) m and Landsat of coarse spatial resolutions (30x30)m multispectral data of same area of interest were used. The results using Landsat image showed that weed covered 0.24, 0.9, 1.45 and 2.74 % area of CHAL in the year 1992, 2000, 2009 and 2018, respectively. The cover of the weed estimated using Landsat images was comparatively higher than the cover obtained from the World view-2 images. After evaluating all the available results the knowledge-based algorithm using Landsat produced very promising results, with >77% overall accuracy and a Kappa index of 0.54 in CHAL. The overall accuracy varied between 78 and 83% and Kappa indices of 0.56 and 0.66 for Landsat; the highest overall accuracy was achieved in Makwanpur district where as the lowest was achieved in Kaski district. The overall accuracy varied between 81 and 88% and Kappa indices of 0.62 and 0.76 for the WV2. The highest overall accuracy was achieved in Nawalparasi district where as the lowest was achieved in Kaski district. When compared, the accuracy was higher in the WV2 image than in the Landsat image. The largest area of distribution was found in Middle Mountain followed by Siwalik and High Mountains. The methods adopted in this study can be used for testing other types of satellite data or other classification algorithms. This investigation revealed the strength of mapping shrub weed using Landsat images which is freely available in the archives. Keywords: Supervised classification, Landsat imageries, World view 2, Digital number, Confusion matrix, Kappa index