A comparison of the accuracy of Nakhon Si Thammarat’s forest area classification methods; Normalized Difference Vegetation Index (NDVI) and Supervised Classification of LANDSAT 5 satellite data using geographic information system technique

Abstract

Geographic information system (GIS) is a widely technology used in the study on the earth exploration planning and creating mapping using satellite images data for national development planning, land use changes and agricultural area planning, natural resources and environmental management and applying for the analysis of spatial data i.e.; forest area, water resources, water qualities, and agricultural areas, etc. The objective of this study is to compare two analysis methods; Normalized Difference Vegetation Index (NDVI) and Supervised Classification method in the classification of the forest areas in Nakhon Si Thammarat Province. The dataset of Landsat 5 satellite images in 2012, 2007 and 2002 were analyzed by Arc-GIS 9.2 geographic information program. As the NDVI analysis method was classified of the area based on the difference of surface reflection between the near-infrared and wavelengths of the red color in proportional to sum of those two wavebands to adjust to the normalized distribution. Supervised classification method was a maximum likelihood classifier method. The training sample area from database maps of Land Development Department were used as referenced areas (Training Area) to identify the characteristics of the sample data. The classification of forest areas was divided into 3 categories: rain forest, mangrove forest and swamp forest. The accuracy of analysis method was rechecked by overlay analysis and statistical value of intercultural principles with land use information in 2012, 2007 and 2002 of Land Development Department. The results showed that supervised classification method had higher classification accurate than NDVI classification method. The accuracies of forest area classification of supervised classification method and NDVI method were 99.97, 99.78 and 99.78 and 49.43 73.84, and 97.98 percentages, respectively. Because, In Supervised Classification method, use the reference data for land use from the Land Development Department as a group based on color values to classification forest areas. The reason that the accuracy of the NDVI method in 2012 and 2007 was less than 2002 due to the similar NDVI values according to the density of forest areas. Thus, resulting in the classification areas error in which the NDVI value of the forest is between 0.40-1. The analysis of NDVI method can reduce the time to analyze the data is, but the analysis of Supervised Classification method high accuracy because the information is processed and verified by the government agencies in processing.