My first remote sensing project in 2017 - LULC in northern Thailand

This is one of the very first remote-sensing projects I completed when I was at Brown in 2017. I think it is worth showing because the study area itself is where many of my relatives live. Every time I visited them, I can see that the area becomes more urban and that there is less green area. The goal of this project was to track land use changes in Chiang Dao District, Chiang Mai, Thailand (~19°N, ~99°E) between 1989 to 2017. I used Landsat 4 Thematic Mapper (TM), Landsat 5 TM, and Landsat 8 TM from USGS in this project.

I became interested in this region because Chiang Dao (pink area on the left figure) is one of the richest wildlife biodiversity forests in Thailand. It hosts two natural nature reserves with a total area of 1644 sq. km. However, this region has been substantially modified for residential and agricultural uses over the past few decades both legally and illegally. According to the local authority, the biodiversity of wildlife in the Chiang Dao forest had decreased from 340 species (1993) to 282 species (2005). Despite this change, no satellite-based land use land cover (LULC) work had been done in Chaing Dao.

The images were selected during the dry season to avoid cloud cover (November-March). I preprocessed using ENVI by correcting for solar irradiance variants, performing atmospheric correction (dark pixel subtraction), removing clouds, and mosaicking images.

I performed a cloud removal process to remote clouds in some images. I calculated for NDVI and applied supervised classification based on local words ad previous knowledge of this area.

Results

The natural combinations show urban expansion (white) over time. We also saw a shift of crop types from high drought-tolerant crops (corn-peanut-ginger-potato fields in orange color) to low drought-tolerant crops (orange tree-garlic fields in white/light green color) around the center of this district. Although it is possible that this crop difference is due to different observed months, we can still see that there is less forest in 2017, compared to 1989.

The left figure shows the NDVI composite (R:2017, G:1989, B:1989). In this case, red means increased vegetation in 2017, relative to 1989. Cyan means decreased vegetation, while white represents similarly dense vegetation. You can see that most areas are in cyan color! The right figure is a supervised classification result for the 2017 image (blue = water, red = urban, forest green = forest, purple = bare soil, yellow = high drought-tolerant crop fields, and chartreuse = low drought-tolerant crop fields).

To sum up, this region has been significantly transformed into agricultural fields and urban areas. The deforestation area is ~272 sq. km by 2017 which is clearly larger than the area that the Gazette of 1982 allowed locals to modify land (5.485 sq. km).