Updating Regional Land-Use/Land-Cover Maps for Improved Hydrologic Modeling and Water Resource Management in the Lower Mekong Basin

Over 60 million people living in the Lower Mekong Basin (LMB) region depend on the Mekong river for water for many purposes, including water for human consumption, hydro-electric energy, fishing, tourism, transportation, and irrigation for agriculture. However, there are many threats to the human inhabitants of the LMB, including flooding from severe storms, Land Use Land Cover (LULC) change, and hydrologic change (e.g., water flow and in-water sedimentation) associated with increased damming. The Soil and Water Assessment Tool (SWAT) is being used by the Mekong River Commission (MRC) and other institutions and agencies in the LMB region to help aid water resource and disaster management via prediction of surface runoff and stream flow rates. The LMB SWAT models include a LULC map as a primary input.

Prior to this project, the LMB SWAT models employed 1997 LULC data. The latter LULC map was out of date and included inadequate agricultural types than desired. In response, circa 2010 LULC maps were compiled through computer-aided image classifications of MODIS monthly NDVI and dry season Landsat satellite data. Monthly MODIS NDVI data for 2010 was used to map regionally common LULC types with unique vegetation greenness phenology. Circa 2010 dry season Landsat data was used to map LULC types more common at local scales (e.g., urban areas and smaller water bodies). GIS techniques were used to integrate MODIS and Landsat results into 250-meter LULC maps for MRC-defined Sub-Basins (SBs) 1 - 8. The 2010 LULC maps were then validated quantitatively. In particular, random locations on the LULC maps for SB 4 and 7 were compared to various geospatial reference data that included Landsat false color imagery, higher spatial resolution aerial and satellite data, in-situ data from the MRC, and other previously produced LULC maps for the region.

At full scheme specificity, the two tested LULC maps showed more than 80% overall agreement with reference data. Compared to the 1997 map, the 2010 LULC map includes more specific agricultural LULC classes pertaining to crop type and number of crops per year. The agricultural classes in the 2010 LULC maps incorporated crop calendar data supplied by the MRC. Once compiled, the 2010 LULC maps were supplied to project partners to update SWAT models for LMB SBs 1-8. The SWAT modeling products are in turn being used to aid water resource and disaster management and decision-making in the region.

The 2010 LULC map was later compared to the 1997 LULC map to quantify and assess LULC change for each sub-basin being modeled with SWAT, including flood prone riverine areas where towns and agriculture can occur. The LULC change map enabled identification of areas in which forests were converted to agriculture, industrial forest plantations (e.g., rubber), and reservoirs from 1997 to 2010. The LULC change map was also employed to view LULC change across the entire LMB. The 2010 LULC map is also being used in conjunction with SWAT to aid the monitoring and assessment of impacts from extreme weather and climatic events, such as flooding and drought effects on agricultural and urban areas. The 2010 LULC map includes multiple agricultural cover types (e.g., rainfed and irrigated rice according to number of crops per year). This agricultural class specificity in the 2010 LULC map may be useful for monitoring and assessing food security issues. The 2010 LULC map is a more up-to-date and specific base map that could be used in part to derive a more current LULC map. Preliminary assessment of MODIS NDVI-based change maps for 2018 indicate that additional transition of native forest to agriculture and industrial forest plantations has occurred in some portions of the LMB since 2010. However, more work is needed to further update the 2010 LULC maps to be more indicative of current conditions (e.g., in 2018).