Developing Land Use Land Cover Maps for the Lower Mekong Basin to Aid Hydrologic Modeling and Basin Planning

Abstract

This paper discusses research methodology to develop Land Use Land Cover (LULC) maps for the Lower Mekong Basin (LMB) for basin planning, using both MODIS and Landsat satellite data. The 2010 MODIS MOD09 and MYD09 8-day reflectance data was processed into monthly NDVI maps with the Time Series Product Tool software package and then used to classify regionally common forest and agricultural LULC types. Dry season circa 2010 Landsat top of atmosphere reflectance mosaics were classified to map locally common LULC types. Unsupervised ISODATA clustering was used to derive most LULC classifications. MODIS and Landsat classifications were combined with GIS methods to derive final 250-m LULC maps for Sub-basins (SBs) 1–8 of the LMB. The SB 7 LULC map with 14 classes was assessed for accuracy. This assessment compared random locations for sampled types on the SB 7 LULC map to geospatial reference data such as Landsat RGBs, MODIS NDVI phenologic profiles, high resolution satellite data, and Mekong River Commission data (e.g., crop calendars). The SB 7 LULC map showed an overall agreement to reference data of ~81%. By grouping three deciduous forest classes into one, the overall agreement improved to ~87%. The project enabled updated regional LULC maps that included more detailed agriculture LULC types. LULC maps were supplied to project partners to improve use of Soil and Water Assessment Tool for modeling hydrology and water use, plus enhance LMB water and disaster management in a region vulnerable to flooding, droughts, and anthropogenic change as part of basin planning and assessment.

Mapping Land Use Land Cover Change in the Lower Mekong Basin From 1997 to 2010

“…Acquired from the MRC as a vector data layer, the 1997 LULC map was derived from image interpretation, delineation, and digitization of LULC patches from 1:250,000 scale hard copies of Landsat Thematic Mapper imagery (Malyvanh and Feldkotter, 1999). The 2010 LULC map dataset was obtained from a previous project discussed by Spruce et al (2018) that used a combination of ∼quarter kilometer resolution 2010 MODIS 8-day reflectance data from the LP-DAAC web site 1 and single date circa 2010 dry season Landsat Thematic Mapper reflectance data at 30 m resolution from the USGS GLOVIS web site 2 . For this project, the 1997 and 2010 LULC maps were nominally gridded to a ∼231.66-m resolution.…”

“…In addition, a vector-based grid map of each pixel's boundaries was superimposed on remote sensing imagery products. Each sample was image interpreted at a 1:6,000 scale which enables viewing individual applicable Landsat pixels of 30 m resolution (∼8 × 8 or 64 pixels) located within a Example field photos of these LULC classes can be found in Spruce et al (2018). FIGURE 4 | 1997-2010 LULC change map for study area with SB boundaries overlain in cyan for SBs 1-8.…”

“…We think it is in part related to the remotely sensed data and mapping method used to derive the 1997 LULC map. The forest deciduousness levels for the 1997 LULC map did not include Landsat image data collected at monthly intervals across the calendar year, while the 2010 LULC map used the monthly MODIS NDVI data to help classify forest deciduousness according to vegetation greenness phenology (Spruce et al, 2018). Given that the 2010 map used higher temporal resolution MODIS data to classify forest deciduousness, we think that the 2010 LULC map probably shows forest deciduousness classes more realistically compared to the 1997 map dataset.…”

“…Given that the 2010 map used higher temporal resolution MODIS data to classify forest deciduousness, we think that the 2010 LULC map probably shows forest deciduousness classes more realistically compared to the 1997 map dataset. The production of the 2010 LULC map leveraged the vegetation greenness phenology to classify forest types based on deciduousness and multiple agricultural types with comparatively unique phenology (Spruce et al, 2018). Other potential factors include conversion to agriculture or transition to deciduous forest as a result of timber harvesting according to Rundel (2009).…”

“…The initial MRC SWAT model employed a LULC map that was based on 1997 Landsat data (Dat, 2013). More recently, MRC SWAT models were updated with a 2010 LULC map for the LMB that was developed using a combination of monthly MODIS NDVI and Landsat data (Spruce et al, 2018). The updated map included more detailed agricultural LULC classes in order to revise and improve SWAT hydrologic modeling products and applications (Mohammed et al, 2018a,b).…”