[A-CG36] Satellite Earth Environment Observation

[ACG36-P22] Development of high-spatial-resolution global surface water map (GSMaWS Version4)

*Shinta Seto¹, Dai Yamazaki², Takuji Kubota³, Kosuke Yamamoto³ (1.Nagasaki University, 2.University of Tokyo, 3.JAXA)

Keywords:surface water, microwave radiometer, flood inundation

Global Satellite Mapping of Wet Surface (GSMaWS) is a product to show spatiotemporal variation of land surface water based on the brightness temperature data observed by microwave radiometers. GSMaWS Version 1 had been produced by using microwave radiometers GMI and AMSR2 globally for the period between 2013 and 2017 with the spatial resolution of 0.1 degrees (Seto and Mine 2018). Later, SSMIS sensors were additionally used and the period was extended (Yamamoto et al. 2024). Normalized Differential Flood Index (NDFI) is calculated from brightness temperatures at 18.7 GHz of vertical polarization (TB18.7V) and at 23.8 GHz of vertical polarization (TB23.8V) as NDFI = (TB23.8V - TB18.7V) / (TB23.8V + TB18.7V). Water cover ratio was calculated by NDFI divided by 0.06. GSMaWS Version 2 was produced for the period between 2015 and 2019 over Japan with the spatial resolution of 1.5 arc seconds (Seto 2020). To realize such high spatial resolution, land cover data and expected flood inundation area data of Digital National Land Information are used to assign water cover possibility level to each 1.5 arc second grid. For example, level 0 is assigned to ocean, level 1 is assigned to river, level 2 is assigned to paddy field, and level 8 is assigned to the other land cover types outside the expected flood inundation area. Water cover probability is given to each grid inside a footprint according to the water cover possibility level so that the average of water cover probability becomes the same with water cover ratio calculated by NDFI at the footprint. GSMaWS Version 3 was produced for the period between 2015 and 2019 over the area including Thailand (5-25 N, 90-110 E) with the spatial resolution of 30 arc seconds. Considering data availability to produce water cover possibility level, the spatial resolution is coarser than that in Version 2.
In this study, GSMaWS Version 4 is produced globally with the spatial resolution of 15 arc seconds. As land cover data, GLCNMO (Kobayashi et al. 2017) is used. Occurrence of surface water data in Global Surface Water (Pekel et al. 2017) and flood inundation frequency calculated by Zhou et al. (2021) by means of CaMa-Flood are used. The former (GSW) may not detect surface water under dense vegetation as visible and infrared sensors are mainly used. The latter (CMF) may not detect inland inundation or other types of surface water as CaMa-Flood is a simulation of flood inundation. For a 15 arc second grid, larger value of GSW and CMF are taken (the value is denoted by F). Water cover possibility level 0 is assigned if the land cover type is water body or if F is equal to 100 %. For the other land cover types, the level is determined according to the value of F. Level 11 is assigned if F is equal to 0 %. Next, the relation between NDFI and water cover ratio is prepared. As NDFI is affected by land surface temperature and precipitation, precipitation rate data is taken from GSMaP Version 7 and land surface temperature data is taken from Today’s Earth. Excluding the case that the precipitation rate is higher than 0.1 mm/h, for each degree C bin of land surface temperature, month, sensor and orbit type (ascending/descending), NDFI and the ratio of level-0 grids in the footprint are analyzed and the relation is used to convert from NDFI to water cover ratio. Then, water cover probability is determined at 15 arc second grid according to water cover possibility level. The products have been made by AMSR2, GMI, and SSMIS (F16 and F18) for the year of 2018 (Figure shows water cover probability around the downstream of Amazon river on July 1, 2018).