IMPROVING ACCURACY OF GROUNDWATER STORAGE IN THAILAND USING GRACE DATA ASSIMILATION TECHNIQUE
Groundwater is critical for supporting ecosystems and facilitating human adaptation to climate change, which may exacerbate water accessibility and food security issues. Accurate groundwater data is thus essential for improving the reliability of the country’s strategic plan. However, obtaining groundwater data from ground measurements is difficult due to the scarcity of sampling sites, despite providing accurate groundwater information. In some provinces, for instance, there may be only a few available measurement sites. As a result, groundwater information is frequently obtained through a hydrology or land surface model. A strong point of models is their ability to generate spatially distributed estimates and comprehensive environmental variables. The model’s limitation is the significant uncertainty surrounding, for example, inaccurate parameter calibrations and underrepresented model physics, and relying solely on model simulation may result in faulty or invalid predictions.
Groundwater data can also be estimated by measuring the Earth’s regional gravity fields over time. These have been observed by the Gravity Recovery and Climate Experiment (GRACE) and GRACE Follow-On (GRACE-FO) satellite missions since 2002 (Tapley et al., 2004). GRACE-FO is a continuous gravity mission to its successor (GRACE), and this study refers to both missions as GRACE. GRACE is a twin satellite tracking system that flies in a near-polar orbit 500 km above the Earth’s surface. The K-band ranging system is used to calculate the range (or range rate) deviation between the twin satellites as a result of changes in Earth’s gravity. This range measurement is used to derive variations in terrestrial water storage (TWS), which is the sum of soil moisture, groundwater, canopy, snow, and surface water storage. Satellite gravimetry is distinct from other remote sensing techniques in that it can detect total column of mass variations (including groundwater), whereas other techniques are only sensitive to a few centimeters of depth. The disadvantage of GRACE is its limited spatiotemporal resolution, which only provides monthly-average data at a footprint of approximately 100,000 km2. This limits GRACE applications to a large area, while direct use of GRACE data is unlikely to benefit local farmland or urban planning.
This study employs the data assimilation (DA) technique to combine the strengths of model simulation and GRACE data. In DA, the model states are statistically adjusted using satellite observations, taking into account the uncertainties in the model states and observations. Despite the success of GRACE DA techniques, they have never been used in Thailand. The purpose of this paper is to demonstrate the GRACE DA application in improving groundwater information in Thailand. The development is demonstrated in the northern part of the Ping River Basin, where in situ groundwater measurements are available to evaluate the results (Fig. 1). GRACE data are assimilated into the Community Atmosphere–Biosphere Land Exchange (CABLE) land surface model (LSM) to improve the groundwater storage (GWS) at a resolution of approximately 5×5 km2. The analysis from this study highlights the relevance of the GRACE DA technique in assisting a groundwater monitoring system in Thailand.
Figure 1. The northern part of Ping River Basin in Thailand and geolocations of in situ groundwater sites. The inset shows the geolocation of the study area (black rectangle).
The GRACE-derived TWS variations are assimilated into the CABLE model using the 3-dimensional ensemble Kalman smoother (EnKS 3D). The EnKS 3D accounts for spatial correlations in model and observation errors, given that the latter are highly correlated at neighboring 0.05°×0.05° grid cells (e.g., within 3°×3° observation space). This section provides a high-level overview of the EnKS 3D concept, while more in-depth information can be found in Tangdamrongsub et al. (2021). The EnKS 3D comprises forecast, analysis, and update distribution steps (Fig. 2). The forecast step propagates the ensemble model states forward in time. The analysis step calculates the monthly model state correction using GRACE observations (and uncertainties). The final step reinitializes the initial states and repeats the forecast step with the correction distributed across the month.
Figure 2. Data processing diagram of GRACE data assimilation
The evaluation is carried out by comparing the correlation (R) of GRACE DA GWS results (with respect to in situ data) with the correlation value of the model estimates. Specifically, the difference is computed by R(GRACE DA) – R(model), where positive and negative values reflect GRACE DA improving and degrading GWS estimates, respectively. Figure 3 shows that GRACE DA provides a significant improvement in GWS estimations, with higher correlation values of up to 0.53, or by 0.3 on average (Fig. 3a). It is of particular note that this improvement is substantial in light of the fact that the model estimate is already quite accurate. The positive impact of GRACE DA on GWS is consistent with the most of GRACE DA research. The 2019 – 2020 drought characteristic is found in both GRACE DA GWS estimations and in situ groundwater level (Fig. 3b), similar to TWS, indicating the sensitivity of the groundwater component to droughts. Model simulations are unable to depict the declining GWS.
Figure 3. (a) Differences in correlation between GRACE DA and model estimations (DA minus model). Positive (red) and negative (blue) readings imply that GRACE DA improves or degrades GWS estimates, respectively. The given values are the correlation differences at the measurement sites. (b) The average GWS estimates from model and GRACE DA between 2004 and 2020. The time series from all sites are used in the averaging. The average in situ groundwater level (H) is also shown.
The success of using satellite data assimilation to enhance the accuracy of regional groundwater storage in Thailand is demonstrated in this study. The method improves hydrologic variables at the very high spatiotemporal resolution, and spatially/temporally continuous fields can always be produced even where/when satellite observations are absent. These accurate groundwater products may improve the robustness of water resources and climate-related decision-making, including (but not limited to) agriculture and urban planning. Future work aims to improve product resolution and incorporate multiple satellite data into a multivariate data assimilation framework to produce a suite of accurate hydrologic and climate variables for interdisciplinary studies.
Full paper can be downloaded here.
RELATED DATA
ALICE-LAB: Asian Land Information for Climate and Environmental Research Laboratory
Asian Hydrometeorological Simulation (AHS)
Access high-resolution monthly hydrometeorological data at 1 km scale for Asia, empowering advancements in water resources management, climate research, and environmental decision-making. Unlock precise insights for optimized conservation strategies, refined hydrological forecasts, and effective climate change mitigation.
Thailand Hydrometeorological Simulation (THS)
ALICE-LAB provides daily high-resolution hydrometeorological data for Thailand at a 1 km scale, enhancing water resource management and climate research. This detailed dataset refines hydrological forecasts, supports conservation strategies, and offers critical insights for effective climate change mitigation, driving better decision-making and operational efficiency.