Using data assimilation to improve land subsidence prediction from a data-driven and physics-based modeling approach: An application to Bangkok, Thailand

Abstract

Accurate and precise subsidence simulation is greatly influenced by limited availability of data, specifically input forcings/drivers and calibration data. In this study, an ensemble-based data-assimilation method is used to improve the estimates of land subsidence in Bangkok, Thailand, which is simulated by a linked data-driven and physics-based modeling approach. The approach models land subsidence caused by groundwater pumping at observation well nests and deals with limited data availability. A data-driven time series analysis method is first utilized to model groundwater heads in aquifers, which then serves as boundary conditions to a one-dimensional land subsidence model.  Simulated land subsidence near an observation well nest is a compaction-based, vertical estimate that assumes elastoplasticity. The assimilation of uncertain head observation data results in an estimate of the probability distributions of various state and parameter values based on the model, data, and their uncertainties. Results consist of an improved land subsidence estimation and uncertainty quantification. In Bangkok, the approach is applied with limited groundwater and subsidence observations and only an estimate of basin-wide pumping. Prior to data assimilation, groundwater and subsidence dynamics are successfully captured at 23 well nest locations. The application of the data assimilation method provides an improved understanding of these dynamics in Bangkok through uncertainty quantification of heads and subsidence as well as related parameters. Ultimately, this study demonstrates the applicability of data assimilation to improve land subsidence estimates when dealing with data scarcity.  Risk assessments of relative sea-level rise of Bangkok and other deltaic regions depend on improving land subsidence estimates and associated uncertainty, which is essential for future flood mitigation efforts.