Data-driven modeling and analysis of pumping-induced subsidenceĀ 

Overview

One significant cause of over-pumping groundwater is land subsidence. Land subsidence causes permanent groundwater storage loss and can exacerbate the effect of relative sea level rise; therefore, mapping land subsidence in globally groundwater-stressed regions and understanding its key drivers is of great importance. Satellite observations from InSAR and GNSS provide reliable methods to measure subsidence with high accuracy. However, InSAR is computationally challenging and expensive to process, while GNSS records are limited, presenting challenges to map land subsidence at larger scales.

Our research group develops data-driven approaches to map subsidence at global and regional scales by incorporating the limited available InSAR and GNSS-based subsidence observations. Our developed machine learning models utilize various remote sensing and modeled gridded datasets, which represent hydrologic fluxes, land use, and geologic properties, as input variables to map groundwater pumping-induced land subsidence at a high resolution. These models, like the one developed in Hasan et al. (2023), help us explore the global drivers of land subsidence, such as the presence of fine-grained sediments and irrigation activities. Additionally, the models enable us to identify future hotpots of land subsidence by estimating land subsidence probability.

Our group also evaluates empirical relationships between subsidence and various geological and anthropogenic drivers at more local, watershed scales. These include the Parowan Valley of southwest Utah, and the Central Valley of California. While empirical methods provide many advantages, often process-based models are needed to understand complex drivers of subsidence and forecast future subsidence. Our group works extensively with those models as well.


Funding

NGA (National Geospatial Intelligence Agency), NSF GRFP

Study areas

Utah, California, western US, global


Figure Caption:

The figure to the right is taken from Hasan et al. (2023) and shows (a-c) Global and regional land subsidence mapped by the developed machine learning model, (d) analysis of global drivers of land subsidence.


Papers

Hasan, F.*, Smith, R., Vajedian, S.*, Pommerenke, R.*, Majumdar, S.*, 2023, Global land subsidence mapping reveals widespread loss of aquifer storage capacity. Nature Communications, 14 (1), e2022WR034095. https://www.nature.com/articles/s41467-023-41933-z

Smith, R.G., Majumdar, S.*, 2020, Groundwater Storage Loss Associated with Land Subsidence in Western US Mapped Using Machine Learning. Water Resources Research. [link] [preprint]

Smith, R.G., R. Knight, J. Chen, J.A. Reeves, H.A. Zebker, T. Farr, and Z. Liu, 2017, Estimating the permanent loss of groundwater storage in the southern San Joaquin Valley, California. Water Resources Research [link].

Smith, R. G., Hashemi, H., Chen, J., & Knight, R. (2021). Apportioning deformation among depth intervals in an aquifer system using InSAR and head data. Hydrogeology Journal, 29(7), 2475-2486. [link]

Li, J.*, Smith, R., Grote, K. (2023). Analyzing Spatio-Temporal Mechanisms of Land Subsidence in the Parowan Valley, Utah, Hydrogeology Journal. https://link.springer.com/article/10.1007/s10040-022-02583-5