Modeling Groundwater Depletion and Sea Level Rise

By Jacob Wessel

There is general scientific consensus that global mean sea level (GMSL) has risen in the 20th century and at an accelerating rate. This rise is expected to continue throughout the 21st century, with many contributing factors attributable to observed climate change, both natural and anthropogenic. Two primary drivers of sea level rise (thermal expansion of warming seawater and melting of ice sheets and glaciers) account for the bulk of this change, but several other groups of processes also contribute to varying degrees. One such factor is land water storage (LWS), which is generally defined as all forms of terrestrial water excluding land ice, which is modeled separately. Changes in LWS will alter the fluxes of freshwater with the oceans, contributing to a changing GMSL.

Figure 1: Aggregated drivers of sea level change. Highlighted in this research is the land water storage component, specifically the role of groundwater depletion on future global mean sea level.

Among the various components of LWS, groundwater depletion is of particular interest as it continues to be a major issue in water resources for many key basins worldwide. As groundwater is depleted from underground reservoirs, an estimated 80% returns to the ocean, where it will contribute to sea level rise. Other components of LWS directly affected by human activity include reservoir impoundment (trapping water behind dams) and the terrestrial biosphere (altered by deforestation and other land use changes). The contribution of LWS changes to global sea level rise has been studied and quantified in attempts to close the sea level budget over the 20th century. In addition, a few studies have used projections of future groundwater depletion and dam construction to estimate future impacts on sea level. However, these studies are limited, and many papers use the same underlying datasets and simple projections. Estimates of the LWS contribution to future sea level rise given in the IPCC Sixth Assessment Report exhibit no scenario dependence across a wide range of SSP/RCP combinations. As a result, the uncertainty of these estimates is poorly characterized and is limited to propagating parameter/data uncertainty, defining simple fixed confidence intervals, or averaging results from just a few studies. Although the contribution to GMSL of groundwater depletion and LWS in general is thought to be relatively small, there is an incomplete discussion in the literature about the range of possible future outcomes. Adding robustness to the analysis of future groundwater depletion and its uncertainty as it relates to sea level rise can aid decision making and policy planning, and doing so at a global level is a first step to identifying particularly vulnerable regions when it comes to local sea level rise.

My Ph.D. research in the Lamontagne Lab utilizes large-scale modeling of coupled systems and techniques to characterize uncertainty using large ensembles of many model runs. My previous research centered around macro-scale energy systems modeling, specifically the bulk electric power system, and so my work with integrated assessment models tends to favor energy discussions. However, the multi-sector dynamics space that we occupy is inherently interdisciplinary, and engaging with research in hydrology and climate science is a natural extension.

To address the problem, we leverage an ensemble of 900 simulations from the Global Change Analysis Model (GCAM), which handles complex interactions across highly interconnected energy, water, land, climate, and socioeconomic systems through balancing market equilibria. The model runs were constructed as a factorial ensemble which varies six key scenario inputs: socioeconomics (consistent with five SSP’s), climate forcing pathway (consistent with four RCP’s), the underlying global climate model (five GCM’s), three groundwater availability scenarios, two surface water storage scenarios, and two hydrological models used for calibration. From these model runs, we extract basin-level non-renewable groundwater withdrawal data through the year 2100, which considers the evolution of these interconnected systems and the effects of parameters like population growth, global temperature increase, and reservoir expansion. Using these time series, we map the outcome space associated with this set of 900 future pathways and translate these outcomes to changes in GMSL to compare with previously published estimates.

Figure 2: Makeup of scenario ensemble run using the Global Change Analysis Model. In total, 900 model simulations result from a factorial sampling of these six sensitivities. The ensemble excludes implausible SSP/RCP combinations.

To show the effect on GMSL of the groundwater depletion in our model runs, we use cumulative non-renewable groundwater withdrawals from 2020-2100 and assume 80% is transferred to the ocean. Figure 3 shows this groundwater depletion across the ensemble of model runs by continent, which is an aggregation of the 235 individual basins modeled in GCAM. High variability can be seen across the ensemble for several continents, while the inset subplot shows this variability over time on a global scale. We find that the highest groundwater withdrawals are concentrated in Asia and the Middle East.

Figure 3: Cumulative groundwater depletion across all model runs, shown by continent and globally (inset).

Eustatic sea level change results from any net change in the volume of water in the ocean, and to calculate it, one needs only a measurement for the total surface area of the ocean along with the volume added or removed. Here, we assume this surface area to be 361 million km2. The leftmost boxplot in Figure 4 gives the distribution of GMSL rise outcomes in our scenarios, while other boxes show the distribution split by sensitivity. Results show the median GMSL rise due to groundwater depletion by 2100 as 83 (17-205) mm. Background lines and shaded regions give the total LWS contribution to GMSL rise reported in two IPCC reports and its reported confidence interval (given as a uniform range). Note that this also includes effects from surface reservoir impoundment, which can work to lessen the total sea level rise. Figure 4 shows strong agreement with reported values, with a more robust characterization of uncertainty and variability. Additionally, IPCC AR6 gives the full range of GMSL projections for 2100 as 280-1,600 mm, meaning that the relative contribution of LWS changes according to our results could vary greatly and become a more significant driver.

Figure 4: Distribution of GMSL rise by 2100 due to groundwater depletion in our model ensemble, compared to IPCC estimates.

Over the course of this project, I performed a wide literature review of sea level rise and its contributing factors in order to understand both reconstructions of the sea level budget and projections into the future. My primary advisor, Dr. Jonathan Lamontagne (CEE), worked with me to leverage our expertise on large ensemble modeling of coupled human-earth systems in a novel way to address the identified gap in the literature. This modeling adds robustness to the sea level literature through its integrated framework, which combines agricultural, water resources, energy, and human systems dynamics to produce detailed groundwater projections. My co-advisor, Dr. Andrew Kemp (ECS), provided invaluable knowledge in sea level reconstructions to guide the project and bridge the gap between the modeling and application of results. I also connected with several other experts for advice and guidance.

Sea level rise is a complex but pressing issue in regard to our changing climate. However, not articulated here is the geographical diversity in sea level rise and its effects. In addition to GMSL, relative sea level (RSL) describes a more regional or localized response to changing sea level, which can vary widely due to vertical land motion (VLM), gravity, rotation, & deformation (GRD) effects, and ocean dynamics. The next steps of this research are to describe the effects of LWS changes on RSL rise using our ensemble, which will further add to the literature and give planners and policymakers more information with which to steer planning decisions. Figure 5 shows the geographical distribution of groundwater depletion in 2100, normalized by basin area, for the median case as well as the 1st and 99th percentiles of total global groundwater depletion. The concentrations seen in the Middle East, South Asia, the Nile basin, and the Western US could be large enough that gravitational effects become a relevant contributor to RSL, similar to the melting of the ice sheets.

Figure 5: Basin-level groundwater depletion in 2100 for three representative cases: median, 1st percentile, and 99th percentile of total global groundwater depletion scenarios. Future work will explore effects on relative sea level, which varies by location.

Jacob Wessel is a PhD candidate at Tufts University.