To the Editor

Pokhrel et al.1 attempt to estimate sea-level change caused by changes in terrestrial water storage, using an integrated water resources assessment model. I suggest that they have substantially overestimated the contribution of terrestrial water storage to sea-level change, because their model assumptions are unrealistic.

First, the model used by Pokhrel et al. does not actually simulate groundwater flow processes. Instead, all groundwater use is assumed to be unsustainable and therefore to equate to groundwater depletion. This assumption contradicts the principle that well withdrawals from an aquifer are partly or even mostly balanced by increased recharge and decreased discharge, induced by lowered water levels in the aquifer, and do not correspond to decreased storage2,3. Furthermore, the fraction of pumpage derived from storage decreases with time2,3. For the US, with its broad spectrum of aquifer types, boundary conditions and types of water uses, total groundwater withdrawals amounted to 5,340 km3 over the period 1950 to 20054. Cumulative groundwater depletion for that period was about 823 km3 (ref. 5), only 15% of the total extractions. If a percentage of 15–20% reflects average global conditions, then the assumption by Pokhrel et al. that all pumpage is equivalent to depletion could lead to an overestimation of global depletion by a factor of 5 to 7.

Secondly, Pokhrel et al. derive a cumulative total of 18,000 km3 of groundwater depletion from 1950–2000. This is grossly out of bounds with what has been observed. The largest, and best documented case of depletion in one aquifer system is the High Plains (Ogallala) aquifer in the central US, where the cumulative groundwater depletion amounted to 243 km3, or 28% of extractions, for 1950 to 20006. Another 50 to 70 similarly depleted systems simply do not exist throughout the world.

Finally, their consideration of reservoir seepage to groundwater is also flawed. They assume that seepage losses directly equate to a volumetric increase in terrestrial storage, using a model of total water seepage increasing indefinitely as √t_. This model works if the aquifer is infinite in areal extent and the rising groundwater levels do not encounter any interfering boundary conditions7. Both assumptions are, however, unrealistic: aquifers are not infinite, and interfering boundaries, such as a river valley below the dam, are almost always close by. Pokhrel et al. estimate that reservoir storage attributed to seepage is about 50% of reservoir capacity. One of the few studies documenting reservoir bank storage indicates that increased groundwater storage adjacent to the Hungry Horse Reservoir, Montana, represents only about 7% of the reservoir capacity8. If this difference of a factor of 7 holds more generally, the computed effect of reservoir storage on sea-level change would be much smaller. Storage increases in aquifers adjacent to surface reservoirs are mostly local in nature9 and will often stabilize within a few months or years following the filling of the reservoir. After a reservoir starts to fill, adjacent groundwater levels rise and the seepage into the aquifer will become increasingly balanced by increases in groundwater discharge, which limits increases in groundwater storage as the groundwater flow system attains a new equilibrium condition. In many areas, groundwater seepage into certain parts of a reservoir will be large relative to seepage losses out of other parts10.

I conclude that Pokhrel et al.1 have substantially overestimated storage changes from groundwater depletion and reservoir seepage losses, and their model is not valid for these purposes.