| 14 Aug 2026
Antecedent groundwater depletion amplifies hydrological drought severity: aquifer-river connectivity and cascading recovery in the 2022 Po Plain extreme event
Abstract. The 2022 Po River drought produced a 78 % reduction in annual discharge at Pontelagoscuro despite a precipitation deficit of only 20–25 %, a discrepancy not attributable to meteorological anomalies alone. Using GRACE-FO groundwater storage anomalies (GWSA), ERA5-Land soil moisture fields, and a 25-year daily discharge record, we show that antecedent aquifer depletion altered aquife-river connectivity, amplifying the hydrological response beyond what meteorological forcing alone would predict.
GWSA had reached −149 mm by August 2021, more than a year before the drought peak. During the 2022 irrigation season, groundwater abstraction substantially exceeded recharge once Alpine-sourced surface water became unavailable, producing the largest single-season storage loss in the 22-year satellite record (−246 mm, Z = −2.32). The recession coefficient k increased by 90.7 % between 2021 and 2022, exceeding the storage-dependent trend by 90.5 % and consistent with a threshold transition in connectivity near GWSA = −150 mm, a level corresponding to roughly the lowest quintile of the 2002–2024 August GWSA record. Direct runoff collapsed by 72 %, leaving baseflow as the dominant hydrograph component.
Recovery proceeded in three tiers: meteorological indices within 8 months, river discharge within 12 months, and soil moisture within 14 months; GWSA had not returned to historically normal levels within the observational record, leaving a period of persistent hydrological drought of at least 7 months undetected by standard precipitation-based indices. Antecedent groundwater depletion thus alters river discharge sensitivity to meteorological forcing, with recovery in groundwater-dependent irrigated basins extending well beyond the meteorological signal. We propose a diagnostic framework based on GWSA, k, and BFI as an early-warning tool for data-sparse irrigated basins.
This study looks at a very important topic – the response of major, groundwater-dependent rivers to drought, in cases where there is pre-existing groundwater depletion within the catchment. The authors have (I believe) demonstrated their case that antecedent groundwater depletion results in far more significant and long-lasting reductions in streamflow than could be predicted using meteological data and indices. I think this paper is worthy of publication and an important contribution to knowledge. However, there is some further work required to strengthen the evidence base for the main findings, and integrate the results with relevant literature and data that has examined similar questions elsewhere.
Specific comments:
Line 29: Typo – correct to aquifer.
Line 81: Wada et al 2010 reference is missing from references list. Famiglietti et al 2014 did not specifically look at the Mediterranean region. You might re-phrase this to indicate that groundwater depletion is a growing global issue, but that there has been limited study (using GRACE or other tools) in the Mediterranean region.
Line 93: This last point may be controversial – does GRACE really remove the need for in-situ observations? I think there is a strong case to say the two must be used in conjunction wherever possible.
Line 113-114: Is this meant to read 2 to 10? Seems like a small typo.
Line 117: If characteristic of the groundwater, it would be good to include one or more hydrographs showing groundwater levels through time (either individual wells, or an aggregated average level). The same applies for the surface water gauging site – would be great to present a typical hydrograph early in the paper for background.
Line 179: Were other possible recession filters applied and assessed for the level of fit? There are many available filters – some discussion of how you chose the Lyne & Hollick filter as being the most appropriate for the data would be helpful.
Line 249: Clarify whether the low-flow deficits are relative to 2021 mean discharge, or another metric (e.g. baseflow during 2021).
Figure 3: Label size is far too small on the figure. Please enlarge the figure label and axis text by at least 50%. Same applies to Figure 4, including axis labels and all other text.
Line 350: It appears the mean baseflow is almost 100% of the mean streamflow (when comparing the mean baseflow values reported here, and the mean 2021 and 2022 total streamflow at line 248). This seems unlikely to be realistic, or at least requires some further discussion and evidence in support. This point underscores the need to carefully assess the most appropriate baseflow/hydrograph separation method and check the realism of the output. Adjusting either the filter method (exploring other options like Boughton, Chapman, Eckhardt), and/or the adopted coefficient(s), and checking whether this significantly affects the mean baseflow estimates, is an important step to ensure estimates are robust and the method appropriate. Is there literature supporting the Po River as being a dominantly baseflow hydrograph?
Line 372 to 388: I’d like to see some more reference to literature on baseflow recession and streamflow hydrograph responses to drought here. You have made some observations and claims about a possible ‘threshold’ change in aquifer-river connection, reflected in the baseflow coefficients. It is important to support this with some relevant literature on the topic – e.g. where else has this been observed and what is the theoretical underpinning? The work of Dr Timothy Peterson and colleagues comes to mind (e.g., Peterson et al. Watersheds may not recover from drought, Science 372, 745-749).
Line 416-417: Can you give some further supporting evidence about the ‘suppressed’ baseflow during this period? Are you basing this conclusion just on the elevated baseflow recession constant, or is there other evidence you can point to in the hydrograph - e.g. comparing flow percentiles during low-flow periods across different years?
Line 426: Your earlier text and Figure 8 suggest that the lag is longer than 7 months – please check for consistency throughout.
Line 440: Is there strong evidence that the antecedent aquifer condition (deficit) was significantly greater in the lead-up to the 2022 drought than previous droughts? This is implied, but no strong evidence is cited here. The GWSA data are an obvious place to look.
Line 444-447: This should be testable, by examining hydrographs from previous years with similar precipitation deficit, but less of a negative GWSA. It is important to attempt this to ensure the conclusion is supported.
Line 478: The California Central Valley comes to mind here – please have a look at the relevant literature and data from this region to see if the same holds true.
Line 496: Is there a particular point on the streamflow hydrograph where you can identify this change specifically, and show the reader?
Line 523-531: Indeed, these are important findings for river management.
Line 537: Is such monitoring sparse within the study area, specifically?
Section 5.5: This section does not really convince the reader that you have identified a specific threshold where baseflow recession behavior has qualitatively shifted to a new state. This would require some more in-depth analysis of the hydrograph to demonstrate the key change point (e.g. change in recession slope or shape that corresponds to a threshold GWSA). You have done this semi-quantitatively by means of looking at recession constants but the work is not very precise. This part doesn’t necessarily have to be part of the scope of the paper – your findings are significant enough regardless, but I think if you are going to make this claim, you need to substantiate it better.
Looking at recent (July-August 2026) SPEI data for the European alpine region, there has been a very significant meteorological drought this summer. Somewhere in your introduction or conclusion, I would suggest mentioning that your research has very important implications for subsequent drought periods, including the current severe drought being experienced in central Europe.