Satellites can see Earth’s hidden water move, but what causes those changes? Across six closed basins, this study finds that precipitation tracks the year-to-year GRACE signal far better than evapotranspiration. Seasonal to interannual variations in surface water storage revealed by the GRACE satellites have received less attention than storage trends in the literature.

The analysis focuses on six large endorheic basins and develops variability attribution diagnostics against independent precipitation and evapotranspiration datasets. Endorheic basins, also known as closed or terminal basins, are landforms characterized by their lack of an outflow to the ocean or a major river.

These basins are often found in semi-arid and arid regions, where there is low precipitation and high potential evapotranspiration. Those characteristics make them extremely sensitive to both climate change and anthropogenic activities. GRACE is a satellite mission launched in March 2002 that provides accurate estimates of water storage anomalies from measurements of changes in Earth’s gravity field.

GRACE data have a relatively short time series that is likely to be dominated by natural variability. Earlier work found that GRACE storage anomalies in the Tarim River Basin were consistent with regional precipitation anomalies, with synchronous peak and trough events.

During 2002 to 2016, the global endorheic system experienced a widespread water loss of about one hundred six point three gigatons per year. The study investigates whether GRACE storage variations can be explained by precipitation and evapotranspiration observations according to budget considerations.

The questions include how balanced precipitation minus evapotranspiration is in long-term mean and interannual storage variability, and how well it reproduces seasonal storage variability. The study also asks how well precipitation minus evapotranspiration reproduces interannual storage variability and whether that variability can be attributed to precipitation or evapotranspiration.

Because precipitation and evapotranspiration observations often do not reproduce GRACE storage variations well, the study produces optimized estimates consistent with GRACE on all timescales. The datasets were chosen because they were derived from satellite data, and the study uses a single product for each budget component rather than an ensemble of different products.

The study tests whether the budget closure method can bring these independent products into consistency, without depending on the accuracy of the initial product. Each dataset has monthly resolution and was interpolated at a zero point five degree spatial resolution.

Data for each selected basin were separated using a mask and then spatially averaged. The study uses GRACE data, which provide estimates of total water storage anomaly. GRACE maps Earth’s gravity field every thirty days, using two identical satellites orbiting Earth in tandem and separated by about two hundred twenty kilometers.

The chosen data version is MASCON JPL RL06v2, and the JPL GRACE data cover January 2002 to January 2020. Some months with missing data after 2012 were filled with monthly climatology plus temporal interpolation of monthly storage anomalies. The analysis focuses on six different endorheic basins.

The study analyzes larger contiguous areas that are more accurately resolved by GRACE data, even when some regions contain several smaller basins. Four regions are located in Eurasia, and the study also includes a large endorheic basin in South America and Central Australia.

The method defines flux-inferred storage, or FIS, by integrating precipitation and evapotranspiration fluxes into and out of a region so they can be compared with GRACE surface water storage anomalies. Figure two compares GRACE-measured water storage with flux-inferred storage, or FIS, derived from observed precipitation and evapotranspiration across six basins from 2002 to 2020.

The curves often diverge substantially, and FIS frequently shows a stronger seasonal cycle, especially in the Asia-east and Asia-west panels. This matters because the mismatch tests whether observed hydrological fluxes can account for the storage variations detected by GRACE, including the prominent downward trends in the Caspian and Asia-west basins.

To quantify how much interannual variability can be explained by each flux contribution, the analysis calculates correlations between GRACE and flux-inferred storages. The overall trend is forced to zero for both GRACE and FIS to avoid correlation from trend matching.

The storage is deseasonalized by removing the monthly mean seasonal cycle to avoid correlation signals from the seasonal cycle. Observed fluxes are adjusted to satisfy the water-budget equation while accounting for the uncertainties of each component. The budget balancing ensures consistency with GRACE on both short and long timescales.

The energy budget is solved simultaneously and coupled to the water budget through a shared latent heating term. Figure three compares original seasonal water fluxes, shown as dashed lines, with budget-closed solutions, shown as solid lines, alongside their integrated storage cycles.

In Asia-east, precipitation and evaporation both peak around July, and the adjusted fluxes nearly cancel, producing very little seasonal storage variation, consistent with GRACE. Asia-west shows a February precipitation peak, later evaporation, and a stronger precipitation-driven storage contribution, while Australia displays a pronounced seasonal exchange between precipitation- and evaporation-driven storage.

Figure four compares seasonal fluxes on the left with the corresponding storage cycles on the right for the Caspian, Aral Sea, and Altiplano Plateau basins. Dashed lines show the original observations, while solid lines show budget-closed solutions constrained by GRACE.

The storage panels make clear how integrating precipitation, evaporation, and net flux produces smoother seasonal anomalies, while the differing shapes across basins reveal that the dominant driver of storage variability is basin-dependent. On seasonal timescales, precipitation and evapotranspiration data show insufficient cancellation, implying over-large seasonal variations in surface storage.

In most basins, precipitation drives the seasonal storage cycle while evapotranspiration dampens storage amplitudes; in the Caspian Basin, seasonal storage is driven by evapotranspiration, with precipitation remaining seasonally constant over the whole drainage basin. Budget closure mostly adjusts evapotranspiration, which has larger uncertainties, when fitting the GRACE data.

Figure five compares detrended, deseasonalized storage anomalies from GRACE, precipitation-based FISP, and evaporation-based FIS–E across six regions. The blue FISP curves generally track the timing and shape of the orange GRACE variations, while the red FIS–E curves often show little interannual variation or oppose storage changes.

This matters because the visual supports the authors’ attribution of much observed interannual storage variability to precipitation, while highlighting limitations in the original evaporation estimates. Interannual storage variability detected by GRACE can be clearly attributed to precipitation variations detected through the FISP metric in all these basins.

The original evapotranspiration data show very little interannual variability, and their variations generally dampen interannual storage variations. Following budget closure, FISE displays greater variability, enhancing its storage dampening role.

The quantitative metric is obtained by correlating the storage time series shown in Figure 5, and the correlations are presented in Table 1. A strong positive correlation between observed FISP and GRACE storage supports precipitation as the key driver of storage variability.

The correlation reaches plus zero point eighty-eight in the Caspian Basin and is preserved in each basin after budget closure. All basins apart from the Caspian show negative FISP and FISE correlations, indicating compensating flux impacts on storage. Table one correlates GRACE storage variations with precipitation-driven, evaporation-driven, and combined flux-inferred storages after trends and seasonal cycles are removed.

Observed precipitation-driven storage correlates positively with GRACE across all regions, reaching zero point eight eight in the Caspian, while closed-budget values are zero point eight nine there and zero point nine eight in both the Aral Sea and Asia–east. The final columns show that precipitation and evaporation components are often strongly anticorrelated after closure, highlighting how budget adjustment changes the inferred contributions to total storage variability.

On year-to-year and multi-year timescales, the correlation between precipitation-driven storage and observed GRACE variability ranges between zero point fifty-five and zero point eighty-eight across all basins. That strong correlation is maintained after budget closure.

Storage changes driven by precipitation alone from GPCP are too large compared to GRACE, while FLUXCOM evapotranspiration data generally have only very weakly compensating interannual variations. After budget closure, interannual evapotranspiration variability is substantially increased.

The initial time-mean water-budget imbalance ranges between twenty-one and forty-five percent of mean precipitation across all basins. These imbalances mean that the long-term storage trends seen by GRACE cannot be sensibly attributed. Even though precipitation and evapotranspiration are fundamentally out of balance, they still provide insight into seasonal and interannual GRACE variability.

Figure six compares annual precipitation with annual storage changes across six basins, using separate axes and trend lines over the eighteen-year record. The inset translates those trends into cumulative storage changes: for example, the Caspian shows minus thirteen point one centimeters from precipitation and minus eight point one centimeters from storage-change trends, while Asia-west shows plus nineteen point four and minus three point two centimeters, respectively.

This matters because it exposes how precipitation trends relate to, but do not necessarily explain, long-term GRACE storage changes. Linear trends in GRACE storage cannot be attributed because they depend on constant differences between precipitation and evaporation that are typically highly out of balance in the initial data.

Over the Caspian, the downward trend in precipitation is more than sufficient to explain a downward trend in storage. Without any initial imbalance in precipitation minus evaporation, that trend is still sufficient to explain seventy percent of the GRACE total storage loss over eighteen years.

Although the trends are not statistically significant because of the amplitude of interannual variability, they are consistent with precipitation as the dominant driver of low-frequency storage changes in the Caspian Basin. The analysis covers GRACE water storage variations from 2002 to 2020 and evaluates seasonal storage variations, interannual variability, and longer-term trends.

It considers raw observational data and data adjusted through inverse modeling to guarantee closed coupled water and energy budgets on monthly timescales. The energy balance allows year-to-year seasonal amplitude to vary through different energy flux components, particularly latent heat variations coupled to the water cycle.

Those energy variations could perhaps be verified against reanalysis or land surface temperature data, but a full energetic analysis is left for future work. Water flux trends cannot be detected with significance because of the large interannual variability.

The strong agreement between multi-annual GRACE storage and precipitation variations, especially over the Caspian Basin, lends no support to the suggestion that climate-change-driven evapotranspiration changes are responsible for the water storage trends seen by GRACE.

Across these endorheic basins, precipitation is the strongest explanation for seasonal to interannual GRACE storage variability, while long-term trends remain difficult to attribute because the initial water budgets are substantially imbalanced.