Many thousands of large dam reservoirs have been constructed worldwide during the last 70 years to increase reliable water supplies and support economic growth. Because reservoir storage measurements are generally not publicly available, so far there has been no global assessment of long-term dynamic changes in reservoir water volumes. We overcame this by using optical (Landsat) and altimetry remote sensing to
reconstruct monthly water storage for 6695 reservoirs worldwide between
1984 and 2015. We relate reservoir storage to resilience and vulnerability
and investigate interactions between precipitation, streamflow, evaporation, and reservoir water storage. This is based on a comprehensive analysis of streamflow from a multi-model ensemble and as observed at ca. 8000 gauging stations, precipitation from a combination of station, satellite and forecast data, and open water evaporation estimates. We find reservoir storage has diminished substantially for 23 % of reservoirs over the three decades, but increased for 21 %. The greatest declines were for dry basins in southeastern Australia (
Globally the number of large reservoirs – dams impounding more than 3 million m
Adding to the challenge, evidence is emerging that existing reservoirs in some regions have experienced diminished water storage. Recent water supply failures or near-failures have occurred in the US Colorado River Basin since 2000 (Udall and Overpeck, 2017), southeast Australia between 2002–2009 (Van Dijk et al., 2013), Barcelona, Spain, in 2007–2008 (March et al., 2013), São Paulo, Brazil, in 2014–2015 (Escobar, 2015), and Cape Town, South Africa, in 2015–2017 (Sousa et al., 2018). However, it is unclear if these events are part of a global climate trend or due to local supply or demand changes. The underlying causes are also not necessarily the same in each case; reservoir storage dynamics are the net result of river inflows, net evaporation (i.e., evaporation minus direct precipitation onto the reservoir), and dam water releases to water bodies and users downstream. A change in the balance between these three terms leads to a change in the storage level. There are also interactions. The physical connection between precipitation, streamflow generation, and atmospheric moisture demand creates positive feedbacks in storage volume changes. For example, assuming the entire water supply system experiences comparable dry conditions, inflows will decrease while net evaporation and downstream demand for water releases for consumptive use will increase. To mitigate this feedback, reservoir operation rules will typically aim to reduce dam releases in response to lowering storage levels. Only a detailed analysis of the water balance of an individual reservoir can conclusively separate the contributions of these three processes to a change in water storage. However, in practice, a loss of reservoir water storage in the presence of a decrease in upstream or downstream river flows within the river system indicates that reduced precipitation conditions are the most likely primary cause, whereas the absence of such a precipitation and streamflow decrease, or even an increase, points towards less prudent reservoir operation, possibly in response to increased demand. Therefore, knowledge of temporal trends in reservoir storage and river flow can be combined to interpret whether trends in reservoir water storage are widespread globally, and if so, whether they are likely to be due to changing climate conditions or due to other factors. For the majority of large reservoirs, operators keep records of releases and estimated storage volume, inflows, and net evaporation. Unfortunately, these data are typically not publicly available for a variety of commercial, logistical, political, and security reasons. Probably mainly because of this, so far, there has been no attempt at a global assessment of long-term dynamic changes and attribution of trends in water reservoir storage.
Satellite remote sensing has been widely used to measure reservoir water
height, extent, and storage. Mulligan et al. (2020) developed a global
georeferenced database containing more than 38,000 georeferenced dams and their associated catchments, but without any descriptive features and
measurement information. The Database for Hydrological Time Series over Inland
Waters (DAHITI) (Schwatke et al., 2015) and the US Department of Agriculture's Foreign Agricultural Service (USDA-FAS) Global Reservoirs and Lakes Monitor (G-REALM) (Birkett et al., 2010) are the two most comprehensive datasets offering global surface water body height variations derived from satellite altimetry, such as Jason-1, Jason-2, Jason-3, TOPEX/Poseidon, and ENVISAT. Several regional and global time series of reservoir water extents have been produced based on MODIS, Landsat or Sentinel-2 imagery (Khandelwal et al., 2017; Ogilvie et al., 2018; Yao et al., 2019; Zhao and Gao, 2018; Schwatke et al., 2019). Reservoir volume dynamics can be estimated at either a regional or global scale using existing datasets and approaches to derive both height and extent from remote sensing, but this approach is only suitable for a limited subset number of reservoirs worldwide due to wide spacing of the satellite altimetry tracks (Busker et al., 2019; Tong et al., 2016; Medina et al., 2010; Crétaux et al., 2011; Duan and Bastiaanssen, 2013; Gao et al., 2012; Zhang et al., 2014). Messager et al. (2016) estimated the volume of lakes and reservoirs with a surface area greater than 0.10 km
In this study, we combined Landsat-derived surface water extents, satellite altimetry, and geo-statistical models to reconstruct monthly reservoir storage globally for 1984–2015, and examined long-term trends of global reservoir water storage and changes in reservoir resilience and vulnerability over the past three decades. Part of our objective was to determine the extent to which climate variability and human activity each affected global reservoir dynamics over the past three decades. It is currently impossible to analyze the influence of human activity at a global scale directly as there are very few in situ reservoir water release records available publicly, and no hydrological models that can provide reliable estimates. Instead, we consider all climate terms in the reservoir water balance and infer the influence of the remaining unknown term – water releases. First, we investigated trends in precipitation, streamflow, and storage at both the reservoir and basin level. If the trends between these variables show similar spatial patterns globally, then this increases the likelihood that climate variability mostly explains storage changes. Second, we examined the temporal correlation between precipitation, reservoir inflow, and storage change to further understand potential causative relationships. Third, beyond reservoir releases, net evaporation is the only other potential loss term, and we examined what fraction of observed trends in storage was attributable to net evaporation. Using the combined insights, we deduced the role of human activity on reservoir storage changes, noting that a direct attribution would require in situ records of reservoir water releases. To support our inference, we analyzed the trends of global water withdrawal to discuss whether it could be a significant factor in reservoir storage changes.
The Landsat-derived Global Surface Water Dataset (GSWD) (Pekel et al., 2016) provides statistics on the extent and change of surface water at the global scale over the past three decades at a spatial resolution of 30 m. Clouds, cloud shadows, and terrain shadows cause errors or missing data for individual months, but Zhao and Gao (2018) developed an automated method to fill gaps in contaminated image classifications and enhance the accuracy and consistency of reservoir surface water extent estimates. They applied this method to produce a monthly time series of a surface water extent dataset for 6817 reservoirs worldwide, based on mapping of the location and high-water mark as contained in the Global Reservoir and Dam database (GRanD) (Lehner et al., 2011). The average Pearson correlation (
List of the spatial data used in the analyses with source, resolution, and temporal coverage of data.
n/a: not applicable
The US Department of Agriculture's Foreign Agricultural Service (USDA-FAS)
provides near real-time surface water height anomaly estimates every 10 d
for 301 lakes and reservoirs worldwide. The water surface height product (G-REALM) was produced by a semi-automated process using data from a
series of altimetry missions, including Topex/Poseidon (1992–2002), Jason-1
(2002–2008), Jason-2 (2008–2016), and Jason-3 (2016–present) (Birkett et al., 2010). The root mean square error (RMSE) of G-REALM altimetry data is expected to better than 10 cm for the largest water bodies (e.g., Lake Victoria; 67 166 km
Daily and monthly in situ river discharge observations were collated as part of previous research (Beck et al., 2020) from different national and
international sources (Table S1 in the Supplement). In total, we archived 22 710 river gauging records. Global monthly surface runoff estimates for 1984–2014 were derived from the eartH
In total, 132 large reservoirs had records of both the surface water extent and
height for the overlapping period 1993–2015. The interpretation of the Pearson
correlation depends on the adopted
The total storage capacity in Group A (red) and B (brown), that left unaccounted for (blue), and the combined capacity of reservoirs for which the data were suitable (teal) or unsuitable (pink) for long-term analysis.
For 6637 reservoirs with water extent observations only (Group B; Fig. 1),
we used the HydroLAKES method (Messager et al., 2016) to estimate storage. There are typically two ways to estimate bathymetry based on the digital elevation model (DEM) for reservoirs that have no satellite altimetry measurements from space. The first approach is to develop an area–elevation curve based on a DEM (Avisse et al., 2017; Bonnema and Hossain, 2017). The second method is to extrapolate surrounding topography from the DEM into the reservoir to estimate the mean depth (Messager et al., 2016). Although the accuracy of these methods depends on errors inherent in DEM data, the latter one has been proven to be a reliable and effective way to estimate the bathymetry of global lakes and reservoirs. A Pearson correlation between predicted and reference depths of
Messager et al. (2016) proposed a geo-statistical model that provides the empirical relationship of the mean lake or reservoir depth, with water surface area and the average slope within a 100 m buffer around the water body. The main assumption of this model is that lake bathymetry can be extrapolated from surrounding topography using slopes. Four empirical equations to predict the depth from area and slope were developed by Messager et al. (2016) for different lake size classes (i.e., 0.1–1, 1–10, 10–100, and 100–500 km
There are 6862 reservoirs reported in the GRanD database (Lehner et al., 2011), with a total 6196 km
We calculated linear trends between 1984–2015 in annual reservoir storage,
observed streamflow, modeled streamflow, and precipitation for each basin
(HydroBASINS Level 3). Trend significance was tested using the Mann–Kendall
trend test (
Changes in reservoir resilience and vulnerability between 1984–1999 and 2000–2015 were analyzed at the scale of river basins. The reliability,
resilience and vulnerability (RRV) criteria can be used to evaluate the
performance of a water supply reservoir system (Hashimoto et al., 1982;
Kjeldsen and Rosbjerg, 2004). The calculation requires that an unsatisfactory state can be defined, in which the reservoir cannot meet all water demands, leading to a failure event. “Reliability” indicates the probability that the system is in a satisfactory state:
Validation of monthly reservoir storage time series reconstruction against in situ storage data.
In situ monthly storage records from the US Army Corps of Engineers, US Bureau of Reclamation, and Australian Bureau of Meteorology were used for
error assessment. There are a total of 131 reservoirs with at least a 20-year
overlapped time series between in situ data and satellite-derived data. We
did validation for all these 131 reservoirs (5 for Group A and 126 for Group B). The averaged correlation between observed and estimated volumes is 0.82 and
Validation of monthly reservoir storage time series reconstruction for Group B against results obtained using the method for Group A.
The trends (
The trends of storage
The resilience of reservoirs in southwest America (including the Mississippi Basin), central Chile, eastern South America, southeastern Australia, the
coast of southeastern Africa, and central Eurasia has reduced sharply between 1984 and 2015, and the vulnerability of these reservoirs has increased by more than 30 % (Fig. 5). In contrast, reservoirs in western Mediterranean basins, the Nile Basin, and southern Africa have stronger resilience and less vulnerability than before (Fig. 5). All these changes are attributed to changes in reservoir storage, as we found there is a robust positive relationship (
The change in resilience
The relationship (dashed gray line:
Linear trends in annual, basin-average
We summed the storage for individual reservoirs to calculate the combined storage in
134 river basins worldwide. Basins losing or gaining more than 5 % of their combined storage over the three decades could be found on every continent (Fig. 7c). Among these, 27 (20 %) showed a significant decreasing and 37 (28 %) a significant increasing trend in reservoir storage (Fig. 7c). If precipitation and runoff trends show the same direction as reservoir storage trends, then it is plausible that climate variations play an important role in reservoir storage trends. On the other hand, if rainfall and runoff show opposite trends to those in reservoir storage, then that could suggest a dominant influence from either net evaporation or water releases. For the majority of the 64 basins with a significant trend, the trends were of the equal sign for storage, runoff, and precipitation, suggesting that precipitation changes are commonly the most likely explanation for observed trends (Fig. 7a and b). Opposite trends in precipitation (or runoff) and storage were found for 12 out of 134 basins, with six showing decreasing and six increasing storage trends. Most of these could be explained by spatial variations in runoff trends within the respective basins (Fig. S3). The linear changes in modeled streamflow were validated against changes in observed streamflow, and the Pearson's correlation between them is 0.77. This indicated that modeled streamflow can reliably represent trends in river flow globally (Fig. 8b). There is a robust positive relationship
(
The relationship (dashed gray line:
The correlations of annual storage changes and reservoir inflow (as approximated by basin-modeled streamflow)
The ratio of the linear trends in net evaporation and storage for each basin.
The greatest storage gains occurred in the Nile Basin (
Time series
This study reconstructed monthly reservoir water storage dynamics from 1984–2015 at the global scale based on satellite-derived water extent (Zhao and Gao, 2018) and altimetry measurements (Birkett et al., 2010). Where no altimetry data were available, geo-statistical models (Messager et al., 2016) were applied to the satellite-derived water extent for reservoir water volume estimations. About a quarter (22.5 %, including most large reservoirs worldwide) of the total reported total combined reservoir capacity (Lehner et al., 2011) around the world was measured by combining satellite-derived extent and height, while 41.1 % was estimated based on geo-statistical models using remotely sensed surface area. There does not appear to be any systematic global decline in global reservoir water availability, but we found significantly decreasing trends in reservoir water volumes in southeastern Australia, southwestern USA, and eastern Brazil, creating the risk that storages fall to low capacity more often (i.e., weakened resilience) and endure larger deficits (i.e., higher vulnerability).
Linear trends in annual water withdrawal between 1984–2010 (gray shade: no reservoir data; black outlines: trend significant at
To understand the influence of the reservoir size distribution on the total basin storage trends, we compared the trend directions of total storage in all reservoirs, the top three largest reservoirs, and the remaining small reservoirs, respectively. We did this for 42 basins with more than 20 reservoirs (4003 reservoirs in total). Combined storage in these three groups all showed the same trend direction in 27 (62.8 %) of basins. The trend in the combined storage for all reservoirs had the same direction as that for the largest few reservoirs for eight more basins and the same direction as the combined remaining smaller reservoirs for another eight basins. This indicates that the largest reservoirs do not always dominate the combined total storage dynamics.
Trends in reservoir storage and river flow showed spatial consistency at both individual and basin scales globally. There was a reasonably strong temporal correlation between precipitation, streamflow, and storage. Changes in net evaporation only accounted for a small fraction of reservoir volume changes. Mady et al. (2020) and various other authors found that evaporative losses can account for much of the loss of water from small reservoirs (e.g.,
Comparison of trends in reservoir storage reconstruction against climate variability and human activities.
There is currently no global hydrological model capable of estimating the
impact of historical operational water management at the reservoir or basin
level with meaningful accuracy. However, to get an indication of the
potential impact of human activity and associated reservoir water releases on reservoir storage changes, we analyzed the global water withdrawal estimates produced by Huang et al. (2018). The gridded monthly withdrawal time series for 1971–2010 were spatially and temporal downscaled from 5-year temporal resolution estimates from FAO AQUASTAT and USGS, which were based on national assessments and surveys (Huang et al., 2018). Their estimates provide separate water withdrawal estimates for irrigation, hydroelectricity, domestic, livestock, manufacturing, and mining, respectively. The withdrawals are from reservoirs, rivers, and groundwater and, as such, cannot be compared directly to reservoir water release, but may provide useful context. We calculated total withdrawals from the six sectors combined and examined trends from 1984–2010 at the basin scale. The results show significant increasing trends in withdrawals in 78 basins, mainly in South America, Africa, and Asia, and significant decreasing trends in 29 basins in Europe, Australia, and parts of Northern America, noting, however, that the magnitude of withdrawals varied widely compared to, for example, total river inflows or reservoir capacity (Figs. 12 and S6). The global pattern in water withdrawal trends is different from the spatial patterns in precipitation, inflow, and storage (Fig. 7). We calculate that (either significant or non-significant) water withdrawal trends are associated with about equal numbers of increasing and decreasing water storage trends (Table 2). By contrast, rainfall and inflow trends lead to a change in storage in the same direction for around 80 % of basins (Table 2). These observations further support the notion that climate trends rather than water withdrawals are primarily responsible for the observed trends in reservoir storage. Nonetheless, there are basins where storage trends may have been influenced by water withdrawals. For example, inflows increased by 43 % in northern Venezuela, while total reservoir storage deceased by 15 %, conceivably because water withdrawals tripled from 1984–2010 (Figs. 7 and 12). A comparable scenario also occurred in coastal basins in Angola, Mozambique, Tanzania, and Kenya. Storages in Iran, Turkmenistan, and northern India decreased by an average 33 %, which may be attributed to an unknown combination of reduced inflows (
Accurate temporal pattern estimates were the main purpose in this study because relative water storage and long-term change are more relevant information for water resource management. Our validation results show that
82 % of the reservoirs evaluated show a strong correlation (
The uncertainties and limitations of reservoir storage estimates are mainly from the errors in satellite altimetry data and satellite-derived water extent data. The quality and accuracy of these altimetry measurements depend on the size and shape of the water body, surrounding topography, surface waves, major wind events, heavy precipitation, tidal effects, the presence of ice, and the position of the altimeter track (Birkett et al., 2010; Busker et al., 2019). The RMSE of water level estimations of a narrow reservoir in steep terrain will be many tens of centimeters (Birkett et al., 2010; Schwatke et al., 2015). DAHITI altimetry data, with RMSE between 4–36 cm for lakes (Schwatke et al., 2015), should have a similar accuracy as G-REALM, although its water level observations have so far received less evaluation. The classifier used to produce GSWD surface water data performed quite well, with less than 1 % commission error and less than 5 % omission error (Pekel et al., 2016). But no-data classifications in GSWD data caused by cloud, ice, snow, and sensor-related issues could lead to large data gaps in time series and an underestimation of actual reservoir extents (Busker et al., 2019). In general, a no-data threshold is applied to monthly GSWD data to remove imagery with a large percentage of contamination before deriving lake and reservoir water extents. It helps reduce the issue to some extent, but contaminated imagery would still remain in the rest of GSWD data. Zhao and Gao (2018) developed an automatic algorithm to repair contaminated Landsat imagery. This has increased the number of effective images by 81 % on average and produces continuous reservoir surface area dynamics.
The higher hypsometric correlation we used, the less uncertainties volume
estimations would have (Crétaux et al., 2016). We selected a correlation threshold of 0.7 in this study, which is lower than Tortini et al. (2020) (
The total number of Landsat images over North America, southern South America, southern Africa, central Eurasia, and Australia over the past three decades is much larger than in the rest of the world, and particularly in tropical regions (Pekel et al., 2016; Wulder et al., 2016). Regions with sparse Landsat observations can have additional uncertainties in their long-term trend analyses, although this issue has been mitigated to some extent by the approach developed by Zhao and Gao (2018). In principle, the inflow of sediments into reservoirs could contribute to decreasing storage. However, Wisser et al. (2013) showed that sedimentation caused a total decrease of global reservoir water storage of only 5 % over a century (1901 to 2010) and, hence, we expect the effect of sedimentation on our 32-year analysis to be small. There are studies showing higher sedimentation rates (e.g., Syvitski et al., 2022), so the impact of sedimentation on reservoir trend analyses cannot be discounted entirely. Thus, decreasing storage volumes could be exacerbated by sedimentation, while increasing storage volumes could potentially be (partly) explained by it.
Regional storage trends in the dam reservoirs found here are consistent with trends reported in a previous study for 200 lakes (including a few reservoirs) across North America, Europe, Asia, and Africa during 1992–2019 (Kraemer et al., 2020). Both lakes and reservoirs are influenced by changing inflow and net evaporation in response to climate variability. Although human regulation has more influence on reservoirs than on natural lakes, our results suggest that for the majority of basins, natural influences dominate human impacts, although human impacts on the hydrological regime still exist, of course. For example, Cooley et al. (2021) found that human interventions have resulted in larger seasonal variability in reservoirs than in lakes globally. In line with the study carried out by Kraemer et al. (2020), we also found that the distribution of global lake and reservoir storage or level long-term trends do not fully reflect the “wet gets wetter and dry gets dryer” paradigm that some have predicted to occur due to anthropogenic climate change (Wang et al., 2012). Reservoirs in dry regions, such as southwest America, southeastern Australia, and central Eurasia, have indeed seen deceasing combined storage, while those in wet regions, such as northern North America, have increasing storage. However, we found increasing storage in dry southern Africa and decreasing storages in wet southeastern South America. Additionally, total terrestrial water storage (i.e., the sum of groundwater, soil water, and surface water) derived from GRACE satellite gravimetry for the shorter period 2002–2016 showed decreases in endorheic basins in Central Eurasia and the southwestern USA, and increases in Southern Africa consistent with our storage changes (Wang et al., 2018).
Reservoir storage dynamics are the net result of river inflows, net evaporation, and dam water releases. We found a reasonably strong relationship between changes in river flow and reservoir storage, while changes in net evaporation do not seem to have affected storage trends significantly. We infer that reservoir water releases are unlikely to be the dominant driver of the three-decadal trends in reservoir storage for the majority of basins. However, we acknowledge that this particular conclusion is based on deductive rather than observational evidence, and would benefit from corroboration for any individual reservoir using actual release records, which often exist but are not publicly available. Although there are no water demand and supply or dam operation data available globally that could serve as direct evidence, there have been local studies. For example, reservoir operating rules (i.e., reservoir outflow) were inferred from a combination hydrologic modeling and satellite measurements for the Nile Basin, the Mekong Basin, northwest America, and a region of Bangladesh (Bonnema and Hossain, 2017; Bonnema et al., 2016; Eldardiry and Hossain, 2019). It was not possible to apply the techniques used in these studies at global scale because of the resulting uncertainties in inferred reservoir inflows. To distinguish the respective influences of human activity and climate variability on reservoir dynamics, greater collaboration and public sharing of in situ data on reservoir storage, water release, and downstream water use would be required. In some basins, satellite-derived upstream and downstream river discharge dynamics (Hou et al., 2018, 2020) and changes in irrigation area or evaporation (Van Dijk et al., 2018) may be able to provide additional information to better understand the drivers of reservoir water security. The algorithm from Zhao and Gao (2018) could, in principle, be used to calculate reservoir surface water extent time series beyond 2015, but is reliant on the availability of Landsat-derived GSWD (Pekel et al., 2016). These data could also be derived from MODIS or Sentinel 2 and help understand how reservoir water storage changes from 2015 onwards. The new NASA Surface Water and Ocean Topography (SWOT) satellite mission should also provide new opportunities to cover a larger number of reservoirs (
We reconstructed monthly storage dynamics between 1984–2015 for 6695 reservoirs using satellite-derived water height and extent. For reservoirs with water extent data only, the storage was estimated from the surrounding topography using a geo-statistical model. This approach introduces uncertainty but is inevitable as lake bathymetry data based on surveys are typically unavailable, at least in the public domain. The estimated reservoir storage dynamics show strong correlations with averaged
Code is not publicly available, but further information or details can be obtained contacting the corresponding author.
The Global Reservoir Surface Area Dataset (GRSAD) is
available from the Gao Research Group, Texas A & M University (
The supplement related to this article is available online at:
JH and AIJMVD conceived the idea. AIJMVD, HEB, LJR, and YW guided the study. JH carried out the analysis and wrote the manuscript with contributions from all the co-authors.
The contact author has declared that none of the authors has any competing interests.
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
This study was supported by the ANU-CSC (the Australian National University and the China Scholarship Council) Scholarship. Calculations were performed on the high-performance computing system, Gadi, from the National Computational Infrastructure (NCI), which is supported by the Australian Government. We also thank Bernhard Lehner of McGill University for his feedback on an earlier version of this paper.
This paper was edited by Thom Bogaard and reviewed by three anonymous referees.