We have used a series of analyses to operationalise the safe and just ESBs for blue water. The first was to determine where we are already outside the ESBs for surface and groundwater. The second was to quantify whether we have sufficient surface water flows to the minimum water needs for all people to escape from poverty relying on surface water alone. The third was to quantify what proportion of groundwater recharge we would need to draw on to meet human needs while respecting the surface water ESB. The fourth was to quantify the trend in annual groundwater recharge and annual rainfall volumes.
Determining where the ESBs for surface and ground water cannot be met
We identified when a location is already outside the safe and just ESB for surface water by comparing modelled observed (altered) monthly flows with modelled pristine (unaltered) monthly flows. We calculated 20% of the long-term mean annual pristine flows at grid cells throughout global river networks as a spatially distributed volume of annual alteration that is within the safe and just ESB, leaving 80% of annual flows unaltered to protect aquatic ecosystems and the ecosystem services they provide. To quantify the extent to which river flows are outside the safe and just ESB in a given catchment, we first calculate the number of months in a year where the contemporary altered flows are more than 20% different from pristine flows using the long-term mean gridded discharge data. We then represent this data as the proportion of months in a year with more than 20% difference for each grid cell in the global river networks. For the purposes of the spatial analyses, we defined a catchment as being outside the safe and just ESB when the long-term mean observed total annual discharge at the river mouth was more than 20% different from the long-term mean pristine total annual discharge. We used total annual discharge for this analysis for comparison with the groundwater ESB, which is on an annual time step.
We identified regions that were outside the safe and just ESB for groundwater by comparing the long-term trend in groundwater storage volumes. Regions where the average annual drawdown exceeded average annual recharge showed an ongoing decline in groundwater storage and were defined as being outside the safe and just ESB for groundwater. Complementing this analysis was an analysis of the trend in annual groundwater recharge.
Quantifying whether there is sufficient surface water flows to meet minimum human needs
We compared the volume of water that was available under the safe and just ESB at the catchment scale with different volumes of water to meet human needs. We calculated total annual volumes of water required for basic human needs based on two different Minimum Access levels28 for per capita daily water needs (Table 2, adapted from Rammelt et al.28). The two levels of Domestic access needs in Table 2 were defined according to the volume of water required to meet daily domestic needs to live a dignified life (Level 1) and to escape from poverty (Level 2). The two access levels of All needs represent the demand on the hydrological cycle and include the same domestic needs and the volume of water required to produce food, energy and infrastructure at the two access levels. See Rammelt et al.28 for the full methodology used to derive these numbers.
Table 2
Minimum water needs defined by Rammelt et al.28 to maintain a dignified life and to escape from poverty.
Water Demand Metric
|
Litres/capita/day
|
Litres/capita/year
|
Domestic access Level 1 (dignity)
|
50
|
18,250
|
Domestic access Level 2 (escape from poverty)
|
100
|
36,500
|
All needs access Level 1 (dignity)
|
292.85
|
106,890
|
All needs access Level 2 (escape from poverty)
|
405.67
|
148,069
|
The daily per capita water needs were converted to spatially distributed gridded annual volumes by multiplying the demand metrics by a distributed population dataset for the year 202063 and then summed over river basins. Long-term mean annual discharge and available surface water discharge at the basin mouth is used to define integrated water flows for the river basins. Where the annual alteration budget under the ESB for surface water was greater than the per-capita water needs for the resident population, we determined that it is possible to meet human needs from water within that catchment while meeting the ESB. Where the annual alteration allowed under the ESB was less than the per-capita water needs for the resident population, we determined that it is not possible to meet human needs from water within that catchment while meeting the ESB, creating a safe water deficit. For the purposes of this analysis, we made no assumptions around water storage capacity or monthly alteration levels that would be required to meet these human needs.
Quantifying what proportion of groundwater recharge is needed to meet human needs
In catchments where we cannot meet safe water needs with surface water from within the catchment, we may be able to rely on groundwater recharge to meet these needs. To quantify the extent to which groundwater recharge would be required, we converted safe water deficits for All needs at access level 2 to a proportion of the total annual groundwater recharge. We summed groundwater recharge volumes over river basins for these basin-level calculations. We calculated the proportion of groundwater needed to meet the safe water deficits as the ratio of the Safe Water Deficit and the average annual recharge volume.
Assessing where groundwater recharge volume is in decline
We identified pixels, and then regions, where groundwater recharge volume was in decline by quantifying the annual recharge in a given pixel and then quantifying the trend in the annual recharge. At the catchment scale, we calculated the average trend of all pixels in each catchment to define the status of whether recharge has been in decline in that catchment or not. We accompanied this with similar analyses of annual rainfall volumes to identify where declining groundwater recharge volumes is associated with declining annual rainfall.
Global surface water hydrology
We derived the pristine and disturbed monthly river flow datasets from the WBM water balance model river discharge outputs64 at 6-minute grid cell resolution using the TerraClimate high resolution data set of monthly climate forcings65 for the period 2000–2020. River basin delineation and flow routing configurations are defined by the WBM 6-minute topological river network used to establish local discharge and river flow64. The pristine and disturbed WBM runs use the same climate forcings for the 2000–2020 time period but only employ human alterations to the water cycle, including water extraction for irrigation and large reservoirs, in the disturbed runs. The modelled long-term mean contemporary global annual discharge of 38,000km3 under this scenario is consistent with other results from the literature66,67.
Long-term mean monthly discharge is calculated for the modelled pristine (non-human impacted) and disturbed (human impacted) discharge from the WBM model over the 2000–2020 time domain to determine the extent of altered flow. The analysis is limited to only the perennial or actively flowing river extents by applying a 3mm/yr upstream monthly average runoff exceedance threshold68 occurring for at least 10 years out of the 2000–2020 time domain. We also mask out upstream headwater areas (smaller than 250km2) that have modelled irrigation depths below the median irrigation depth for small headwater cells (3.6 mm/yr). This mask is applied to eliminate noise in the modelled data associated with very low irrigation and discharge values in headwater grid cells. River network and basin extents are defined by the WBM water balance model with naming convention taken from the GRDC Major River Basins of the World69.
Global groundwater dynamics
Hydrological measurements of volumetric changes in aquifer storage are critical in assessing groundwater status but these measurements are considerably limited in several regions of the world. Given that the aquifers in some regions are typically not monitored, global scale assessments of baseline aquifer volumes are difficult. To circumvent this, the Gravity Recovery and Climate Experiment (GRACE70) mission has been used to track changes in several large aquifer systems around the world71. In this study, changes in groundwater were quantified using the GRACE data covering the period 2003–2016 (data files accessed at http://www2.csr.utexas.edu/grace/RL06_mascons.html). GRACE measures monthly changes in terrestrial water storage (TWS), being the sum of soil moisture, groundwater, surface water, snow water, and canopy storage and is expressed as:
\(TWS=SMS+GWS+SWE+SWS+CS\) Eq. 1
where SMS is the soil moisture storage change, GWS is the change in groundwater storage, SWE is the change in snow water equivalent, SWS is the change in surface water storage (i.e., inland surface and reservoir storage) and CS is the water storage change in canopy. To quantify GWS, Eq. 1 was rearranged such that SMS, SWE and CS, which are model-derived outputs from the Global Land Data Assimilation (GLDAS) NOAA Land Surface Model L4 v2.172 were subtracted from TWS. Outputs from hydrological (e.g., SMS, SWE, and CS) obtained from model simulations may be characterised by large uncertainties due to inadequate in-situ data for calibration and parameterization, as well as the presence of strong inter-annual and seasonal variability in surface reservoirs and snow/ice cap storage in some regions. In some regional groundwater studies, the effects of interannual changes in surface water component such as those from major lakes and reservoirs are significantly strong and have been reasonably managed and removed from the GRACE-observed TWS using data reconstruction and synthesised kernel functions73,74. However, a global scale groundwater processing protocol or the isolation of surface water footprint from the GRACE hydrological water column using model simulation is rather impracticable and not feasible. Alternatively, the water storage components (SMS, SWE, SWS) in Eq. 1 above have been captured in the WaterGAP Global Hydrology Model75 and by directly subtracting these WGHM outputs from TWS will result in groundwater changes. The uncertainties in these WGHM water storage components are unknown and arguably could amplify the estimated groundwater changes from TWS, especially in regions where WGHM outputs performed poorly (e.g.,71,74).
To cushion the effect of such errors and uncertainties that may be propagated from this approach, much of the regions with substantial inland surface water storage changes (e.g., Caspian Sea, Black Sea, Lake Victoria, and other significant water bodies) were masked out (regions with no groundwater signal). Additionally, uncertainties caused by residual ice/snow cover from areas (e.g., Patagonia, Alaska, Himalayas, Swiss Alps, etc.) with large variations were minimised by masking such regions using the world distribution of glaciers and ice caps extents (geospatial data layer showing boundaries of such glaciers). This decision acknowledges the higher uncertainties in the simulations of these quantities by the GLDAS model. Further, some glaciers are small and may be obscured but a buffer zone of 1 arc degree was therefore created to help flag and remove such glaciers. Overall, the groundwater estimation process here is based on the water budget approach, which has been widely used in GRACE-derived groundwater storage studies (e.g., 76,77). There are several GRACE-TWS products available from different providers, but the TWS data used in our study is a mass concentration (mascon) product (GRACE RL 06 version 02) obtained from the Center for Space Research. These mascon products are preferred to other GRACE solutions (e.g., those based on spherical harmonics) since it exhibits less signal leakage and a posteriori filtering is unnecessary as the mascon product relies on geophysical constraints to suppress noise in the data (e.g., 78).
The estimated annual recharge volume used in this study was based on the time series of groundwater anomalies. Annual recharge was estimated by first aggregating the monthly groundwater data into annual values. Finally, for each groundwater pixel, recharge was calculated by quantifying the difference between the maximum groundwater depth in a given year and the shallowest observation of the preceding year (Fig. 5). Areas categorised as under risk of groundwater stress (i.e., groundwater in decline) were identified by computing the difference between the estimated annual recharge and draw down (also complemented by trend analysis in groundwater). Using the aggregated monthly groundwater data, the latter was quantified as the maximum groundwater values of a specific year less the observed minimum values of the following year). Notably, this draw down varies in time and space and could be human or climate driven. Accurate assessment of recharge using modelling techniques and chloride mass balance could be challenging because groundwater recharge is governed by complex interactions, including the relationship of climate change (e.g., prolonged drought) and human water abstraction with land surface conditions (e.g., increased evapotranspiration), geology and differences in water yield, among other factors. However, we found that our recharge estimates broadly aligned with some proposed estimates in the literature and other reports79. Moreover, the spatial distribution of trends in the time series of global groundwater and annual recharge were estimated using the least squares approach.
Trends in annual rainfall volume
The trends (mm/yr) in annual rainfall were based on monthly Global Precipitation Climatology Centre (GPCC) precipitation data (mm). The GPCC-based precipitation is one of the widely used gridded rainfall products because it consists of quality-controlled observational data from 67,200 gauged stations world-wide80. The GPCC data is available on a 0.25° spatial resolution and was accessed from NOAA’s repository (https://psl.noaa.gov/data/gridded/data.gpcc.html). The monthly grids (spatial and temporal dimensions) of GPCC rainfall were generated from the scientific file format (popularly called Network Common Data Form) and accumulated to annual values using scripts written in Matlab R2018A version, underpinned by the Mapping and Aerospace Tool boxes. The least squares approach was then used to estimate the trends in the time series of the annual rainfall data for the period between 2002 and 2016, consistent with other data used in this study.