The fate of domestic and manufacturing wastewater after production can follow a number of paths (Fig. 1). Wastewater from these activities can be collected, typically in sewers, septic tanks or pit latrines, or uncollected and discharged directly to the environment (e.g. open defecation). Collected wastewater can undergo treatment, ranging from basic primary treatment (removing suspended solids) to specialised tertiary or triple-barrier treatment (nutrient removal and toxic compound removal), or can be discharged to the environment untreated (Mateo-Sagasta et al., 2015). When treated, wastewater can be released to the environment or intentionally reused as a “fit-for-purpose” water source. Untreated wastewater can similarly be discharged to the environment or intentionally used as a source of freshwater. Furthermore, both treated and untreated wastewater can be used unintentionally where wastewater is incidentally present in a water supply (“de facto reuse”).

Country-level wastewater data were collated from four online databases (Table 1): Global Water Intelligence (GWI, 2015), the Food and Agriculture Organization of the United Nations (FAO AQUASTAT, 2020), Eurostat (Eurostat, 2020) and the United Nations Statistics Division (UNSD, 2020). For consistency, the year 2015 was selected for all wastewater data. Where wastewater data from the online sources were reported in a different year (up to a maximum of 10 years, i.e. 2006 onwards), all wastewater data were standardised to 2015 based on data from the most recent reporting year (see Table 1 for the standardisation method).

Data from different sources were cross-examined to check for consistency and to remove implausible data. Where large discrepancies (>1 order of magnitude) existed between different data sources for a country, data points were excluded. For example, the GWI reports Kazakhstan to produce 6205×106 m3yr-1, whereas the FAO reports just 411×106 m3yr-1. Similarly, the FAO reports Moldova to produce 46.7×106 m3yr-1 compared to 672.1×106 m3yr-1 by the UNSD. In total, reported data for 11 countries were excluded for wastewater production. For wastewater collection and treatment, percentage data were cross-referenced with reported connection rates (i.e. percentage of population connected to wastewater collection and treatment). Six and seven countries were excluded for collection and treatment, respectively. For example, the GWI reports a 95.2 % collection rate for Azerbaijan, while the UNSD reports that only 32.4 % of people are connected to wastewater collection systems. Similarly, the GWI reports a 17 % treatment rate in Hong Kong SAR, whereas the UNSD reports that 93.5 % of people are connected to wastewater treatment plants. No data points were excluded for wastewater reuse. In a small number of cases where percentage values obtained were marginally illogical (i.e. wastewater collection<wastewater treatment; wastewater treatment<wastewater reuse), percentage values were set to the proceeding level in the wastewater chain (Fig. 1).

Table 1 displays the data sources and the associated number of countries with wastewater data for production, collection, treatment and reuse. The procedure for standardising data to the year 2015, when required, is presented. Wastewater production is assumed to be dependent upon both population size and per capita production (related to per capita wealth). Hence, we standardise wastewater production linearly with GDP (gross domestic product), a combined metric of population size and wealth. Conversely, wastewater collection and treatment are assumed to be more dependent on economics; hence we linearly apply GDP per capita for standardisation. The methods used to compile wastewater production, collection, treatment and reuse data from multiple sources to provide a single quantification per country are also displayed. Lastly, the population coverage in percentage terms and by the number of unique countries is displayed both per geographic region and by economic classification. The number of unique countries and the population coverage of data at each stage of the wastewater chain are also displayed in Fig. 1. Reported wastewater data were available for the majority of the world's most populous countries. This results in a high-percentage population coverage relative to the number of countries. Both the number of countries and population coverage reduces through the wastewater chain, with available wastewater data decreasing from 118 to 37 countries (90 % to 60 % population coverage) from wastewater production to wastewater reuse data.

Datasets of predictor variables for regression analyses were downloaded from multiple sources (see overview Table 2). Datasets spanned a wide range of predictor variables covering social (e.g. total and urban population), economic (e.g. GDP or Human Development Index (HDI)), hydrological (e.g. irrigation water scarcity) and geographic (e.g. land area and agricultural land) dimensions. The selected predictor variables were expected to have a physical basis for correlation with wastewater production, collection, treatment or reuse. Where appropriate, datasets from these sources were combined to produce comparable metrics for countries of different populations and geographic sizes (e.g. GDP per capita in USD per capita and desalination capacity per capita in m3yr-1 per capita). Values were taken for the year 2015, where available, or from the closest reporting year when unavailable. Irrigation water scarcity and desalination capacity were taken from 2019 and 2018, respectively. Data were transformed, either using a log or square root transformation, to reduce the skew in the independent variables and to ensure normality.

Multiple linear regression was used to predict country-level wastewater variables (production, collection, treatment and reuse) for countries without reported data. Stepwise elimination was used for feature selection to remove redundant predictor variables and reduce overfitting. Wastewater production was predicted in volumetric flow rate units (106 m3yr-1). Conversely, wastewater collection, treatment and reuse were predicted as a percentage of wastewater production. Predicted values of percentages were bounded to the 0 %–100 % range (i.e. <0=0 and >100=100). Predicted percentages were subsequently applied to reported or predicted values of wastewater production to obtain wastewater collection, treatment and reuse in volumetric flow rate units. Bootstrap regression was used to quantify the uncertainty in the predictions (by geographic region, economic classification and at the global scale) at the 95th confidence level. In total, 1000 regressions with random sampling and replacement were fit to provide predictions at countries lacking data. Wastewater observations were combined with these 1000 bootstrapped predictions, with the 2.5th and 97.5th confidence intervals taken as lower and upper confidence limits, respectively.

Wastewater data (reported and predicted) are at the national level, for the 215 countries as listed by the World Bank (https://data.worldbank.org/country, last access: 5 January 2020). Wastewater data are also aggregated to eight geographic regions based on the World Bank regional classifications: (1) East Asia and Pacific, (2) eastern Europe and Central Asia, (3) Latin America and Caribbean, (4) Middle East and North Africa, (5) North America, (6) South Asia, (7) sub-Saharan Africa, and (8) western Europe. Furthermore, data are also aggregated to four economic classifications based on the World Bank Atlas method: (1) high income (>USD 12 056 GNI – gross national income – per capita), (2) upper-middle income (USD 3896 to USD 12 055), (3) lower-middle income (USD 966 to USD 3895) and (4) low income (<USD 995). Predicted wastewater data were used to supplement reported data and, where unavailable, to develop a comprehensive global outlook of wastewater production, collection, treatment and reuse.

Country-level wastewater production, collection, treatment and reuse data were downscaled to 5 arcmin resolution (∼10km at the Equator) based on the sum of averaged annual domestic and industrial return flow data (henceforth “return flow”). Return flows represent the water extracted for a specific sectoral purpose, but which is not consumed, and hence it returns to and dynamically interacts with surface and groundwater hydrology (de Graaf et al., 2014; Sutanudjaja et al., 2018). Return flows used for downscaling are calculated as gross - net water demands from the Water Futures and Solutions (WFaS) initiative for the years 2000–2010 (Wada et al., 2016). The WFaS water demand dataset follows the approach developed for PCR-GLOBWB (PCRaster Global Water Balance; Wada et al., 2014). Domestic return flows only occur where the urban and rural populations have access to water, whereas industrial return flows occur from all areas where water is withdrawn (Wada et al., 2014). Both domestic and industrial return flows are dependent on country-specific recycling ratios based on GDP and the level of economic development (Wada et al., 2011, 2014).

Grid cell return flow was divided by the country's total return flow to obtain the fraction per grid cell. Wastewater production was downscaled directly proportionally to return flows by multiplying the grid cell return flow fraction per grid cell with wastewater production at the country level. Wastewater collection is assigned sequentially to grid cells with the largest downscaled produced wastewater flows. Thus, collected wastewater is preferentially allocated to grid cells with the highest levels of municipal activities, where central wastewater collection (and treatment) is assumed to be most economically feasible. Wastewater treatment is assigned to grid cells only where wastewater collection exists, at an average treatment rate calculated at the country level. The treatment rate is calculated as the proportion of collected wastewater that undergoes treatment and, hence, can differ from the country-level wastewater treatment percentage (which is calculated as the proportion of produced wastewater that is treated). For the downscaling of wastewater reuse an additional criterion was introduced to represent water scarcity, a key driver of wastewater reuse. The ratio of water demand to water availability was calculated. Grid cells within a country with a treated-wastewater allocation are then ordered based off this ratio, and treated-wastewater reuse was assigned sequentially to these grid cells.

The location and design capacity of individual wastewater treatment plants were used to validate the downscaled wastewater treatment data. Reported data for 25 901 wastewater treatment plants located across Europe were obtained from the European Environment Agency (EEA, 2019). Data for a further 4283 wastewater treatment plants were obtained for the contiguous United States from the US Environmental Protection Agency (US EPA, 2020). An additional 478 wastewater treatment plants, distributed globally (excluding Europe and the US), were extracted from the GWI wastewater database (GWI, 2015). For EEA and GWI wastewater treatment plants, treatment capacity reported only in population equivalent (PE) was approximated in volume flow rate units based on the linear regression obtained for wastewater treatment plants reporting capacity in both population equivalent and volume flow rate (EEA: R2=0.80, p<0.001; GWI: R2=0.81, p<0.001). Wastewater treatment plants were assigned to their nearest grid cell, and treatment capacities were aggregated per cell. In total, wastewater treatment data were available for 22 133 unique grid cells. For validating downscaled wastewater reuse, only plants (with treatment capacity >1×106 m3yr-1) using tertiary or higher wastewater treatment technologies were considered. In total, 572 wastewater treatment plants in the EEA database met this criterion. A further 78 wastewater treatment plants, which are specifically designated as wastewater reuse facilities, were sourced from the GWI database. In total, wastewater reuse data were available for 601 grid cells. Downscaled wastewater treatment and reuse were compared to wastewater design capacities.

To account for the large variation in the treatment capacities of wastewater treatment plants considered, in addition to the geographical mismatch between where wastewater is produced and treated (i.e. wastewater treatment plants are typically located on the outskirts of urban areas), validation occurred at differing geographical scales. Wastewater treatment plant capacity was divided by wastewater production per capita to approximate the number of people that the wastewater treatment plant serves. If the population served by a wastewater plant exceeds the grid cell population, the validation extent was expanded to the directly neighbouring cells. This is allowed to occur, until the population served by the treatment plant is reached, only up to a maximum of three iterations, reflecting a radius of ∼30km around the wastewater treatment plant. The total downscaled wastewater treated over the extended area was then compared to that of the treatment plant.

To quantify the performance of the downscaling approaches, the root-mean-square error (RMSE) and mean bias (BIAS) were calculated. Normalised values of RMSE and BIAS were calculated (nRMSE and nBIAS) by dividing by the standard deviation of the wastewater treatment plant capacity. Pearson's (r) coefficients were calculated to quantify the linear dependence, with R2 values based on both the linear and log–log relationship between downscaled and observed values also being calculated.

The results of the regression analysis for wastewater production, collection, treatment and reuse are summarised in Table 3. All regression models were significant at the p<0.001 level with adjusted R2 values ranging between 0.61 and 0.89. Country-level observed versus simulated wastewater production (log 106 m3yr-1), collection (%), treatment (%) and reuse (%) data are displayed in Fig. 2. The regression equations were applied for 97, 113, 122 and 178 countries with no or excluded data representing 10 %, 14 %, 22 % and 40 % of the global population for wastewater production, collection, treatment and reuse, respectively.

Wastewater production was best predicted based on total population, GDP per capita and access to basic sanitation. A significant regression equation was found (p<0.001) with an adjusted R2 value of 0.89, with all predictor variables also significant at the p<0.001 level. While the number of people within a country was found to have the strongest influence on total wastewater production (β=0.96), the average economic output per inhabitant (β=0.31) and the level of access to wastewater services (β=0.19), such as flushing toilets to piped sewers, are important for determining the amount of wastewater produced per capita. These three factors therefore account for the combined effect of population size and variations in wastewater production per capita linked to economic and development factors in determining total wastewater production in a country. Comparing observed with predicted total wastewater production data demonstrates the overriding importance of a country's population, with wastewater production spread across multiple orders of magnitude for countries irrespective of geographical region or economic classification (Fig. 2a).

Wastewater collection was predicted (adjusted R2=0.69, p<0.001) based on the Human Development Index (HDI), urban population and wastewater production per capita. HDI, an overarching proxy for the level of development, was found to be the strongest influence over wastewater collection (β=0.50, p<0.001). Urban population (β=0.14, p<0.01) and wastewater production per capita (β=0.25, p<0.01) were also significant but less important predictor variables of wastewater collection. For urban populations, a greater proportion of a population living in urban areas resulted in higher collection rates for the country, while higher levels of wastewater production per capita corresponded to larger collection rates. The observed versus predicted wastewater collection rates are depicted in Fig. 2b, which displays the trend across different geographic zones and economic classifications.

Wastewater treatment was predicted (adjusted R2=0.80, p<0.001) based on GDP per capita (β=0.28, p<0.01) and wastewater collection (β=0.66, p<0.01). Countries with larger economic outputs per capita likely have more resources for wastewater treatment, resulting in higher overall treatment rates. As wastewater treatment is dependent upon wastewater collection, countries with higher wastewater collection rates typically also treat a greater proportion of their wastewater. Observed versus predicted wastewater treatment rates are displayed in Fig. 2c.

The amount of wastewater treated will determine the maximum potential for treated-wastewater reuse within a country. Water scarcity, particularly when driven by high irrigation water demands, is also a primary driver of wastewater reuse (Garcia and Pargament, 2015). To account for this relationship, the fraction of wastewater undergoing treatment processes and irrigation water scarcity was multiplied to give an integrated metric indicating the “availability of treated wastewater for irrigation water scarcity alleviation”. Wastewater reuse was predicted (adjusted R2=0.70, p<0.001) from this metric (β=0.60, p<0.01) in combination with the desalination capacity per capita (β=0.29, p<0.1), as an indicator of the prevalence of unconventional water resources in a country. The observed versus predicted wastewater reuse rates from this regression are displayed in Fig. 2d. Irrigation water scarcity data were unavailable for 53 countries, mostly small island nations. Here an alternate regression model was constructed based on desalination capacity per capita (β=0.63, p<0.01) and wastewater treatment (β=0.24, p<0.1) only, resulting in a slightly lower explained variance (R2=0.61). While these countries represent <1 % of the global population, this alternate regression was necessary to account for wastewater reuse occurring particularly in water-scarce small island nations. These islands typically lack renewable water resources and hence unconventional water resources such as desalinated water and treated wastewater represent a substantial proportion of the water availability (Jones et al., 2019).

Globally, this study estimates that 359.4×109 m3yr-1 (358.0×109–361.4×109 m3yr-1) of wastewater is produced annually, with a global average of 49.0 m3yr-1 (48.8–49.2 m3yr-1) per capita. Global annual wastewater collection and treatment is estimated at 225.6×109 m3yr-1 (224.4×109–226.9×109 m3yr-1) and 188.1×109 m3yr-1 (186.6×109–189.3×109 m3yr-1), respectively. These values indicate that approximately 63 % and 52 % of globally produced wastewater is collected and treated, respectively, with approximately 84 % of collected wastewater undergoing a treatment process. Wastewater reuse is estimated at 40.7×109 m3yr-1 (37.2×109–47.0×109 m3yr-1), representing approximately 11 % of the total volume of wastewater produced. This estimate also indicates that approximately 22 % of treated wastewater undergoes intentional reuse, with the remaining 78 % (totalling 147.4×109 m3yr-1) discharged to the environment. This compares to the estimated 171.3×109 m3yr-1 of wastewater discharged directly to the environment without undergoing any form of treatment. It is worth highlighting that the vast majority of wastewater data are from reported sources, with just 2.4 %, 4.8 % and 5.2 % of global wastewater production, collection and treatment being from predicted values using regression. This occurs both due to the high population coverage and due to the missing data primarily being from developing countries, where wastewater production per capita and percentage collection and treatment rates are lower. The global quantification of wastewater reuse relies more heavily on predicted values, constituting 23.4 % of reuse volume globally. This occurs primarily due to poor data availability, particularly in countries with large populations in eastern Europe and Central Asia (e.g. Russia, Turkey and Poland) and western European countries, where wastewater treatment rates are generally high but the proportion of wastewater reused relies on simulations (e.g. Germany, Italy and Greece).

Table 4 displays wastewater production per capita (m3yr-1 per capita) and wastewater production, collection, treatment and reuse (109 m3yr-1), aggregated from the country data (reported + simulated) at the global scale and by region and level of economic development. Figure 3 displays wastewater data plotted at the country scale in proportional terms (m3yr-1 per capita for production; percentage of produced wastewater for collection, treatment and reuse), facilitating direct comparisons between countries.

Substantial differences in wastewater production, collection, treatment and reuse occur across different geographic regions and by the level of economic development. Wastewater production per capita is notably highest in North America at 209.5 m3yr-1 per capita, over double that of western Europe (91.7 m3yr-1 per capita), the next highest producing region per capita. When considering individual countries in these regions, the USA (211 m3yr-1 per capita) and Canada (198 m3yr-1 per capita), in addition to small, prosperous European countries (e.g. Andorra at 257 m3yr-1 per capita, Austria at 220 m3yr-1 per capita and Monaco at 203 m3yr-1 per capita), are the highest producers per capita. Comparatively, the larger western European countries have lower wastewater production per capita, with Germany, the UK and France at 92, 92 and 66 m3yr-1 per capita, respectively. Conversely, most sub-Saharan African countries produce less than 10 m3yr-1 per capita. Wastewater production values are comparable to the World Health Organization's absolute minimum water requirements for survival of 2.7 m3yr-1 per capita (WHO, 2011) in countries such as Niger (2.7 m3yr-1 per capita), Burkina Faso (3.4 m3yr-1) and Ethiopia (4.2 m3yr-1 per capita). Aggregated for the region, sub-Saharan Africa produces approximately 20 times less wastewater than North America per capita, at 11.0 m3yr-1 per capita.

In volumetric flow rate terms, the East Asia and Pacific region produces the most wastewater (117.6×109 m3yr-1), coinciding with the largest population share (∼31%). Conversely, South Asia produces just ∼7% of global wastewater despite a population share of ∼24%, whereas the ∼5% of people living in North America account for ∼20% of global wastewater production. Wastewater production also varies greatly with level of economic development. The prominent discrepancies between economic classifications indicate a strong relationship between wealth and wastewater production regardless of geographic location. Wastewater production per capita more than doubles at each income classification level from low income (6.4 m3yr-1 per capita) to high income (126.0 m3yr-1 per capita). With respect to population size, people living in high-income countries (∼16% global population) produce ∼42% of global wastewater, compared to low- and lower-middle-income countries (∼50% global population) producing ∼20% of global wastewater.

Wastewater collection and treatment rates are highest in western Europe (88 % and 86 %, respectively) and lowest in South Asia (31 % and 16 %, respectively) and sub-Saharan Africa (23 % and 16 %, respectively). Wastewater collection is notably low in the East Asia and Pacific region, where total wastewater production is high. Conversely, wastewater collection in the Middle East and North Africa region is relatively high at 74 %, likely resulting from the lack of renewable water supplies. Wastewater treatment percentages follow similar regional patterns. Notably, wastewater treatment is substantially lower than wastewater collection in the Latin America and Caribbean and South Asia regions, potentially indicative of high rates of untreated-wastewater reuse in these regions. Wastewater collection and treatment percentages follow similar patterns as wastewater production with respect to income level, with high-income countries collecting and treating the majority of their wastewater (82 % and 74 %, respectively) down to low-income countries with small collection and treatment rates (9 % and 4 %, respectively). The proportion of collected wastewater being treated also decreases with income level, at 91 %, 73 %, 60 % and 47 % for high-, upper-middle-, lower-middle- and low-income classifications, respectively. The fact that 40 % and 53 % of collected wastewater is untreated in the lower-middle- and low-income classifications, respectively, may also be indicative of the higher prevalence of intentional untreated-wastewater reuse (whereby collected wastewater is reused without undergoing treatment).

High utilisation of treated-wastewater reuse occurs predominantly in the Middle East and North Africa, with the United Arab Emirates, Kuwait and Qatar reusing more than 80 % of their produced wastewater. Water-scarce small island developed countries, including the Cayman Islands, US Virgin Islands and Malta also have high rates of intentional treated-wastewater reuse of 78 %, 75 % and 67 %, respectively. Treated-wastewater reuse is prohibitively low in areas with low wastewater treatment rates, such as sub-Saharan Africa and South Asia. In addition, treated-wastewater reuse is also low in areas with sufficient availability of conventional water resources such as across Scandinavia (where reuse is <5%).

In volumetric flow rate terms, intentional treated-wastewater reuse is estimated to be largest in the East Asia and Pacific region (11.9×109 m3yr-1) and North America (9.1×109 m3yr-1) and lowest in South Asia (0.5×109 m3yr-1) and sub-Saharan Africa (1.6×109 m3yr-1). Conversely the Middle East and North Africa (27.8 %) and western Europe (17.5 %) dominate in percentage terms. In volumetric flow rate units, the Middle East and North Africa (15 %) and western Europe (16 %) account for almost a third of treated-wastewater reuse globally, despite only accounting for 5.8 % and 5.7 % of the global population, respectively. Approximately half (52 %) of intentional treated-wastewater reuse occurs in high-income countries, with 37 % from upper-middle-income countries. Intentional treated-wastewater reuse is contingent upon the availability of treated-wastewater resources, which is typically more prevalent in high-income countries (who both produce more wastewater per capita and treat a higher percentage of the resource). However, the proportion of treated wastewater intentionally reused is higher in the upper-middle- (25 %) and lower-middle-income (25 %) groups than in the high-income group (19 %).

Figure 4 displays gridded wastewater production, collection, treatment and reuse, allowing for the identification of hotspot regions and zones at 5 arcmin resolution. Wastewater production occurs across the globe, with hotspots coinciding with the largest metropolitan areas (e.g. Tokyo and Mumbai) where the largest concentration of domestic and industrial activities occurs (Fig. 4a). In contrast, wastewater production is close to zero in world regions with low concentrations of people and industrial activities, such as the Sahara, inland Australia and the high-latitude climate zones (e.g. northern Canada and Russia). In countries where municipal activities are heavily concentrated in a small number of cities, such as in the Middle East and Australia, small clusters of grid cells with very high wastewater production (>5×106 m3yr-1) occur. Wastewater collection (Fig. 4b) and treatment (Fig. 4c) are typically more concentrated in urban areas within individual countries. This is particularly prominent in South America and sub-Saharan Africa. Conversely, downscaled wastewater collection and treatment reflect wastewater production in regions where wastewater collection and treatment rates are very high, such as western Europe and Scandinavia. Wastewater reuse is constrained to the lowest area (number of grid cells), being concentrated in regions where treated-wastewater resources are available and where water scarcity issues are of particular concern.

Figure 5a displays the global distribution of the wastewater treatment plants and designated wastewater reuse sites considered in this study. Plant capacities were compared to downscaled quantifications for validation of wastewater treatment (Fig. 5b) and wastewater reuse (Fig. 5c). Overall, a reasonable performance is obtained at most wastewater treatment and reuse plants with linear R2 values of 0.57 (p<0.001) and 0.50 (p<0.001), respectively. The observed negative normalised biases suggest that downscaled wastewater treatment (-0.32) and reuse (-0.51) were underestimated with respect to the observed treatment capacities. This may occur due to discrepancies between the design (i.e. maximum) capacity of wastewater treatment plants, which is commonly the capacity that is reported, versus the actual treated-wastewater volumes. Factors such as the construction year of wastewater treatment plant are important, as plants are constructed to be larger than current requirements in anticipation of future increases in wastewater flows. Furthermore, uncertainties in the data used as basis for downscaling wastewater production (i.e. PCR-GLOBWB return flows) directly impacts the downscaled results of wastewater treatment. For example, the underprediction of return flows in urban areas and overprediction in rural areas could lead to the overprediction of wastewater treatment in areas without treatment plants and underprediction of wastewater treatment for grid cells with large treatment capacities.