Prospective water scarcity footprint under climate change applied to bio-based sustainable aviation fuel production pathways

Summary

The transition of aviation from fossil to sustainable aviation fuels (SAFs) raises concerns about its water requirements in an increasingly water-stressed world, highlighting the need for integrated water footprint assessments under climate change. This study first develops novel water scarcity footprint factors projected monthly and annually until 2099 at a 0.5° × 0.5° global resolution under eight climate change scenarios. Significant regional and temporal disparities are revealed, with future factors differing by more than 50% from historical values in many regions. Applying these dynamic factors to projected global SAF production shows that Asia contributes to more than 50% of the global future water scarcity footprint, while North America exhibits lower impacts despite high production volumes. By demonstrating that historical factors underestimate future impacts, particularly under high-emission scenarios, these findings emphasize the importance of prospective environmental assessments to ensure that energy transition does not endanger water security.

Graphical abstract

Highlights

• •Global annual and monthly water scarcity footprint factors modeled up to 2099
• •Insights on prospective water impacts of sustainable aviation fuel production
• •Supports strategic planning by identifying future water risks in energy transitions

Earth sciences; Environmental science; Geochemistry

Introduction

The energy-water nexus has sparked a long debate over sustainability.¹ Energy production requires water, and energy is essential for water processes. From electricity to fossil fuels and biofuels, the energy sector accounts for around 10% of the global freshwater use.² The transition toward non-fossil energy, coupled with the growing energy demand, may put significant pressure on freshwater availability when energy pathways are water intensive, such as those based on biomass. When freshwater availability is insufficient to meet demand, water scarcity emerges as a challenge to the progress toward achieving the UN SDG6, which aims to ensure clean water for all.³ As countries transition to low-carbon energy systems, accurately assessing the water intensity of these new technologies becomes essential. This is necessary to maintain a sustainable balance between energy production and water security.

Among these technologies, sustainable aviation fuel (SAF) is emerging as an alternative energy solution to reduce the aviation sector’s carbon footprint. Notably, SAF production doubled from 2022 to 2023, reaching over 600 million liters. In addition, 65% of the emissions reductions required for the aviation sector to achieve net zero CO₂ emissions by 2050 are expected to come from SAF.⁴ However, lower carbon technologies do not necessarily require less water. While some lower carbon energy pathways such as wind and solar photovoltaic (PV) require very little water, decarbonization pathways that rely strongly on biofuel production can even exacerbate water availability.⁵ Therefore, as the SAF industry grows in response to climate change policies, its water footprint under future conditions should be carefully assessed.

When a new technology is being designed and employed, prospective environmental studies, such as life cycle assessments (LCAs), are typically performed to support decisions toward large-scale deployment. In this context, projected production levels, expected process characteristics, and forward-looking background datasets are commonly used. The early assessment of environmental impacts helps to optimize processes, identify trade-offs, and align relevant policies with sustainability goals. For this reason, multiple water footprint methods, such as the water stress index (WSI),⁶ the available water remaining (AWARE),⁷ and the withdrawal-to-availability (WTA),⁸ have been proposed to quantify water impacts.

One major weakness of the existing water footprint impact methods is the lack of consideration of spatiotemporal aspects in the characterization factors, such as geographic variability and progressing climate change. These factors are used to quantify the potential impact of water consumption but rely mostly on average historical data aggregated at watershed or national levels. In particular, subnational factors have been proposed,⁹ while local water data have been used to develop regionalized factors in Thailand¹⁰ and Brazil.¹¹ However, these factors are calculated as average annual/monthly datasets over a historical period, ignoring the future variability.

Water availability is dynamic though; it fluctuates across different regions and seasons.⁵ Climate change is expected to affect global water ecosystems through changes in seasonal hydrological processes and extreme events.^12,13,14 Some regions will benefit with increased water availability,⁵ while, on the contrary, it is expected that 3.9 billion people will live in regions with water scarcity by 2050.¹⁵ Therefore, climate change adds a critical temporal aspect that is usually overlooked in common water assessments. Reliance on historical averages fails to capture these temporal shifts. This can lead to an inaccurate water impact assessment of emerging energy systems in the future.

This limitation becomes particularly apparent when trying to assess the future water scarcity footprint (WSF) of SAF production. As part of a broader sustainability analysis, it is found that existing characterization factors are based on average hydrological conditions and are not designed to account for future changes in water availability under climate change. Pfister et al.¹⁶ calculated WSI factors in 2050 based on the IPCC A1B scenario¹⁷ to estimate the water footprint of agriculture for different food supply strategies. Similarly, AWARE factors for 2050 have been developed by Baustert et al.¹⁸ at a country and basin level under two scenarios, which were applied on a water desalination plant in Spain. Both studies identified significant differences in future water scarcity. However, their results are restricted to limited future scenarios and coarse spatial and temporal resolutions. Large-scale studies are needed to produce more consistent results. Such studies should integrate both the temporal dynamics and spatial variability of climate change and water resources. This would provide a better understanding of how climate change and shifting freshwater availability impact energy systems globally. This gap presented a major obstacle to accurately evaluating SAF’s future environmental performance. It also reflected a broader challenge in prospective LCAs: the lack of spatiotemporally dynamic water characterization factors that reflect evolving environmental changes.

To address this gap, this study develops spatiotemporally resolved WSF characterization factors, hereby indicated as CFs, based on projections from 2024 up to 2099 for eight climate change scenarios,^19,20 offering a broader perspective on climate change impact compared to previous studies. Mean factors are also estimated based on historical data for the period of 1861–2005 for comparison. The final datasets are provided at a monthly and annual step with a 0.5° × 0.5° spatial resolution and global coverage. Given the applicability of these factors, aggregated results at basin and country level are also generated, to emphasize the critical role of spatial resolution.

With both SAF and water needs set to increase, it is essential to understand the impact of SAF production on water scarcity in the future, which remains currently unexplored. This study addresses the challenge of estimating and understanding future water impacts of such technologies. It focuses specifically on the growing role of bio-based sustainable aviation fuel and how its WSF may evolve under climate change. By unraveling the linkages between the two sectors, two main objectives are covered: (1) the development of spatiotemporally resolved WSF CFs under multiple climate change scenarios in the future and (2) their application to estimate the WSF of projected global SAF production over the short-term future.

The SAF industry case study directly motivates the methodological development and serves to illustrate the practical importance of dynamic CFs for decision-making. In doing so, this study reveals how reliance on historical water data can significantly underestimate future water stress, especially under high-emission scenarios. Moreover, future water stress points are revealed. These provide a spatially explicit overview of where and when SAF-related water scarcity risks may intensify. Such insights can help with the adoption of sustainable practices and policies, ensuring that SAF production not only supports energy transition but also aligns with global water conservation goals. Finally, the importance of spatiotemporal dimensions under various climate change scenarios in the environmental assessments of emerging energy technologies is highlighted as a key to accurately evaluate the future environmental footprint and address associated risks.

Results

Spatiotemporal WSF factors

The spatiotemporal variation of WSF CFs, calculated at a 0.5° × 0.5° resolution, reveals significant disparities across the globe. First, factors over the historical period are presented in Figure 1, calculated for four different general circulation models (GCMs), using historical climate and CO₂ concentrations.¹⁹ The calculated mean factors over the historical period indicate that regions such as North and South Africa, the Middle East, parts of Central Asia and Australia, and western parts of the United States have consistently experienced high water scarcity stress (Figure 1). Indeed, these observations align with findings in a relevant study by Mekonnen and Hoekstra,²¹ who performed a similar analysis on water scarcity based on a 10-year historical period and identified these regions as water stress hotspots. These areas are mostly characterized by high population density, intense irrigation, and arid climate.^21,22,23 The local variations reveal how water scarcity factors vary within the same countries (i.e., large factors in western India compared to the northeast), highlighting the importance of high spatial resolution.

Figure 1.

Mean water scarcity footprint factors CF (m³ world eq./m³) over the historical period (1861–2005) for four different general circulation models

(A) GFDL-ESM2M.

(B) HadGEM2-ES.

(C) IPSL-CM5A-LR.

(D) MIROC5.

Maps are projected at EPSG:4326, with a global 30 arc-min resolution (excluding Antarctica). No data are available for gray-colored grid cells.

Except for the spatial variability, which is well documented in literature, novel insights are provided in the interannual variability under climate change scenarios. Figure 2 shows the global distribution of seasonal WSF factors CF during 2024 under the RCP 2.6-IPSL-CM5A-LR scenario (see Figure S1 for the RCP 6.0-IPSL-CM5A-LR scenario), highlighting a clear intra-annual variability. Distinct seasonal hotspots are evident, for example during the summer months (June to August) in south Europe and during December to March in Siberia, where in general higher CFs are observed. The complete dataset of monthly WSF factors CFs for all years (2024–2099) and all eight climate change scenarios is available in the supplemental information and the associated online repository.

Figure 2.

Seasonal water scarcity footprint factors CF (m³ world eq./m³) during 2024 for the RCP 2.6-IPSL-CM5A-LR scenario

Results are presented as seasonal mean over December, January, February (A), March, April, May (B), June, July, August (C), September, October, and November (D).

Maps are projected at EPSG:4326, with a global 30 arc-min resolution (excluding Antarctica). No data are available for gray-colored grid cells.

Besides the seasonal variability, notable differences are observed between the historical period and 2024 (beginning of the future analysis) and 2099 (end of the future analysis) (Figures S2 and S3). A comparison between the historical period and 2024 indicates significant changes (over ±50% in both directions) in most regions under the RCP 2.6 scenarios, as shown in the first column of Figure 3. The CFs have notably increased in Central and South Europe, Eastern USA, the Middle East, Western Africa, and parts of South Asia. By 2099, the factors continue to indicate a significant increase in most of the aforementioned areas (see second column of Figure 3). On the contrary, there are many areas where factors remain stable across years or even notably decrease, due to increased water availability, associated with high precipitation (e.g., monsoons in tropical climates) and glacier melting (e.g., high northern latitudes).²⁴ Similar observations are made for the RCP 6.0 scenarios, as shown in Figure S4.

Figure 3.

Water scarcity footprint factor (CF) difference (expressed as percentage) for the low GHG concentration

RCP 2.6 scenario: between the historical period and 2024 (first column) and between 2099 and 2024 (second column). Results are shown for four different GCMs. Negative difference indicates a decrease and positive difference an increase of water scarcity in the future. Maps are projected at EPSG:4326, with a global 30 arc-min resolution (excluding Antarctica). No data are available for gray-colored grid cells.

Variation across the four different GCMs highlights the uncertainty of such projections, particularly in regions where water availability is highly sensitive to climate conditions (Figure 4). Indeed, the variability is much higher for the future projections in 2024 and 2099, being the highest in water-stressed regions, such as northern and southern Africa, Australia, the Middle East, Central Asia, western USA, and southern America. To quantify the uncertainty arising from different GCMs, statistics of the inter-model range in the CFs were calculated globally (Table S1). The mean inter-model range increased from 5.1 m³ world eq./m³ during the historical period to 15.6 m³ world eq./m³ in 2024 and 16.3 m³ world eq./m³ in 2099, alongside a notable rise in the 90th percentile and maximum range values in both future time points. This substantial spread demonstrates that uncertainty in future water scarcity estimates grows significantly over time, especially in already stressed regions. Moreover, these findings highlight the importance of including multiple GCMs in prospective impact assessments to capture a wide range of potential outcomes.

Figure 4.

Range of the water scarcity footprint factor across four different GCMs (calculated as the maximum minus the minimum value) covering

(A) the historical period, (B) 2024, and (C) 2099 for the RCP 2.6 scenario.

Notable differences are also revealed by comparing the two future climate-related forcing scenarios with the historical period (Figure S5). This figure illustrates that, compared to RCP 2.6, RCP 6.0 shifts the distribution toward higher intensity changes, thereby revealing critical emerging hotspots. Smaller increases or decreases in the future factors are observed globally for the RCP 2.6 scenario when compared to the historical data. In contrast, the RCP 6.0 scenario exhibits more pronounced changes with a larger share of grid cells experiencing increases above 50%–70%, indicating either a significant exacerbation or improvement in the regional water stress. It is thus evident that this scenario reveals the areas of concern that were previously underrepresented, highlighting the need to include climate change scenarios in prospective water scarcity studies.

WSF of future SAF production

Bio-based SAF production is expected to grow significantly globally in the next few years, though not all regions will follow this trend due to differences in policies, resources, and technological advancements (Figure 5A). Compared to 2024 levels, global SAF production is projected to increase substantially over the period 2025–2030, primarily driven by binding policy mandates and national targets.²⁵ In the European Union, the ReFuelEU Aviation,²⁶ as part of the “Fit for 55” package, establishes mandatory SAF blending requirements starting in 2025, with progressive targets set up to 2030 and 2050. Similarly, the United Kingdom has introduced a SAF Mandate beginning in 2025,²⁷ while in the United States, the Sustainable Aviation Fuel Grand Challenge sets annual production targets through 2030.²⁸ In total, global SAF production is expected to increase from the 2024 level of approximately 2.3 bn L to about 22 bn L by 2027, further rising to approximately 29 bn L by 2030.²⁵ The United States, Canada, China, and Brazil are expected to be major players in the SAF industry. In particular, SAF production in North America is expected to reach more than 10 million tonnes per year (Figure S6). Asia and Europe are following, with Asia seeing a notably rapid expansion. Alcohol-to-ethanol (ATJ) and Fischer-Tropsch (FT) pathways are also expected to become commercially available over the next years, compared to the currently employed hydroprocessed esters and fatty acids (HEFA) process.

Figure 5.

SAF production capacities and their water scarcity footprint globally

(A) SAF production (ktonne) in 2030 based on own calculations from Argus Media data.²⁵ Three different production pathways are covered: HEFA, ATJ, and FT. Map insets show the logarithmic WSF in 2024, 2027, and 2030 for three countries: China, Spain, and United States. Values calculated for four different GCMs (yellow for GFDL-ESM2M, red for HadGEM2-ES, purple for IPSL-CM5A-LR, and green for MIROC5), using RCP 2.6 (solid line), RCP 6.0 (round dot line), and historical (dash dot line) basin-level CFs.

(B) Water scarcity footprint (WSF) per continent under two RCP scenarios across four GCMs in 2024, 2027, and 2030. Results are calculated using basin-level CFs. Table below shows the contribution of each continent to the global SAF production per year of analysis (expressed as mass percentage).

The WSF of SAF production is found to significantly vary across different countries and time periods (see map insets in Figure 5A). Indeed, the WSF ranges from values lower than 1 × 10⁻⁴ million L_eq/tonne SAF to over 1 × 10⁻¹ million L_eq/tonne SAF in highly water-stressed areas (Figures S6 and S7), such as the USA, India, Pakistan, Spain, and Australia. In contrast, northern countries such as Canada, as well as countries located in tropical climates such as Brazil and Thailand, show lower WSF values due to more abundant water resources and lower factors (Figure 1). Detailed basin- and country-level factors are available in supplemental information.

Map insets in Figure 5A provide a comparative analysis of the WSF in three major SAF producers already present in the industry, i.e., USA, China, and Spain. The results show the variance (as order of magnitude) between the historical data (HIST) and climate scenarios RCP 2.6 and RCP 6.0, evaluated through four GCMs based on basin-level factors. While the WSF starts at relatively similar levels across scenarios in Spain, the trend changes as time advances. Under the RCP 2.6 scenario, a general increase is observed in 2030 compared to 2024 for all GCMs, while the RCP 6.0 scenario presents a more variable outcome, with more evident peaks and valleys across years. This difference is even more noticeable in China, while, on the contrary, the USA shows a steadier WSF profile over the years, with similar trends between the RCP scenarios.

The comparison of these map insets (Figure 5A) highlights the pronounced difference between estimating impacts in the future using historical and projected factors. All inset graphs in Figure 5A show that the WSF based on historical factors is underestimated, compared to the RCP projected factors. For instance, the WSF of China differs by 1° of magnitude between the historical and future factors. The variability in future scenarios, particularly under RCP 6.0, suggests more pronounced future water scarcity impacts due to climate change. Thus, the historical factors fail to capture anticipated changes in climate variability, leading to higher and more varied WSF projections.

Asia consistently contributes the largest share to the global WSF under all scenarios (more than 50% in the future) (Figure 5B), reflecting its dominant role in SAF production water impacts, due to high production capacities and water-stressed regions in the continent. Europe is the second largest contributor, while North America, despite being the largest SAF producer in the future, exhibits a much lower WSF. A steady increase of the WSF over time is generally observed for both RCP scenarios, driven by both SAF demand and exacerbating water scarcity conditions in the corresponding areas. The RCP 6.0 scenario shows in general higher values across all models, indicating more pronounced impacts.

In order to investigate the robustness of results to uncertainties in operational water consumption, a Monte Carlo simulation was performed with 5,000 iterations, applying a ±10% triangular distribution around the baseline water use for HEFA, ATJ, and FT pathways. This uncertainty range is based on observed multi-year changes in industrial water intensity across various countries and sectors.^29,30,31,32 A triangular distribution was chosen for the Monte Carlo simulations because the empirical distributions of the key inputs are unknown, but credible minimum, most-likely, and maximum values can be specified.^33,34,35 This provides a clear and bounded representation of uncertainty without requiring unwarranted assumptions about the distribution’s shape. This approach is common in risk and environmental modeling when data are limited.³⁶ Based on the resulting distributions, the mean, 5th–95th percentile range, and minimum to maximum range for each case were calculated (see Spreadsheet in supplemental information). Figure 6 presents the relative deviations from the base case for the WSF across continents under eight climate change scenarios for 2024, 2027, and 2030. The boxplots consist of boxes representing the 5th–95th percentile range and whiskers representing the minimum to maximum values. Across all scenarios, deviations remain relatively small, with interquartile ranges generally within ±5%, confirming the robustness of the results.

Figure 6.

Boxplots of Monte Carlo-derived relative deviations of water scarcity footprint from the baseline (boxes: 5th–95th percentiles, whiskers: min–max)

Results are shown per continent and year of analysis, for eight different climate change scenarios.

Except for the temporal variance of the results, attention should be given to the spatial resolution of the applied factors. The difference on the calculated WSF for each country based on country- and basin-level factors is presented in Figure 7, with a variance higher than ±50% being observed for most countries. For example, in regions like South Asia, significant positive differences indicate that basin-level factors reveal higher water scarcity impacts compared to country-level factors, while an opposite trend is observed for other regions, such as parts of Europe and Northern America. This suggests that country-level aggregations may underestimate local water stress hotspots or overestimate impacts in certain basins. The importance of using accurate spatial scales in accordance with the available data resolution is highlighted, allowing for more accurate regional water footprint assessments.

Figure 7.

Difference (expressed as percentage) between the WSF of SAF production calculated using country- and basin-level aggregated CFs

Results are presented for the two RCP scenarios and for four GCMs in 2030. Positive difference indicates higher WSF for basin-level CFs.

Discussion

Development of spatiotemporal WSF factors

This study provides critical insights into the spatiotemporal variability of WSF factors (CFs) and their implications for future SAF production. Unlike previous studies that mostly rely on historical or static water scarcity factors at coarser resolutions,^37,38,39 such as case studies using local historical data in the USA,^40,41 Thailand,¹⁰ Peru,⁴² Brazil,¹¹ Australia,⁴³ and China,^44,45 this work explicitly integrates future climate projections at high global spatial and temporal resolution. This approach offers a more realistic basis for global assessments, as the results show clear differences across regions and time horizons. These findings highlight the importance of considering both spatial and temporal variability when assessing the water footprint of emerging energy technologies.

A key novelty of this work lies in the development of high-resolution CFs at a 0.5° × 0.5° resolution, monthly and annually projected up to 2099 under eight climate change scenarios, representing a significant advancement over existing assessments. Although annual averages are useful for applicability and consistency in environmental assessments, the availability of monthly CFs provides further value by capturing intra-annual water stress dynamics. These are especially critical in regions where water stress might occur only a few months per year, similarly to what has been observed by Mekonnen et al.²¹ By incorporating projected climate data, this approach contributes to including future climate variability in water scarcity assessments. The developed CFs are applicable to various emerging energy technologies beyond SAF, particularly relevant to water-intensive ones, such as hydrogen production. The full set of monthly and annual CFs at a fine resolution is available as open datasets to support future applications.

Prospective WSF of global SAF production

As SAF production is scaling up globally, novel insights on its water footprint evolution are revealed, with a substantial variability both across space and time. The analysis shows that Asia is the largest contributor to the WSF of global SAF production, accounting for over 50% by 2030, followed by Europe at around 20%. Although North America has the third highest contribution to the global WSF, it still exhibits relatively lower values despite being the biggest SAF producer by 2030. A previous study on the water footprint of road transport biofuel production by Gerbens-Leenes et al.⁴⁶ projected that the USA, China, and Brazil would dominate the global water footprint by 2030. This was largely attributed to their reliance on first-generation crop-based biofuels. In contrast, the present study finds that the WSF for SAF pathways shifts the burdens away from traditional agricultural hotspots, like the USA. By incorporating climate change scenarios, new emerging hotspots (such as in Europe) are revealed. These findings suggest that the geographic distribution of SAF production facilities has a substantial influence on the overall WSF of the industry, while each technology must be carefully assessed in relation to regional water availability and future climate risks.

Another key contribution of this study lies in demonstrating that reliance on historical data can underestimate the WSF of SAF production. While factors based on historical data are commonly used for simplicity and availability, they fail to account for the dynamic nature of water availability driven by climate change. Similar concerns over the neglection of temporal dimensions in environmental assessments have also been raised by Gheewala et al.⁴⁷ and Li et al.⁴⁸ This is particularly relevant for areas expected to experience significant changes, such as regions in China. This underestimation can lead to misguided decisions in SAF production planning, with investments taking place in potential water-stressed regions in the future.

While temporal dynamics of water scarcity highlight the importance of considering climate change scenarios, spatial resolution is equally critical in water footprint assessments. The comparison between the basin- and country-level analyses in this work allows for a better understanding of local water stress hotspots. Aggregated data at coarser spatial resolution can either lead to significant underestimation or overestimation of water impacts, as also evidenced by earlier works.^6,9,49 Higher resolution data can capture finer geographic characteristics, which are essential for accurately identifying water scarcity hotspots at local levels. For example, the location and water source of each SAF production facility are crucial, as the WSF can be notably reduced in less water-stressed regions.

Limitations of the study

Country- and basin-level CFs were derived using a human water consumption-weighted aggregation, assuming uniform distribution within each grid cell. While it ensures consistency with the AWARE regionalization methodology,⁹ it overlooks spatial clustering of sectoral water consumption, such as industrial zones, population density, and irrigated agricultural areas. Future work could improve spatial accuracy by incorporating high-resolution, sector-specific water use data when they become available at a high spatial and prospective temporal resolution.

Regarding calculation assumptions, this study uses the natural runoff instead of pristine flow to calculate the environmental water requirements (EWRs), following similar practices in recent extensions of the AWARE methodology when pristine flow data are unavailable.⁴¹ This substitution is not expected to introduce large errors because EWRs are defined relative to seasonal flow regimes rather than absolute values. The classification of low-, intermediate-, and high-flow months depends on relative deviations from the mean annual flow, making the method less sensitive to systematic offsets between pristine and natural runoff. At the global scale and 0.5° spatial resolution considered here, this proxy provides a consistent upper-bound representation of water availability in the absence of explicit human water consumption. The resulting uncertainty is likely smaller than uncertainties arising from climate model variability and water demand projections. Therefore, it is not expected to affect the observed trends, scenario comparisons, or main conclusions of this study.

The selection of spatial resolution depends on the purpose of the analysis, the availability of data, and the scale of the assessment. Higher resolution is needed for more localized assessments, such as evaluating the specific impact of a production facility. Coarser resolutions may be sufficient when detailed data are unavailable. At the same time, the annual averages in the main analysis ensure comparability across scenarios and provide a consistent basis for reporting at the country and continental level.

The development of both annual and monthly CF datasets constitutes a key methodological contribution. The annual CFs reveal long-term trends and interannual changes in water scarcity driven by climate change, while the monthly CFs capture also the intra-annual variability due to the hydrological seasonality. Together, these datasets provide complementary insights into how water scarcity evolves over time in the future, both across years and within individual years. The choice of temporal resolution in the application of these datasets depends on the purpose and scope of the assessment, as well as on the temporal resolution of the available inventory data. For the specific case of SAF production, the monthly attribution of water use impacts remains challenging, as production levels and feedstock supply chains are not yet established with sufficient temporal detail to reliably link to monthly water availability. Nevertheless, the use of annual CFs in the SAF case study does not diminish the relevance of the monthly CFs. On the contrary, the availability of both annual and monthly prospective CFs highlights the flexibility of the proposed datasets and their applicability beyond the present case study.

While the developed CF datasets extend until 2099, the application to SAF production is limited to 2030. The choice to generate CFs up to 2099 is data driven, as the climate and hydrological input datasets used in this study provide projections up to that year. Given the availability of these inputs and the long-term impact of climate change in water scarcity, it is considered methodologically appropriate to fully propagate these projections through the CF calculations rather than limiting the analysis to an earlier date. However, as reliable spatially explicit projections of SAF deployment beyond the near term are not yet available, particularly regarding plant locations and technology choices, the case study was limited to 2030. As more detailed spatial and temporal data on SAF production become available, CFs at a finer resolution can be applied. Beyond SAF production, the generated CFs can also be used for other applications, offering greater accuracy in water impact assessments across various energy sectors.

The use of four different GCMs and two RCPs provides a comprehensive view of potential future conditions, highlighting the uncertainty inherent in climate projections. Under RCP 6.0, more pronounced changes in WSF factors are observed. These include both exacerbated stress in certain areas and improved conditions in others due to changes in water runoff patterns and water availability. Similarly, large differences are identified across GCMs, with water-stressed areas being more sensitive to the model choice. These observations stress the complexity of predicting future trends, as different models can yield divergent outcomes based on underlying assumptions. Despite this uncertainty, which is inherent with the assumptions and limitations of the used models, the application of multiple future scenarios helps capture a broader range of possible futures. Although these differences between climate scenarios appear moderate for SAF (Figure 6), probably due to the projected concentration of facilities in less climate-sensitive regions, the developed CFs have broader applicability. For other water-intensive technologies, such as hydrogen production, climate-driven differences across scenarios may be more pronounced, further emphasizing the need to integrate climate variability into water scarcity assessments.

The studied water footprint refers only to the SAF production stage, covering ATJ, HEFA, and FT pathways and excluding feedstock supply due to the lack of highly resolved data for each facility. However, HEFA jet fuel dominates future supply, accounting for 80%–98% of global SAF production, with used cooking oils and animal fats mainly used and expected to continue being used as feedstock.⁵⁰ Such feedstocks are classified as waste, with minor water consumption and land use change impacts. Water consumption during the feedstock supply is attributed solely to their collection and transportation, which is negligible.⁵¹ On the other hand, ATJ jet fuels are mainly produced from agricultural biomass, which can significantly increase their water footprint. For example, approximately 6% of the water consumption of SAF production from corn crops is attributed to the production stage, with the rest almost completely deriving from the corn cultivation stage.⁵² When second-generation feedstocks are used, such as corn stover, the cultivation stage impact becomes more complicated, as this depends on the allocation method chosen. Moreover, transparent and high-resolution datasets linking specific feedstocks, their origins, and irrigation demands to SAF facilities are not currently available. Nevertheless, as governments and policymakers accelerate efforts to meet future climate targets, SAF producers are expected to shift toward utilizing less water-intensive feedstocks, such as agricultural residues and waste.⁵³ As a result, feedstock-related water scarcity may remain limited in most cases. It is also important to note that the water consumption estimates for SAF production are specific to the production phase and can be sensitive to several variables, such as variations in feedstock quality or processing efficiencies.

Therefore, the exclusion of the feedstock stage in the present study was deemed appropriate, based on both the characteristics of the dominant SAF production pathways and limitations in data availability and resolution. In this context, the operational stage provides the most robust and comparable system for assessing the prospective spatiotemporal evolution of the SAF-related WSF. Nevertheless, the results presented here should be interpreted as the operational WSF of SAF production facilities. As a consequence, the reported WSF values represent a lower bound estimate of the total life cycle WSF, particularly for pathways relying on irrigated biomass feedstocks, such as ATJ and FT. Future work could incorporate upstream water consumption, covering a full LCA, when transparent and spatially explicit datasets become available.

Moreover, the operational water consumption for each SAF production pathway was assumed to remain constant over the study period (2024–2030). While technological innovations may reduce water use in the longer term, their timing and scale remain uncertain. Given the relatively short time frame, the main production processes and water use are unlikely to change substantially. Additionally, no robust and pathway-specific projections exist for future water efficiency improvements. To address this potential variability, a Monte Carlo analysis was performed, with results in Figure 7 indicating a small variability. This indicates that the overall trends are robust to variations in operational water consumption, while the use of constant water consumption factors ensures consistency.

Since this study focuses only on bio-based SAF, non-biogenic routes such as power-to-liquid (PtL) are excluded. Among the available bio-based routes, HEFA, ATJ, and FT are included as they represent the majority of existing and planned global production capacity, they are among the most mature technologies, and water consumption data for these pathways are available in the literature. Co-processing technologies, which blend bio-based and fossil feedstocks in conventional petroleum refineries, were also excluded. Attributing water impacts specifically to the bio-based fraction is methodologically complex, while the production volumes remain unknown in most facilities. Nevertheless, co-processing facilities are expected to contribute only up to 4% of the total global SAF production until 2030.²⁵ Other bio-based routes (e.g., pyrolysis and hydrothermal liquefaction) are currently at low technology readiness level (TRL) and lack consistent data for inclusion in a robust analysis.⁵⁴ As these technologies develop and more data become available, future analyses could include them.

Even if the long-term future outlook of SAF production is uncertain, this study provides valuable insights into the potential water footprint and how it may vary in a changing climate. Despite the well-documented water-stressed areas, some SAF facilities are still planned in such regions. The fact that water basins can intersect multiple political boundaries places a significant strain on water resource management policies.⁵⁵ The generation of such high-resolution data based on a large-scale integrated assessment is crucial for informing local policies and strategies that support the sustainable development of SAF in the future. In addition to the WSF, expressed as water quantity, water quality should also be incorporated in future work.⁵⁶ Factors such as nutrient concentration, presence of contaminants, and overall water pollution levels can provide a more holistic evaluation of water stress.

Resource availability

Lead contact

Further information and requests for resources should be directed to and will be fulfilled by the lead contact, Konstantina Vasilakou (konstantina.vasilakou@uantwerpen.be).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• •The generated WSF factors (CFs) have been deposited at Figshare at https://doi.org/10.6084/m9.figshare.27901818 and are publicly available as of the date of publication.
• •All original code has been deposited at GitHub at https://github.com/konnavasil/Water-scarcity-footprint.git and is publicly available as of the date of publication.
• •Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

Acknowledgments

This study was carried out within the framework of the ADV_BIO project financed by the FOD Economie - Energietransitiefonds/SPF Économie - Fonds de Transition Energétique, call 2019–2020 subsidies. P.N. holds an FWO senior postdoctoral fellowship (grant no. 1215523N) granted by FWO Vlaanderen/Research Foundation Flanders. This research also received support from the Special Research Fund Methusalem project NANOlight of the University of Antwerp.

Author contributions

K.V., writing – original draft, writing – review and editing, conceptualization, methodology, software, data curation, formal analysis, validation, and visualization; P.N., writing – review and editing, validation, and supervision; P.B., writing – review and editing, validation, resources, and supervision; S.V.P., writing – review and editing, validation, resources, and supervision.

Declaration of interests

The authors declare no competing interests.

STAR★Methods

Key resources table

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Deposited data
Daily runoff	Inter-Sectoral Impact Model Intercomparison Project (ISIMIP2b)	https://doi.org/10.48364/ISIMIP.626689
Monthly human water consumption	Inter-Sectoral Impact Model Intercomparison Project (ISIMIP2b)	https://doi.org/10.48364/ISIMIP.626689
SAF production facilities	Argus Media	https://www.argusmedia.com/en/news-and-insights/topical-market-themes/sustainable-aviation-fuels-saf
Country borders	Natural Earth	https://www.naturalearthdata.com/
Sub-basins	WaterGAP Input Data of AWARE method	https://wulca-waterlca.org/aware/input-data-watergap/
Water scarcity footprint factors	This paper	https://doi.org/10.6084/m9.figshare.27901818
Software and algorithms
Code for methodology (Python 3.11.5)	This paper	https://github.com/konnavasil/Water-scarcity-footprint.git

Experimental model and study participant details

Omitted as our study does not involve biological models.

Method details

This section is organized into five subsections that follow the analytical workflow of this study. First, the subsection “water scarcity footprint factors” introduces the AWARE framework and defines the water scarcity footprint characterization factors (CFs) along with assumptions. Then, the subsection “water data” describes the hydrological input data and climate change scenarios used. Third, the “spatiotemporal water scarcity footprint factors calculation” describes in detail the calculation of monthly and annual CFs (as defined in the first subsection), and their spatial aggregation to basin and country levels. The last two sections correspond to the second part of this study: the application of the developed CFs to the global SAF production. The “SAF production facilities” subsection describes the chosen case studies (along with selection criteria) of bio-based SAF production. Finally, the “water scarcity footprint of SAF production” explains how the developed CFs are applied to estimate the prospective water scarcity footprint (WSF) of SAF production under different climate scenarios and time horizons.

Water scarcity footprint factors

In this study, water refers to freshwater (blue water), due to its importance to living species and ecosystems, its competition with other sectors and its relevance to biorefinery processes.^57,58 The water scarcity footprint characterization factors (CFs) are calculated following the available water remaining (AWARE), which is commonly used as a representative blue water scarcity footprint indicator in life cycle assessments.⁷ This indicator derives from a consensus model proposed by the Water Use in Life Cycle Assessment (WULCA) working group of the UNEP-SETAC Life Cycle Initiative and is also used by the Environmental Footprint Impact Assessment Method of the European Commission.⁵⁹ It quantifies the spatial and temporal water stress by measuring the amount of available water remaining (AMD) in a given location after accounting for both human and environmental water demands. This locally available water is then compared to the global average (weighted by human water consumption) to produce a dimensionless factor. A higher factor indicates greater scarcity, i.e., less water available relative to demand.

The factors ${CF}_i^t$ are expressed as the available water remaining (AMD) per unit of surface in grid cell i at a given time t⁷:

$${CF}_i^t=\frac{{AMD}_{world}^t}{{AMD}_i^t}$$

Where ${AMD}_{world}^t$ is the consumption-weighted average AMD over the world at a given time t and ${AMD}_i^t$ is the available water of grid cell i at a given time t. The ${AMD}_{world}^t$ is calculated by averaging the AMD values across all grid cells, weighted by the human water consumption in each respective grid cell, following the AWARE methodology. The availability is quantified as the natural runoff, while water demand covers both human and environmental requirements⁷ as follows:

$${AMD}_i^t={Naturalrunoff}_i^t-{HWC}_i^t-{EWR}_i^t$$

Where ${HWC}_i^t$ is the human water consumption in grid cell i at a given time t and ${EWR}_i^t$ is the environmental water requirement to sustain freshwater ecosystems in grid cell i at a given time t. Human demand is taken as human water consumption (water withdrawal not returned to the water source), as water withdrawals discharged back to the same watershed do not have a significant impact on the local water scarcity.⁷ Human water consumption covers water use for irrigation, domestic use, manufacturing, livestock and electricity.

Following the AWARE methodology, ${CF}_i^t$ are subject to cut off values, ranging between 0.1 and 100.⁷ If water demand is higher than water availability or the ${AMD}_i^t$ is lower than 1% of the ${AMD}_{world}^t$, then the CF becomes maximum. On the other hand, if the ${AMD}_i^t$ is 10 times higher than the ${AMD}_{world}^t$, then the CFs become minimum. The ${CF}_i^t$ is a dimensionless indicator representing the relative water scarcity footprint of a specific grid cell i at a given time t, expressed as m³ world eq/m³. This factor reflects a relative value of impact score of water consumption, as m³eq refers to the equivalent cubic meters of world-average. Therefore, a characterization factor equal to 1 corresponds to a region with the same amount of remaining water per area as the global average.

Water data

The required water data, i.e., runoff and water consumption, were taken from the Inter-sectorial Impact Model Intercomparison Project.¹⁹ The ISIMIP2b global water simulations were used for different climate change scenarios, by using four general circulation models (GCM) (taken from the Coupled Model Intercomparison Project Phase 5 (CMIP5)). To capture the complexity of Earth’s climate system and its response to rising greenhouse gas concentrations, this study uses four General Circulation Models (GCMs), i.e., GFDL-ESM2M, HadGEM2-ES, IPSL-CM5A-LR and MIROC5, included in the 5th IPCC Assessment Report. GCMs simulate physical processes in the atmosphere, ocean, land, and cryosphere, and using multiple models helps account for uncertainties and improve the robustness of climate projections. Three different climate-related forcing scenarios are considered for a historical period and two future periods based on the Representative Concentration Pathways (RCP): (a) historical climate and greenhouse gas concentration for a historical period (1861–2005), (b) low greenhouse gas concentration (RCP 2.6) for a future period (2024–2099), and (c) high greenhouse gas concentration (RCP 6.0) for a future period (2024–2099). In total, eight future scenarios are included, according to the Inter-Sectoral Impact Model Intercomparison Project (ISIMIP)¹⁹ and the Intergovernmental Panel on Climate Change (IPCC) AR5 projections.²⁰ Details on the chosen scenarios can be found in the ISIMIP2b simulation protocols.⁶⁰

These scenarios were applied on each GCM and output data on the daily runoff and monthly water consumption (hydrological variables) from the WaterGAP2 hydrological model were taken, following the AWARE methodology.^5,7 The two datasets were available at a 0.5° x 0.5° (30 arc-min) spatial resolution.¹⁹ These datasets take into consideration water demand based on the work of Wada et al.⁶¹ In this case, water use includes aggregated data for agriculture (irrigation and livestock), industries and households (domestic) under the Shared Socioeconomic Pathway 2 (SSP2) scenario. The industrial sector includes both manufacturing and energy water use, without distinguishing between different energy sources. As such, biofuel- or SAF-specific water use is not explicitly modeled.

Spatiotemporal water scarcity footprint factors calculation

The runoff datasets provide data on the observed runoff, after considering human interventions on the water cycle (e.g., human consumption). Therefore, the natural runoff was calculated as the sum of the observed runoff and the human water consumption.⁶² The environmental water requirement was calculated based on the monthly flows, according to Pastor et al.⁶³ The hydrological seasons are characterized as low-flow if the mean monthly flow is less than 40% of the mean annual flow, high-flow if the mean monthly flow is more than 80% of the mean annual flow and intermediate-flow if in-between. For high-flow and low-flow months a 30% and 60% of the mean monthly flow respectively is required to sustain the riverine ecosystem, while 45% for intermediate-flown months. The monthly and mean flows are presented as pristine flows, i.e., without any human influence, however due to the lack of such data in the future ISIMIP2b simulations of WaterGAP2, the natural runoff was used instead.⁴¹

All calculations were performed in Python with a monthly step from 1861 to 2005 for the historical period and from 2024 to 2099 for the future period. The historical spatiotemporal CFs are provided as the human water consumption-weighted average over the historical period for each GCM. The future CFs are provided as annual consumption-weighted factors from 2024 till 2099, for a combination of four GCMs and two RCPs. The produced datasets are available monthly and annually at a 0.5^ox0.5° (30 arc-min) spatial resolution with global coverage.

Except for the initial spatial resolution, future CFs were aggregated at a country and sub-basin level based on a human water consumption-weighted aggregation:

$${CF}_{agg}^t=\frac{\sum _i\left({HWC}_i^t·{CF}_i^t\right)}{\sum _i{HWC}_i^t}$$

The original consumption grid cells might overlap between multiple countries/basins. To address this, it was assumed that the water consumption is equally distributed across each native resolution grid cell.⁹ Therefore, the area coverage of each grid cell within a country/basin border was calculated in QGIS software, by intersecting basin/country borders with a 30 arc-min resolution grid vector. This dataset was then used to calculate an area-weighted human water consumption, before the final aggregation. The consumption-weighted aggregation was applied, following the consensus in literature regarding the appropriateness of this approach for spatial aggregation of water scarcity indicators.^9,64 All calculations were performed in Python and the CFs are provided with an annual step from 2024 till 2099 for 4734 basins and 243 countries, across all future climate change scenarios and the historical period. The basin-level aggregation is performed using the sub-basin spatial grid from the WaterGAP2 model in accordance with the AWARE methodology, consisting of a global set of over 10,000 sub-basins.^7,65

SAF production facilities

Worldwide SAF production volumes were taken from the publicly available “Global SAF capacity” map by Argus Media,²⁵ which provides up-to-date data on the global SAF production facilities, currently operating as well as planned in the near future. The dataset covers SAF production from the following technologies: Alcohol-to-jet (ATJ), Hydroprocessed esters and fatty acids (HEFA), Co-processing, Distillation, Fischer-Tropsch (FT), Power-to-liquid (PtL) and other (e.g., Pyrolysis, Methanol-to-jet, Hydrothermal liquefaction). Since only bio-based fuels are investigated, the SAF facilities were filtered based on three major production pathways: HEFA, ATJ and FT.⁶⁶

The location, employed technology, production capacity and operational start year are reported in the Argus Media dataset for each facility.²⁵ The production facilities were grouped per basin/country and production pathways. Only facilities whose production capacities are known, were included in this study. The data were aggregated across three years: 2024, 2027 and 2030. For each country/basin and year, production volume is obtained by summing the capacities of facilities corresponding to that region and operational start year. The planned facilities are assumed to operate at full capacity, one year after the stated start year.

Water scarcity footprint of SAF production

Water scarcity footprint of SAF production (${WSF}_i^t$), expressed as volume “equivalent” (Leq/tonne SAF), is calculated as:

$${WSF}_i^t={Waterconsumption}_i^t·{CF}_i^t$$

The water consumption of SAF production is the water consumed during the production of bio-jet fuels (expressed as L/tonne SAF), depending on the production technology. Here, water consumption for the three different production pathways was based on the study of Chong et al.⁶⁷ The water requirement for HEFA-SPK, ATJ-SPK and FT-SPK was taken as 1.3977 L/L_fuel, 8.5 L/L_fuel and 4.5 L/L_fuel respectively.

The WSF were calculated for three time periods: 2024, 2027 and 2030 (t = {2024, 2027, 2030}). Two different datasets are calculated: one using basin-aggregated CFs and one country-aggregated CFs based on the SAF facilities location, assuming that water sourced from the same basin/country will be used. The water required for each production pathway was multiplied with the production capacity of that pathway for every country and period. The final basin- and country-WSFs (million L_eq/tonne SAF) were calculated for a combination of four GCMs and two RCPs.

Quantification and statistical analysis

There are no quantification or statistical analyses to include in this study.

Published: March 20, 2026