Partitioning blue and green water sources of evapotranspiration using the water evaluation and planning (WEAP) model

The increasing demand for water due to population growth, changing living standards, alteration of consumption patterns, and rapid expansion of irrigated agriculture has led to an increasing recognition of problems facing river basins in certain parts of the world (Cosgrove & Loucks 2015; Kummu et al. 2016). This is exacerbated by changes in rainfall patterns and the increased frequency of weather extremes, which impact the withdrawal and supply of freshwater and worsen water scarcity in many regions (Wheeler & Von Braun 2013). Regions that struggle to meet rising water demands typically have competing water uses and lack the necessary instruments to better understand water resources through timely and accurate information on the spatiotemporal availability of water (Maestu & Gomez 2011; Parris 2011; Hundertmark et al. 2020). Among the competing water-use sectors, irrigation accounts for approximately 70% of global water withdrawals and is a vital component of global agricultural production (Rockström et al. 2017; FAOSTAT 2020). Increasing water scarcity and competition for limited water resources thus pose significant challenges to the sustainability of irrigated agriculture (Vörösmarty et al. 2010; Borsato et al. 2020).

The water sector has responded to increasing demands by embracing new paradigms of water resource planning and management. A highly favored approach is the development of an integrated water resource management system that offers a comprehensive solution (Mersha 2021). The success and practicality of such approaches, however, rely on accurate estimation of water resources, which vary over space, time, and source. While there is notable improvement in streamflow gauging and water supply measurements, unpredictable weather patterns necessitate a shift of focus toward quantifying water use and investing in water accounting approaches (Marston et al. 2022), particularly in the context of agriculture. Water accounting approaches quantify the spatiotemporal dynamics of water stocks and flows in a geographic region. These approaches provide a comprehensive outlook of when and where water is available and needed and assists water managers and professionals in supporting policy development and long-term decision-making.

The growing need for quantifying water-use components, especially from the agricultural sector, arises from the substantial losses attributed to evapotranspiration (ET). ET accounts for approximately 60% of the average global precipitation but is among the least understood and most complex components, emphasizing the importance of enhancing our knowledge in this area (Li et al. 2009). Idso et al. (1975) highlight the application of ET in the prediction of runoff and groundwater recharge, which can be applied to flood and drought predictions. ET estimates have been primarily used as a decision support tool for irrigation area expansion, changing cropping patterns, and estimating irrigation intensity and efficiency (Yibeltal et al. 2013). Thus, the significance of employing water accounting approaches for estimating ET in agricultural water management is evident. An important step in this regard is differentiating between blue and green water components of ET. Quantifying these components has profound implications for water and food security, ecosystem services, and societal needs. It is key to evaluating tradeoffs between prioritizing irrigated agriculture, which heavily relies on blue water, and rain-fed agriculture, which depends on green water.

In Ethiopia, agricultural water accounts for the largest consumption of water, particularly in arid and semiarid regions such as the Awash Basin. Depending on the landscape being considered, these consumption patterns occur through ET and are undifferentiated and can originate from either rainwater (green water) or water from stream flow or storage (blue water). According to Hoekstra (2019), it is important to differentiate between green and blue water consumption for several reasons, such as (1) to estimate irrigation efficiency, (2) to isolate the impact of blue water consumption on the remaining blue water sources, which helps ensure that environmental flow requirements are met, (3) to estimate crop productivity, which is especially important in irrigated farming, where the goal is to improve crop productivity per unit of blue water consumed, and (4) to explore options for incorporating irrigation into previously rain-fed agriculture. Therefore, quantifying the amount of green water originating from rainwater and when the rainwater supply falls short, the amount of blue water allows for a comprehensive picture of when to use which component – as water availability fluctuates within and across years. With anticipated shifts in climate patterns and pressure from land use changes, accurately partitioning and accounting for blue and green ET in localized regions such as the Awash Basin will better characterize and address water-use variations in the agricultural sector.

Previous studies have used different methods to partition consumption/ET into green and blue water sources and provide a foundation for this study. Velpuri & Senay's (2017) water and energy balance-based ET partitioning approach shows the contribution of blue water ET and green water ET over time and over various landscapes across the contiguous United States (CONUS), with 30% of the water originating from blue water sources and 70% from green sources. The study's high-resolution data and methodology offer valuable insights for improved decision-making in land and water management. Using a global crop water model, Siebert & Döll (2010) show that irrigated crop water use constitutes approximately 31.4% of global consumptive crop water use, out of which (for irrigated crops) 56.2% originates from blue water sources and 43.8% from green water sources. In this study, the necessity of evaluating the spatial distribution of irrigation requirements is emphasized, as the impact of water use in arid and semiarid regions and regions with competing water-use sectors is greater (Siebert & Döll 2010). Liu & Yang (2010) also show the importance of estimating green water sources, which constitute the major contributor to global agricultural water use. A substantial 87% of global crop water use originated from green water sources (Liu & Yang 2010), indicating the need to better manage green water sources. Partitioned ET sources have also been used in assessing local and regional water stress in major cropland-dominated basins to evaluate impact, stress variations, and associated stress drivers (Khand et al. 2021). Hoekstra's (2019) physically based soil water balance approach tracks changes in soil moisture components over time and proposes incorporation of green‒blue water accounting into soil‒water balance models instead of using simplistic assumptions in water-use estimation (Hoekstra 2019). These studies demonstrate how the proposed methodologies are also applicable in water resource management and planning. Understanding the reliance on blue and green water is crucial, especially in irrigated areas, as changes in blue water resources may impact crop production (Velpuri & Senay 2017), and climate-induced precipitation variabilities could affect green water sources, thereby affecting rain-fed agriculture (Khand et al. 2021).

Understanding region-specific water management challenges and preferred modeling approaches are critical for practical applications. In the Awash Basin of Ethiopia, stakeholders have shown a preference for the water evaluation and planning (WEAP) system (Sieber 2006) due to its prior use, widespread adoption, good documentation, and ability to simulate complex water systems to incorporate different water management scenarios. Additionally, WEAP allows stakeholders to visualize and quantify the impacts of water management decisions on the system's water balance. Therefore, this study aims to utilize the soil moisture water balance method within the WEAP model to trace and analyze the origins and dynamics of green and blue water sources contributing to ET in irrigated landscapes of the Awash Basin. By doing so, we assess the extent to which the Awash Basin relies on managed water resources and rainfall and when these dependencies are most significant.

The climate data needed for setting up the WEAP model consists of rainfall, temperature, relative humidity, and wind speed. For precipitation, remotely sensed rainfall data from Climate Hazards Group InfraRed Precipitation with Station data (CHIRPS; Funk et al. 2015) were acquired on a monthly time scale for the period 1981–2020. The 5 km resolution gridded precipitation data were then converted to the node-based data needed by WEAP by computing the area average for each subbasin. The other climate datasets used were the observed average monthly temperature (°C) obtained from the National Meteorological Agency (NMA) and the Climate Forecast System Reanalysis (CFSR) relative humidity (%) and wind speed (m/s) (Saha et al. 2010).

The basin land cover classification comprising forest, grassland, irrigated and rain-fed cultivated land, shrub land, urban land, and water bodies was extracted from the national level WaPOR Land classification portal (FAO 2020). This land classification dataset was chosen for this study because it separates cultivated areas into rain-fed and irrigated areas. To simulate monthly fluctuations in crop water use, dominant vegetation types and their corresponding area coverage were used to estimate an average crop coefficient (kc) per land cover type for rain-fed and irrigated agricultural landscapes. Cereals, flowers, oil crops, pulses, root crops, and vegetables are identified within the irrigated agricultural landscape, while rain-fed agriculture consists mainly of cereals and pulses.

Observed stream flow gauge data from 1981 to 2008 at four stations, namely, Awash-Kuntre, Awash Hombole, Awash-Awash, and Awash Tendaho from the most upstream to downstream of the Awash River (Figure 1), were collected from the Ministry of Water and Energy (MoWE). Volume-elevation curves for dams located on the main river were also collected from the MoWE, which was used to define dam operation levels and storage capacity in WEAP. The maximum pool levels for Koka, Kessem, and Tendaho reservoirs, as read from the volume-elevation curve, are 1,071, 500, and 1,870 Mm³, respectively. Water use based on administrative boundaries was collected from the Ethiopian Central Statistical Agency (CSA), in which water demands were added as domestic, industrial, and livestock demands.

The WEAP system was selected to estimate the spatial–temporal variation in total actual evapotranspiration (AET) and to split it into its respective green and blue water sources. As a water allocation model, WEAP is capable of representing water-use patterns and rates associated with domestic, industrial, and agricultural purposes. Through a representation of basin inflows and outflows, the volume of water consumed can be related to its annual replenishment rate. The model also allows users to analyze current status and future trends in water supply, demand, distribution, accessibility, and water use in time and space within specified domains, making it an appropriate model for this study (Sieber & Purkey 2015).

As irrigation demand constitutes the highest percentage of water demand, having both water supply and irrigation water demand configured on the same subbasin scale allows for better quantification of subbasin inflows and outflows. This subbasin water balance, as a subset of the basin water balance, is expected to provide greater accuracy when accounting for inflows and outflows over space. The remaining demands (industrial, domestic, and livestock) were added to the schematic as demand nodes corresponding to the third administrative level (‘woreda level’). Following model schematization, the soil moisture method was selected among the rainfall-runoff methods to simulate the hydrology of the basin. This entails the representation of each subbasin in three layers, with one surface and two subsurface (soil) layers. Within these layers, WEAP simulates evapotranspiration considering rainfall and irrigation on agricultural and nonagricultural land, runoff and shallow interflow (IF), base flow (BF), and changes in soil moisture. The method uses the modified Penman‒Monteith equation for calculating potential evapotranspiration to be used as input for computing the AET with an empirical relation that considers the soil water percentage in the two subsurface layers. IF separation and BF routing to the river, as well as soil moisture changes in the vertical column, are also simulated in both the upper and lower soil layers (Sieber & Purkey 2015). This method was adopted for simulation by taking climate inputs, including precipitation, temperature, relative humidity, and wind speed parameters, and allows for the characterization of land use and/or soil type impacts on these processes. Overall, the study area is configured as a contiguous set of subbasins that cover the entire extent of the river basin under examination.

The model was calibrated using historical stream flows at four gauging stations. This involved changing land use parameters and modifying the operating rules of flow requirements downstream of dams to improve the fit between simulated and observed flows. The calibration process focused on adjusting the most sensitive parameters, including the runoff resistance factor, SWC, relative storage, and water holding capacity of the root zone. While performing the stream flow calibration, first, the observed and simulated monthly stream flow time series were evaluated using statistical performance measures (see Section 3.3.3.). Next, unmet demands and subbasin/basin water balance were checked, and the calibration parameters were manually adjusted on monthly variation bases and/or year-round values for the land use classes in each subbasin.

Water balance, governed by the law of conservation of mass, dictates that the rate of water entering a specified domain is equal to the rate of water leaving the same domain with any differences attributed to changes in storage. The link between surface water, groundwater, soil moisture content, and the process of evapotranspiration has critical importance in a water balance study. The volume of water flowing into and out of each subbasin was the second measure of calibration and water-use separation evaluated in this study. As a subset of basin-level water balance, subbasin water balance quantifies each element of the hydrological cycle at the subbasin scale. The annual water balance of the basin is accounted for based on the law of conservation of mass that the change in soil moisture storage is the difference between basin inflow and outflow. Through the soil moisture rainfall-runoff method adopted in WEAP, rainfall, and irrigation are accounted for as inflow to the basin system, and evapotranspiration, surface runoff, IF, and BF are accounted for as outflow from the basin system. The difference between an increase in soil moisture and a decrease in soil moisture (DSM) is considered the change in basin storage. Previous studies were used to compare the water balance at both basin and subbasin levels as well as to verify the spatial distribution of open water bodies such as lakes, wetlands, and floodplains.

The separation of green and blue evapotranspiration is a relatively new concept that has significance for water resource management and efficient water resource uses. Different methods using different inputs have been introduced to apply the concept and separate blue and green ET (Van Eekelen et al. 2015; Senay et al. 2016; Simons et al. 2020).

The first conditional analysis was performed to determine the share of water contributed from the DSM, irrigation (I), and precipitation (P).

Estimating the amount of water released through ET from soil (ET_SD): This component originates from a DSM, where DSM is the net decrease in soil moisture compared to the previous time step, which may initially originate from either precipitation or irrigation (in the previous time step) and contribute to IF, BF, and ET. DSM was compared to IF and BF to determine whether the source is large enough to accommodate both.

Estimating the amount of water released through ET from irrigation (ET_IR) and precipitation (ET_PR): This component, in addition to IF and BF, takes into consideration additional sinks such as an increase in soil moisture (SMI), surface storage (SSI), and surface runoff (SR). Four conditions for balancing sources and sinks to determine the amount taken up by ET were defined. The primary sink of irrigation water is considered to be SMI, as irrigation is based on water management with irrigation application taking place when precipitation is not sufficient to replenish soil moisture. As a result, water applied through human management would contribute solely to soil moisture increments. Precipitation sinks were SR and SSI, as overland flow and surface storage are regulated by land cover and are mainly a result of precipitation. Table 1 shows the simultaneous conditions and ET share associated with each source.

Table 1

Estimating the share of ET from soil moisture, precipitation, and irrigation

View Large

Using these sets of conditions, blue and green ET were computed for irrigated landscapes in the entire basin and over the 21 subbasins on an average monthly basis for the period 1981–2020. The basin-level separation was expected to show the overall blue and green ET components in the irrigated land of the Awash Basin, while the analysis of each subbasin is expected to show the variability in blue and green ET, both spatially and temporally, and to reflect the modality and magnitude of rainfall, serving as an indicator of the degree to which regions in the basin rely on rainfall and irrigation for agriculture.

The water balance, evaluated by accounting for inflows and outflows at the basin and subbasin levels, was the second means of model verification used in this study. The annual basin water balance resulted in a total inflow of 84.9 BCM originating from precipitation (79.5 BCM) and irrigation (5.4 BCM), outflows through ET (63.8 BCM), runoff (15.2 BCM), IF (3.3 BCM), and BF (2.6 BCM), and the remaining change in storage through the difference in the increase (addition to soil storage) and decrease (removal from soil storage) in soil moisture, which is considered insignificant at an annual scale.

The water balance of the basin was further disaggregated to the subbasin level to account for the spatial distribution of stream flow and to identify the points of open water systems (lakes, flood plains, and reservoirs) in the basin. Subbasin water balance showed that the basin has 15.7 BCM of annual renewable water in three different forms: surface water flow at the outlet (4.6 BCM), open surface storage or flood plain loss (3.5 BCM), and flow to the groundwater system (7.6 BCM). The identification of water storage and evaporation among the well-known subbasins validates the model's capacity to address the physical characteristics of the basin, which helps to separate evapotranspiration into green and blue water sources.

The first is the Awash-Kuntre subbasin, which is located at the upstream end of the Awash Basin and is characterized by a mono-modal rainfall pattern with the months of June–September making up the rainy season and the average precipitation exceeding 1,300 mm per year. The subbasin also relies heavily on rain-fed crop production, and the contribution of irrigation water or blue ET is very minimal during the rainy season. The second subbasin is the Awash-Awash subbasin, which is located in the middle course of the Awash River below the Koka Reservoir (Figure 7) and is characterized by intensive irrigation, especially for the sugarcane production in the Wonji and Metehara irrigation fields. The Awash-Awash subbasin receives an annual average precipitation of 1,000 mm. We also want to highlight how green and blue ET varies based on the total amount of precipitation in a region in addition to how much irrigated agriculture is present. For that, we present the Awash-Terminal subbasin, which is the downstream part of the Awash River, located below the Tendaho Reservoir. This subbasin receives a total annual average precipitation of 150 mm and heavily relies on irrigated agriculture.

The land cover classification map (from the WaPOR database; FAO 2020) shows that the irrigated area in the Awash-Kuntre subbasin is approximately 1.1%, while rain-fed cropland is 90% of the subbasin. During the rainy season, between May and September, ET from the Awash-Kuntre subbasin is completely green, while during the other months, it is dominated by blue ET. In the Awash-Awash subbasin, irrigated agriculture covers a relatively larger area (587 km²) and has an average maximum monthly ET of 81 MCM (137.8 mm) for irrigated areas. Blue ET dominates all but two months (July and August), where ET is partly due to precipitation. The temporal pattern is an indication that irrigation is the main source of water in the Awash-Awash subbasin for the majority of the time within a year. In the Awash-Terminal subbasin, the monthly maximum evapotranspiration from irrigated areas is approximately 43 MCM. The irrigated agricultural area in the Awash-Terminal is approximately 241 km², which makes the average annual evapotranspiration from the subbasin equivalent to 178.2 mm. With such low annual precipitation, rain-fed agriculture in the Awash-Terminal subbasin is considered negligible, and irrigated agriculture is dominant. In all months, blue ET accounts for total evapotranspiration, which shows that there is a greater reliance on irrigation water for crop production in that region.

Accurately quantifying water sources that make up the largest water-use component, evapotranspiration allows for targeted improvement of crop productivity. Whether to implement better water-use practices that reduce nonproductive ET or evaluate agricultural management practices to increase production while maintaining a consistent ET depends on the accurate spatiotemporal attribution to components of ET. Notably, the Awash Basin is considered to be the most utilized and human-modified basin in Ethiopia, with significant agriculture and multiple competing water users. Given the changing climate patterns and growing food and water demand, potential threats to this basin emphasize the importance of accurately estimating water consumption through ET and discerning its source (whether it originates from blue or green water sources).

By applying the soil water balance method in WEAP and water balance principles (and magnitude of fluxes), conditions were developed to estimate the contributions of soil moisture storage, irrigation, and precipitation to ET. These contributions were then partitioned into green ET and blue ET and were applied to the basin, where, out of the total annual irrigated landscape ET, 88.4% originated from blue ET and 11.6% from green ET. In the basin-level water balance analysis, ET had the largest contribution compared to the other hydrologic components. Precipitation peaks in July–September and April–May, resulting in higher green water consumption (green ET) during these periods. In drier months, irrigation (surface and/or groundwater) takes precedence, driving up blue water consumption. Blue ET is significantly influenced by soil moisture fluxes and ET_SD contributions. At the subbasin level, distinguishing between blue and green ET revealed varying patterns. Upstream irrigated areas heavily relied on rainfall during the wet season and a mix of irrigation and rainfall during the dry season for crop production. Irrigation accounted for less than 3% of the streamflow. Further downstream, the crop production was reliant on both rainfall and dry season irrigation using up to 30% stream flow. In most downstream subbasins, irrigation was crucial, constituting up to 70% of stream flow.

Highlighting blue and green ET in irrigated landscapes reveals the complementary use of rainfall and irrigation, offering insights for expanding agriculture and incorporating irrigation into previously rain-fed agriculture. This analysis enhances understanding of water availability in the Awash Basin and can be further extended by incorporating information on crop yield.

The authors acknowledge the Development Cooperation Agency of the Dutch Ministry of Foreign Affairs' Water Development Partnership Program (DUPC2) project Water Accounting + Phase II 2020–2022.

All relevant data are included in the paper or its Supplementary Information.

The authors declare there is no conflict.