Water use of Prosopis juliflora and its impacts on catchment water budget and rural livelihoods in Afar Region, Ethiopia

Dense impenetrable thickets of invasive trees and shrubs compete with other water users and thus disrupt ecosystem functioning and services. This study assessed water use by the evergreen Prosopis juliflora, one of the dominant invasive tree species in semi-arid and arid ecosystems in the tropical regions of Eastern Africa. The objectives of the study were to (1) analyze the seasonal water use patterns of P. juliflora in various locations in Afar Region, Ethiopia, (2) up-scale the water use from individual tree transpiration and stand evapotranspiration (ET) to the entire invaded area, and 3) estimate the monetary value of water lost due to the invasion. The sap flow rates of individual P. juliflora trees were measured using the heat ratio method while stand ET was quantified using the eddy covariance method. Transpiration by individual trees ranged from 1–36 L/day, with an average of 7 L of water per tree per day. The daily average transpiration of a Prosopis tree was about 3.4 (± 0.5) mm and the daily average ET of a dense Prosopis stand was about 3.7 (± 1.6) mm. Using a fractional cover map of P. juliflora (over an area of 1.18 million ha), water use of P. juliflora in Afar Region was estimated to be approximately 3.1–3.3 billion m3/yr. This volume of water would be sufficient to irrigate about 460,000 ha of cotton or 330,000 ha of sugar cane, the main crops in the area, which would generate an estimated net benefit of approximately US$ 320 million and US$ 470 million per growing season from cotton and sugarcane, respectively. Hence, P. juliflora invasion in the Afar Region has serious impacts on water availability and on the provision of other ecosystem services and ultimately on rural livelihoods.

In the context of climate change and rising demands by a growing human population, water is becoming an increasingly scarce resource 1 . Human societies in arid and semi-arid regions are particularly prone to water scarcity during at least some parts of the year 2 . In those regions, meeting the rising demands for freshwater and protecting or restoring ecosystems so that they can sustainably replenish water stocks will be one of the most difficult and important challenges of this century.
Water is lost through transpiration, interception and evaporation. Evapotranspiration, which can account for more than 90% of the water balance in some catchments, depends on the climatic factors e.g., solar radiation, air temperature, air humidity and wind speed [3][4][5] . In addition, the loss of water to the atmosphere is also affected by the available soil moisture, vegetation structure and the eco-physiology of the dominant plant species 6,7 . Species traits such as plant physiology (e.g. ever greenness), size and root architecture affect water availability and water use both at the individual plant as well as at the ecosystem level 8,9 . For example, at the watershed level, afforestation of grasslands and shrubland can result in significant reduction of stream flow 7 .
Besides land-use change and land degradation, invasive alien plant species (IAPS) are among the most important drivers of change in vegetation composition and ecosystem functioning in terrestrial ecosystems 10,11 . Due to their invasive character, these plants can become the dominant species both at the local as well as at the landscape scale. Some of the well-known examples of changes in vegetation composition due to plant invasions www.nature.com/scientificreports/ combined the stand level water use measurements with the fractional vegetation cover of P. juliflora in the Afar Region that was mapped and published by Shiferaw et al 26 . Finally, we estimated the annual costs of water use by P. juliflora at the regional level.

Materials and methods
Study area. Water consumption of P. juliflora (hereafter Prosopis) was measured in the Amibara District ( Fig. 1) of the Afar Region, Ethiopia (9.16° to 9.21° N and 40.08° to 40.12° E at 740 m a.s.l.). The study area is located in the Awash River Basin and includes both the floodplains of the Awash River and the adjacent dryland. Although relatively water scarce, the Awash River Basin is the most developed and utilized river basin in Ethiopia 27 . The human and livestock population in this basin are estimated at 18.6 and 34.4 million, respectively, and nearly 70% of large irrigation schemes in Ethiopia are located in the Awash Basin 28 . The mean annual river flow/discharge of the basin at the terminal Lake Abe is estimated at about 4.6 billion m 3 of water although exposed to evaporation 27 . Water scarcity, particularly in the lower parts of the river, is the major limiting factor for irrigation development, particularly during the low flow season 27 .
The Afar Region has a mean annual rainfall of about 560 mm 29 . The region is the hottest part of Ethiopia, with a mean annual temperature of 31 °C. The mean maximum temperature reaches up to 41 °C in June, and the mean minimum temperature ranges from 21 to 22 °C between November and December 30 . The biome can be described as semi-arid to semi-desert. The natural vegetation consists of scattered dry shrubs, woodland comprising different Vachellia (Acacia) species, bushland, grassland and wooded grassland 18 . The area has different soil types, including silt fertile soils, sandy soils, heavy clays and rocky outcrops, and a wide range of altitudes ranging from 175 m below sea level to 2,992 m a.s.l. Shiferaw et al. 30 found that Prosopis has primarily invaded areas ranging from rangelands to farmland. Main sources of livelihood are pastoralism and some agro-pastoralism around small rural towns 30 . The main crops grown in the floodplains of Awash River are cotton and sugarcane.
Experimental design. To investigate the temporal and spatial variation in water use by Prosopis across the landscape, data collection was done in the two most heavily invaded habitat types in the Afar Region. These include the floodplains of Awash River and the adjacent non-riparian drylands. Sap flow monitoring stations were established at two sites in the floodplains of the Awash River and at two sites in the dryland area where soil moisture levels were low (Fig. 1). The study areas were representative of other parts of the Afar Region that are invaded by Prosopis.
Site 1 was located near Worer Agricultural Research Center some 200 m away from Awash River and about 30 m from a nearby irrigation canal, and was considered a floodplain site. The area was used for crop production until 2012 after which it was abandoned due to shallow water inducing soil salinity problems. Soon after abandonment, Prosopis invaded the area and established dense stands with 100% canopy cover in most places comprising trees up to 6.5 m height. Soil moisture was relatively high at Site 1 due to proximity to the river. The area invaded by Prosopis was about 6 ha and the soils were temporarily flooded loam and clay soils. Except for annual grasses in some open spaces, there was no undergrowth vegetation, probably due to the dense Prosopis cover.
Site 2 was in the floodplains of a tributary of Awash River and located near Berta locality (Fig. 1). There is a continuous flow of water from a pumped well to less than 10 m from the site where sap flow was measured. The soil is a sandy loam formation, which maintains relatively high moisture content. The Prosopis stand was about 3 ha in size, had a closed canopy and comprised trees of more than 5 m height.
Site 3 was on dryland in the former rangeland of Hallaideghe locality ( Fig. 1), characterized by sandy loam soil formation. There was no surface water source except from rainfall and seasonal flooding from the West Harerghe highlands. This area is now invaded by Prosopis with a closed canopy and tree heights of close to 5 m. The size of the invaded area was about 100 ha.
Site 4 was in the drylands of Berta locality (Fig. 1). The soil is sandy with a high rocky outcrop. The dominant indigenous vegetation around this site consisted of a few Senegalia senegal (L.) Britton (or Acacia senegal (L.) Willd) and other small shrubs and grasses. The area is now dominated by Prosopis stands with a closed canopy with trees reaching up to 4 m height. At this site, a meteorological station was installed next to the sap flow monitoring equipment.
The sap flow data in this study therefore provide insight into how tree water use varied across the landscape including (a) floodplains, (b) dryland areas, and (c) across these two most heavily invaded habitats in the Afar Region. All experimental sites were fenced and protected from animal and human interference. Moreover, safety boxes were also made to protect the equipment from weather and other damages.
Tree and stand water use measurements. The amount of water used by individual Prosopis trees was determined using the heat ratio method (HRM) of monitoring tree sap flow 31 . This technique was selected because it is suitable for measuring low and reverse sap flows which are likely present in desert-adapted species such as Prosopis 14,32 . In total, four sap flow stations were established (Sites 1 to 4) with three trees instrumented per station. Trees with different stem diameters were selected to capture the variation in transpiration rates in the study region. Stem diameter of the instrumented trees was measured just below the branching at about 60 cm above the ground at all sites. Each sap flow station comprised a CR1000 data logger and an AM16/32B multiplexer, as specified by Campbell Scientific, Inc., Logan UT, USA. Each system was powered by a 70 Ah (12 V) rechargeable battery using 50 W solar panels. Four sets of heaters applied heat to each tree for 0.5 s every hour through a custom-made relay control module. Moreover, a pair of equally placed (0.5 cm) T-Type thermocouples was installed on either side of the heater to measure the sapwood temperature before and after pulsing the heat. With a precision drilling rig, two 2.0 mm diameter holes were carefully made for the thermocouples to www.nature.com/scientificreports/ minimize errors due to probe misalignments. Heater holes were about 1.8 mm diameter to ensure a tight fit to facilitate heat transfer to the wood during pulsing. In this sap flow monitoring technique, the heat pulse velocity (Vh, cm/h) is logarithmically related to the ratio of temperature increases upstream and downstream from a heater (v 1 /v 2 ) as shown in Eq. (1), Burgess et al 31 : where Vh is heat velocity cm/hour, k is the thermal diffusivity which was assigned a nominal value of 2.5 × 10 -3 cm 2 /s for wood, x is the distance (cm) between the heater and either temperature probe (~ 0.5 cm), and v 1 and v 2 are increases in temperature before and after pulsing 31 .
The thermocouples were installed in the sapwood at depths ranging from 0.8 to 1.1 cm under the bark to capture the radial changes in sap velocity. Wounding corrections were applied according to the method described by Swanson and Whitfield 57 . The depth of the sapwood was determined visually as it was possible to distinguish between the sap wood and heartwood boundaries from the changes in the color of the wood. The individual tree sap flow volume in liters per hour were converted to stand level transpiration (in mm per hour) using the approach described by Dzikiti et al 14 in which the instrumented trees were assigned to a particular stem size class. The stand level transpiration was then calculated as a weighted sum of the transpiration rates by the trees in each stem size class with the proportion of trees in each size class as the weights. The volumetric soil water content in the root-zone of the trees were measured at each site using a single soil water content reflectometer probe (Model CS616: Campbell Scientific, Inc., Logan UT, USA) installed horizontally at a depth of 50 cm. Sap flow and soil moisture data were measured for 15 months (from November 2016 to January 2018) while evapotranspiration was measured for 11 months (from January to November 2017).
To study the dynamics of total actual evapotranspiration (ETa) from Prosopis stands, an open path eddy covariance (EC) system was installed on Site 1 and data were collected for 11 months from January to November 2017 as it was not possible to continue measuring for more periods due to equipment limitations. The EC was borrowed from Addis Ababa University only for one year so ETa from the other sites could not be measured. The EC system was the IRGASON system which comprised a sonic anemometer (Model: CSAT3A Campbell Scientific Inc., Logan UT, USA) that measured the wind speed in 3-D at 10 Hz frequency. The H 2 O/CO 2 concentrations of the atmosphere were measured using an Infrared Gas Analyzers (Model: EC150, Campbell Scientific, Inc., Logan UT, USA). The collected data was stored by a data logger (Model: CR3000: Campbell Scientific, Inc., Logan UT, USA) on a Compact Flash card module (NL115 or CFM100). To quantify the changes in the energy balance of the study site, two other components of the surface energy balance were measured. These include the net radiation, which was measured using a single component net radiometer (Model: NR-LITE2: Manufacturer: Kipp & Zonnen, Delft, The Netherlands) that was mounted at the top of the tower (~ 7.5 m above the ground). The IRGASON sensor was installed outside the surface roughness layer of the canopy at an average height of about one meter above the Prosopis tree canopy. This ensured a uniform fetch around the tower with a flux foot print of about 100 m radius.
Air temperature and humidity were measured at high frequency using a temperature and humidity Probe (Model HMP155A-L, Campbell Scientific, 2013). The high frequency data were further corrected for 1) lack of sensor levelness (coordinate rotation), 2) sensor time lags, and 3) fluctuations in the air density using the EddyPro version 6.0 software (Li-COR, Nebraska, USA). Sensor separation corrections were not necessary as the IRGA and sonic are a single unit.
Allometric characteristics are one of the major biological factors affecting the eco-physiology of plant species. For example, sapwood area is usually correlated with stem diameter 14 . The sapwood area estimated from the stem diameter measurements was used to calculate the sap flow volumes from the sap velocity measured by the HRM system.

Weather and soil water dynamics.
To measure solar irradiance, precipitation, air temperature, relative humidity and air pressure, an automatic weather station was set up at Site 4, which was located within 7 km from the other three sites. The solar radiation sensor was installed on a horizontal leveling fixture mounted on a south facing cross bar to avoid self-shading errors. A wind sentry was used to measure the wind speed and direction (Model 03,001, R.M. Young; Campbell Scientific, Inc., Logan UT, USA). Rainfall was monitored using a tipping bucket rain gauge (Model TE525-L, Campbell Scientific, Inc., Logan UT, USA). The weather station comprised an Em50 (a 5-channel data logger) and ECH 2 O utility software from Decagon, USA.
Wind speed was obtained from the weather station located at Worer Agricultural Research Center, which was about 500 m away from Site 1. The weather station had a temperature and humidity probe (Model CS500, Vaisala, Finland) installed at a height of about 2.0 m above ground and the station also measured wind speed using a cup or rotational anemometer installed at 2 m high.
The energy transferred into and out of the ground was measured using clusters of soil heat flux plates (Model: HFP01SC-L, Delft, The Netherlands), while soil temperature was recorded using soil averaging thermocouples (Model: TCAV-L: Campbell Scientific, Inc., Logan UT, USA). The soil heat flux plates were installed at 8 cm depth and the soil averaging and soil moisture data measured with the soil water content reflectometers (Model: CS616-L: Campbell Scientific, Inc., Logan UT, USA) were used to correct the soil heat flux for the energy stored by the soil layer above them. At all sites, the sensors were connected to a data logger (Model CR1000, Campbell Scientific, Inc., Logan UT, USA) programmed with a scan interval of 90 s, and data were stored at hourly intervals over the 11 months study period. All data were downloaded every 21 days from data loggers. www.nature.com/scientificreports/

Drivers of water use by the invasive Prosopis. To identify the main drivers of water use by Prosopis
invasions in the Awash River basin of the Afar Region correlations were sought between the various water use variables (transpiration and evapotranspiration) as dependent variables and microclimate factors, i.e. solar radiation, wind speed, vapor pressure deficit of the air (VPD), soil moisture, and ET 0 as explanatory variables.
Upscaling Prosopis water use moderation to the Afar Regional level. To upscale the Prosopis transpiration and ET from the individual study sites to the regional scale, a regression equation was developed using the fractional vegetation cover information mapped by Shiferaw et al 26 . This mapping quantified the Prosopis distribution and cover in the Afar Region at a 15 × 15 m spatial resolution based on explanatory variables including Landsat panchromatic images, other biophysical parameters and field observations. The fractional cover map was generated using a robust modelling approach, Random Forest Algorithm, with a large amount of field observations (> 3000 plots) and seventeen explanatory variables 33 . Then, we estimated the amount of water used by Prosopis stands at each of the four sites in mm/day per pixel with 100% cover. This was extrapolated to all fractional cover levels per pixel indicated in the fractional cover map for Prosopis in the study area. Moreover, canopy cover from the experimental plots of Prosopis trees was estimated. Then, the relationship between sap flow and fractional cover as well as between ET and fractional cover (F ci ) over the invaded area was developed as shown in Eq. (2): where f(x) is either total water use (sap flow) or total ET in mm/day over the whole study area; Fc i is fractional cover at pixel level i , W i is water use either from sap flow or stand ET at 100% canopy cover, and S is a pixel size of 225 m 2 . Finally, we estimated the financial costs incurred from the loss of water through Prosopis transpiration and ET. This was done by taking the water charge of payment for ecosystems services by investors to Awash Basin Organization which was set at US$ 0.00015 per m 3 according to Ayana et al 34 . Also, we estimated the market price and the net benefits of cotton 35,58 and sugarcane 59 , which are major crops grown in the study area, which could be grown with the amount of water used by Prosopis.

Results
Climatic conditions and diurnal variation in sap flow. Sap flow varied with the daily solar radiation ( Fig. 2) and was the highest at high air temperature and low relative humidity (Appendix 1). The highest solar radiation was registered between 10:00 and 16:00, reaching up to 1000 W/m 2 around noon, and the daily average varied between 8.1 and 26.5 MJ/m 2 /day. The highest relative humidity was observed from 03:00 to 07:00 while air pressure was more or less constant throughout the day and was about 91 kPa (Appendix 1). When comparing sap flows of trees in floodplains and drylands, the sap flow was lower in the floodplains than in the drylands though the diurnal trends showed similar patterns (Fig. 2). found proportional with stem diameters of Prosopis tree with R 2 of 0.94 (Fig. 3, left). Moreover, stem diameters of the instrumented trees were similar between the floodplains and the dryland sites. There was a highly significant relationship between sapwood area and stem diameter of the sampled Prosopis trees. The stem diameter distributions were assumed to be representative for trees in the invaded areas (both floodplain and dryland; Fig. 3, right).
Sap flow of Prosopis trees. Water use of Prosopis trees varied with stem diameter, habitat type (dryland and floodplains), and the seasonal changes in the climatic conditions of the study area. On average, the sampled Prosopis trees used about 6.8 ± 2.9 l/tree/day. The instrumented Prosopis trees in the floodplains consumed on average less water (5.1 ± 3.4 l/tree/day) than those in the drylands (8.5 ± 5.8 l/tree/day) ( Table 1). On the other hand, the relationship between stem diameter and sap flow indicated that stem diameter explained about 48% of the variation in sap flow (Fig. 4, left). After normalizing the daily sap flow by sapwood area, the sap flux density also indicated the same trend that Prosopis trees in the dryland area exhibit higher sap flux density than those   www.nature.com/scientificreports/ in the floodplains (Fig. 4, right). Daily average sap flux density (SFD) explained about 64% of the water use of Prosopis across the different habitat types. In both cases, we found that trees in the drylands consume more water than those in the floodplains (Fig. 5). Moreover, regression analysis of sap flow as dependent and tree diameter and habitat type as explanatory variables also confirmed this with R 2 of 0.48, F-stat of 9.22 and significant level of 0.013, which is less than the common p value of 0.05 at 95% confidence level. This result, indicates that the water use of Prosopis trees in dryland differs significantly from that in the floodplains. When comparing seasonal variability between the dry season (January-March) and the rainy season (July-September), Prosopis trees were found to use more water during the dry season (7.5 l/tree/day) than during the rainy season (6.0 l/tree/day; Table 2). During the rainy season, the trees sampled in the floodplains used about 5.0 l/tree/day whereas the trees in the drylands used about 7.1 l/tree/day. On the other hand, during the dry season trees in the floodplains and in drylands used about 5.1 and 10.0 l/tree/day, respectively. Similarly, average sap flow across invaded areas reached 3.0 and 3.8 mm per day during the rainy and dry seasons, respectively (Table 2).
On the other hand, each of the Prosopis trees in the floodplains used less water than the trees in the dryland (Appendix 2). Sap flow was plotted across both habitat types and average sap flow (Fig. 5), indicates that dryland has shown higher sap flow for Prosopis than floodplains of the same species.
Evapotranspiration from Prosopis stand. The daily average of ET a ranged from 1.5 to 9.5 mm, with an annual average of 3.7 mm/day (Fig. 6, top). Moreover, the sap flow and ETa measured at site 1 revealed a similar pattern during the course of the year, although the daily values for ETa were almost always higher than those for sap flow. The average ETo for the 11 months was estimated at 4.99 mm/day. The lowest amount of ET was recorded in January and February due to the low atmospheric evaporative demand while the highest was in June, August and September (Fig. 6, bottom) during summer. Comparing Prosopis ETa with the potential evapotranspiration (ETo), the annual average crop coefficient (K c ) of Prosopis was found about 0.77 for Prosopis invaded areas.  www.nature.com/scientificreports/ Drivers of water use. Regression analyses of possible drivers of water use of Prosopis revealed that ETa was explained by wind speed, soil moisture, VPD, and ETo (all p < 0.0001; Table 3A). Sap flow at individual tree level was significantly explained by soil moisture and, VPD and to a lesser extent also by solar radiation and wind speed (Table 3B). Analyses separated by habitat indicated that soil moisture, wind speed, VPD and solar radiation explained a significant amount of variation in sap flow at tree level in drylands, while only VPD and wind speed explained a significant amount of variation in sap flow of Prosopis in floodplains (Table 3C & 3D). When combining results from floodplains and drylands, sap flow dynamics at tree level was significantly affected by solar radiation, wind speed, soil moisture and VPD (Table 3E). Crop coefficient of Prosopis was also estimated at 0.77 (Table 3A), which is the slope of ETa plotted against ETo. When comparing climatic factors as drivers of water use and measured water uses (in terms of sap flow and ETa), ETa was found the lowest when wind speed was low while sap flow of trees in the drylands changed with VPD, soil moisture, and with solar radiation, but not with wind speed. It was also observed that the higher the solar radiation, the higher the sap flow ( Fig. 2; Appendix 3).
Upscaling Prosopis water use moderation to the Afar Regional level. Depending on Prosopis cover, sap flow and ETa were estimated to range between 3 and 826 mm/day at a pixel size of 15 × 15 m (or the maximum value of 826 mm/day/pixel, is approximately equivalent to 3.67 mm) (Fig. 7). Upscaling the results from sap flow of Prosopis trees and ETa of Prosopis stand from pixels to the regional scale revealed that Prosopis consumed about 3.1 and 3.3 billion m 3 per year, respectively, over the invaded areas of 1.18 million ha in Afar Region, which approximates to 279 mm per year of water use and it is about 50% of the annual precipitation of the region.
Based on current regional water tariffs, the economic impact of water loss due to Prosopis invasion in the Afar Region was estimated at US$ 465,000-500,000 per year. Considering that cotton production requires some 7,000 m 3 /ha/year (consultation with experts), the water consumed by Prosopis corresponds to the amount of water needed to grow approximately 460,000 ha of cotton with an average crop yield of 2.1 ton/ha 35

Discussion
The results of this study indicate that the invasive evergreen Prosopis uses a significant amount of water (3.1 to 3.3 billion m 3 per year over invaded areas) in the Afar Region. Considering that the annual precipitation in the study region is approximately 560 mm 29 , our findings indicate that Prosopis consumes about half of the amount of water received from precipitation in the invaded area in the Afar Region (Appendix 1). www.nature.com/scientificreports/

Sap flow in individual trees and evapotranspiration from stands of Prosopis.
Our study revealed that variation in sap flow among Prosopis trees in Afar Region was affected by sapwood area, stem diameter, habitat type and by climatic factors. This finding concurs with Dzikiti et al 22 that reported a close relationship between sapwood areas and stem diameter as well as between stem diameter and sap flow at the tree level from invasive species, leaf-shedding Prosopis species in South Africa. Tree hydrology is largely controlled by species traits, the regional climatic characteristics and dynamics of water and energy inputs during the growing season 6 . The higher solar radiation, vapor pressure deficit and air temperature in the drylands may thus favor higher water use by Prosopis trees growing in this habitat, relative to trees growing in the floodplains. The higher sap flow of evergreen Prosopis trees in the drylands of Awash River Basin may therefore be explained by the fact that the study area is a flat plain which may provide a sufficiently large groundwater reservoir from which deep-rooting trees of Prosopis can take up water also during the dry season, as groundwater can be found within the range of 10-20 m depth and the depth increased as we move away from the river (expert consultation). Le Maitre et al. 9 reported that in South Africa tree water use of leaf-shedding Prosopis species is higher in riparian settings or shallow water tables where water availability is greater, relative to dryland settings. They therefore concluded that the impact of Prosopis on ecosystems services is greater in riparian than in dryland ecosystems. Similarly, in the southern parts of the USA water use by Prosopis velutina Wooton tended to be higher in the floodplains than in adjacent dryland (Scott et al., 2006). In contrast, the present study revealed that sap flow of the evergreen Prosopis juliflora trees was higher in the drylands than in the floodplains of Awash River Basin. This contradicts the general assumption that trees growing in areas with wet soils consume more water than trees growing in dry soil. While rainwater is the preferred source of Prosopis when available, Dzikiti et al. 14,22 showed that in South Africa Prosopis switched to tapping groundwater once rainwater sources were exhausted. Fritzsche et al 39 found that in South Ethiopia growth of the exotic Eucalyptus globulus Labill. was largely independent of topsoil water content, because tree has the flexibility to resort to groundwater if necessary.
In general, tree water use is expected to vary between the wet and the dry season. In semi-arid to arid regions such as Afar, water availability during the wet season is high and VPD is relatively low. During the dry season, when soil moisture is low and VPD, temperature and solar radiation are high, tree water use should decline to prevent canopy desiccation. However, in contrast to the studies cited above, in savannas of northern Australia tree water use was found to either not differ between wet and dry seasons 40 , or even increase during the dry season 41 . This result was explained on the basis of the fact that rooting depth allowed the plant access to sufficient water stored in the soil also during the dry season to enable transpiration to proceed. Similarly, the study found that the evergreen P.juliflora used more water and transpired more during the dry season than during the www.nature.com/scientificreports/ rainy season. Tanaka et al 42 showed that water use of trees in a hill evergreen forest in Thailand peaked in the late dry season. The factors contributing to higher water use of trees in drylands relative to the floodplains may thus also explain why Prosopis in Afar consumes more water during the dry than during the wet season. Hence, the findings suggest that the evergreen leaf feature of P. juliflora not only leads to water abstraction from the groundwater during both the wet and the dry season, but also it even increases water abstraction during the dry season, which significantly exacerbates regional water shortage. Total ET is acquired from three different sources, namely soil evaporation, interception loss and transpiration, which are mainly influenced by the vegetation composition. For example, the degree of ground cover has a substantial influence on soil evaporation 43 . In this study, however, soil evaporation contributed little to total ET, which might be due to the fact that the soil is very dry and understorey transpiration is low because of absence of undergrowth in the study area. However, the degree of soil evaporation, in turn, is determined by the meteorological conditions 44,45 . In this study ET varied with microclimate, particularly precipitation and its effect on soil moisture and wind speed, whereas sap flow varied with habitat type, solar radiation and sap wood area of Prosopis trees. These results thus suggest that management attempts to reduce water loss due to Prosopis invasion should primarily target invasive trees in the drylands where trees consume particularly high amounts of water, also during the dry season.
The water use of invasive species is of concern in species conservation or ecosystem restoration programs in semi-arid and arid ecosystems, as well as when adapting to climate change by reducing competition for the often scarce water resources 9,14 . Prosopis invasions are having a negative effect on the stability of native tree populations in South Africa and are linked to increased mortality of native trees 46 . While Prosopis has been reported to have allelopathic effects 47 , it is likely that dense thickets of Prosopis also impact native vegetation through competition for water 48 .
Various studies indicate that, although individual native and invasive tree species may have similar sap flow rates, but that invasive-dominated ecosystems were more likely to have higher sap flow rates per unit area than native-dominated ecosystems because invasive species grow in denser stands than natives and dense stands consume more water than sparsely stocked stands 13,14,36 . Dzikiti et al 14 reported that in South Africa Prosopis stands transpired five times more water (544 mm/y) than Vachellia karroo (Hayne) Banfi & Galasso stands (91 mm/y). The results of the present study suggest that in the tropical Afar Region Prosopis stands with 100% cover transpired about 1241-1350 mm/y, which is about six times more water than used by Prosopis stands in South Africa 12 . The results of Dzikiti et al 14 and this study were based on a shared methodology and should therefore be comparable. Nevertheless, there may be several other factors affecting the significantly higher transpiration revealed by the current study. First, the invasive species in Ethiopia and other parts of the tropical region in Eastern Africa, P. juliflora, is the evergreen plant which uses water throughout the year for its photosynthesis activities, whereas the Prosopis taxa in South Africa shed leaves during the cold seasons and are thus likely to have periods with low transpiration and evapotranspiration. Secondly, Prosopis in Afar builds very dense thickets, due to past and partly ongoing utilization for charcoal production, which created multi-stemmed trees that use more water. Also, the deep-rooted Prosopis may benefit from the alluvial nature of the soil in Afar region which allows easy penetrations of the soil layers so as to abstract more water than native species do, although sandy soils drain more water to the groundwater table. This type of soil, in turn, forces the tree to grow deep tap-root systems in search for more water.
Finally, comparing three scenarios for crop specific coefficients (grass reference at standard height of 0.5 m, bare soil and Prosopis invaded areas scenarios), bare soil have a crop coefficient (Kc) of 0.3 to 0.5 while wellwatered grass has 0.7 to 1.0 3 whereas Prosopis invaded areas exhibited a Kc of about 0.77 all year round, indicating that invaded areas extracted more groundwater on an annual basis as grass-dominated habitats in the study area consume water only during the 2-3 months of the rainy season and would dry up during dry seasons, unlike the areas invaded by evergreen Prosopis in Ethiopia.
Upscaling Prosopis water use moderation to the Afar Regional level. Understanding the effect of vegetation change on transpiration and ET across habitat types (in our case drylands and floodplains) is important for scaling up since ET plays an important role in the water balance of hydrological catchments or landscapes 49,50 . Similar to our approach, previous studies assessing forest canopy transpiration or evapotranspiration at the water catchment level used average species sap flux values and/or ET estimates per unit ground area across habitat types to scale-up to the landscape level 51 ). Fractional cover maps, as the one used in this study 26 reflect not only the current distribution but also the different levels of invasion across the region and are thus more accurate in upscaling impact at the water catchment or ecosystem level than presence/absence maps 52 . The Prosopis map we used in the present study for upscaling water consumption by Prosopis was developed using different driving factors (environmental, climatic, remote sensing and other important factors for its distribution) implemented on machine learning algorithms as species spatial patterns are important for understanding spatial estimates of transpiration. This is consistent with Loranty et al 51 that incorporated atmospheric and edaphic drivers to upscale sap flow from trees and ET from stands level to an aspen-dominated forest across a upland-towetland gradient, and doing so in a spatial context, that achieved a more accurate estimate.
The findings of this study indicate that between 8.3-9.1 billion liters of water per day are consumed by Prosopis in the invaded areas of the arid and semi-arid Afar Region. Hence, this fast-growing exotic tree is likely to exacerbate the effects of climate change on the provisioning of ecosystem services through its impact on ecosystem water budget. Based on the extremely low current water tariff for water by the Awash Basin Authority, which can be considered as a payment for ecosystems service by investors, the economic value of water used by Prosopis potentially amounts to US$ 465,000 -500,000 per year across the invaded range, which is equivalent to approximately 15 www.nature.com/scientificreports/ Afar in terms of market prices of cotton and sugar cane production capacity, the value of the water consumed by Prosopis is much higher than that estimated from the water tariffs. As noted by Le Maitre et al 36 , there is a lack of studies assessing the financial impacts of IAPS on yields, which may lead to an underestimation of the economic benefits of managing IAPS for society 53 . Up-scaling of the water use of Prosopis to the regional scale has immense implications to decision makers and land use planners. The Awash River is known to be a losing river system as the water abstracted by vegetation, mainly crop plants and IAPS, can be regarded as an abstraction from the river recharging system to the groundwater resource. Extensive abstraction of groundwater has serious consequences on the water availability to downstream, as well. Moreover, this study revealed that Prosopis uses the groundwater throughout the year as the annual rainfall is by far less than the water consumed by Prosopis, which affects water availability to other services by influencing ecosystem water budget that could be used for the provision of other ecosystem services and thus for supporting rural livelihoods, including crop production or fodder for livestock. Dzikiti et al 22 showed that the decline in groundwater level under Prosopis stands can be attributed to the active water uptake by the trees to meet their transpirational demand. In addition, dense thickets of thorny Prosopis trees create physical barriers for domestic and wild animals to access surface water 19,54 . Moreover, Prosopis affects water availability for indigenous trees and herbaceous species, which also contributes to a degradation of ecosystem functioning and the provision of ecosystem services 55 . For example, indigenous palatable grasses species such as Chrysopogon plumulosus (Durfu), Cenchrus ciliaris (Serdoitas) and Seataria acromelaena (Mussa) and multipurpose trees species such as Acacia tortilis (Eebto), A. senegal (Adebo), A. nilotica (Keselto) and Dobera gelabera (Gersaito) are threatened by Prosopis invasion 18,56 . Thus, there is a need for sustainable management of Prosopis invasion in order to secure benefits from groundwater resources and other services provided by the ecosystems in this semi-arid and arid region.
We acknowledge that the up-scaling of the water use of Prosopis to the regional scale is based on a limited number of experimental sites and some assumptions that entail a certain level of uncertainty. All the four experimental sites were situated in the southern part of Afar Region and may therefore not be representative of the total area invaded by Prosopis (Fig. 1). However, climatic conditions are similar across the lowlands of Afar Region, i.e. the area invaded by Prosopis, with the same rainy seasons and comparable average temperatures across the year. Another caveat of our study is that the upscaling of ETa was based on measurements taken at one experimental site only. As it was not feasible to measure ETa at all four experimental sites, we triangulated it with sap flow values at site 1 and found that our ETa values were reasonable. Nevertheless, investigations on the variation in water use of Prosopis across the invaded range in Afar Region as well as in neighboring countries would further improve our understanding of the impact of Prosopis on the water balance of semi-arid and arid ecosystems in the tropical regions of Eastern Africa.

Conclusions
While Prosopis trees were, among others, introduced in Eastern Africa to increase the availability of wood and to stabilize soil in degraded ecosystems, their invasion has been identified as a serious challenge to biodiversity, environment and the rural livelihoods 15,18,20,30,54,55 . In the context of climate change and an increasing frequency of drought events in dry regions of Sub-Saharan Africa, there is increasing evidence that invasive trees can increase transpiration and ET losses and thus reduce surface water or groundwater recharge 36 . The results of this study provide evidence that in the Afar Region the evergreen P. juliflora not only consumes water throughout the year, but even consumes more water during the dry than during the rainy season. Thus, unless the spread of P. juliflora is contained and the density reduced in areas where it has become established, this invasive tree is likely to have serious consequences for sustainable livelihoods in the region. The estimated net benefit alone would strongly justify the implementation of a coordinated control program.