Analysis on water use efficiency of Populus euphratica forest ecosystem in arid area

Water use efficiency (WUE, the ratio of gross primary productivity (GPP) to evapotranspiration (ET)) reflects the coupled relationship between water loss and carbon gain in the process of plant photosynthetic carbon assimilation. As a dominant tree species in arid area, Populus euphratica plays an important ecological role in slowing desertification. Here, continuous observations of carbon, water, and energy fluxes were carried out in Populus euphratica forest with eddy covariance (EC) technique in 2018. We systematically explained the variation characteristics of energy fluxes and WUE at different time scales, and explored the main controlling factors of WUE in drought-stressed environment based on the synchronous meteorological data. Results showed that the carbon exchange of the Populus euphratica forest ecosystem occurred mainly during the growing seasons (April–October). During this period, the entire ecosystem appeared as a carbon sink with the potential to sequester atmospheric carbon dioxide. The average daily WUE was 2.2 g C/kg H2O, which was lower than other temperate forests (2.57–6.07 g C/kg H2O) but higher than grassland, wetland, and cropland. We also concluded that an increase in carbon dioxide concentration (CCO2) and air relative humidity (RH) could promote the increase of WUE. Nevertheless, WUE was negatively correlated with air temperature (Ta), photosynthetically active radiation (PAR), and normalized difference vegetation index (NDVI). Additionally, WUE increased under moderate soil water content (SWC), but decreased due to the continuously rising SWC. WUE was more strongly affected by factors affecting water consumption than carbon uptake. Under the conditions of high temperature, strong radiation and low humidity in the summer, the growth rate of ET was much larger than that of GPP. This study not only contributes to our understanding of the carbon, water, and energy dynamics of the Populus euphratica forest ecosystem but also provides an important reference for ecological conservation and ecological restoration in arid regions.


Introduction
As global warming intensifies, climate change and its impact on terrestrial ecosystems is receiving increasing attention. Carbon and water cycles are key processes for maintaining nutritional state and energy balance of ecosystems, which are more deeply affected by climate change (Seddon et al. 2016). The development and maturity of eddy covariance (EC) technique enables long-term and continuous measurement of carbon, water, and energy exchanges between land and atmosphere (Amiro et al. 2006;Yu et al. 2006). The global FLUXNET is conducting continuous observations of the carbon, water, and energy exchange processes of terrestrial ecosystems. But there are fewer EC sites for arid area than the southern wet area (Kim et al. 2013). While water is the principal limiting factor in arid area, controlling plant growth and vegetation succession, better insight into the relationship between limited water resources and ecosystem water use efficiency (WUE) would enhance our understanding of plantation-ecosystem processes, services, and feedbacks to the regional climate system (Tong et al. 2014).
In the terrestrial ecosystem, carbon and water cycles closely couple because they both exchange between biosphere and atmosphere via the same pathway, namely the stomata (Yu et al. 2008). Plant photosynthesis absorbs CO 2 along with water loss to regulate the carbon-water balance between ecosystems and the atmosphere (Mammarella et al. 2015). In the same sense, WUE reflects the magnitude of fixed CO 2 when plants consume unit mass of water, and reveals the exchange law of CO 2 and water vapor fluxes in ecosystems and their coupling cyclic relationship (Hui et al. 2001). WUE refers to the ratio of ecosystem carbon flux to water vapor flux. This concept has been extensively used in agronomy, plant physiology, and ecology since its introduction (Jin et al. 2017). Initially, studies of WUE were confined to the leaf physiological or individual levels, with the aim of breeding good crops or directing field management (Keenan et al. 2013). Later, ecologists explored the mechanism of biological invasion and plant adaptation strategies to the environment by studying the WUE (Xu and Hsiao 2004). In recent years, as global climate change issues become more prominent, more perspectives have been directed to native ecosystems such as grasslands and forests. Research scales have also risen to canopy, ecosystems, and landscape levels (Zhao et al. 2007). Many scholars have carried out a large number of WUE researches, involving different spatial scales such as tissues, populations, communities, and regions (Jassal et al. 2009;Skubel et al. 2015;Öquist et al. 2017;Stephens et al. 2018). These studies serve to reveal the interactions of water-carbon cycles, to predict impacts of global change on ecosystem functions (Jin et al. 2017). In the context of climate change, the response of the carbon cycle and water cycle in the arid region of China has attracted special attention (Ma et al. 2019). Some studies have found that drought can reduce WUE (Brümmer et al. 2012;Zhang et al. 2014), while others have found that it can increase WUE (Ponton et al. 2006;Linderson et al. 2012). However, these studies still lack a scientific understanding of the physiological mechanism of the seasonal change of ecosystem WUE during drought period. Studies have shown that the duration of drought also has different effects on WUE (Hui et al. 2001). This is because the impact of drought on ET is long term, including vegetation growth season and subsequent non-growth season (Yu et al. 2008;Govind et al. 2011). Obviously, the ecosystem WUE in arid areas needs more in-depth research.
In arid region, Populus euphratica plays an important role in the ecological protection of arid regions, which can prevent wind and immobilize sand effectively (Chen et al. 2017). In the world, 54% of Populus euphratica is distributed in China, while in China, more than 90% of Populus euphratica resources are concentrated in Xinjiang. Currently, the research of carbon and water cycles on Populus euphratica forest ecosystem mainly focuses on biomass or a certain component of Populus euphratica, such as forest soil, plant body, or leaf. In addition, the observation time is short and discontinuous. Therefore, the process and mechanism of carbon, water, and energy exchanges cannot be seriously studied. Based on the aforementioned scholars' research, we used the CO 2 flux (Fc) to directly estimate the respiration, and solved the problem that the respiration is difficult to be measured. In addition, we compared WUE in the major types of ecosystems around the world. This is an important reference for analyzing whether a Populus euphratica ecosystem is carbon source or a carbon sink, and predicting the impact of global changes and human disturbances on forest carbon budgets.
This study took the Populus euphratica community in the lower reaches of the Tarim River as the research object. Continuous observation of CO 2 , water, and energy exchanges was carried out using EC technique. The main objectives were (1) to clarify the daily and seasonal variations of carbon, water, and energy fluxes caused by diurnal cycles and seasonal changes; (2) to study the carbon exchange process and carbon source/sink changes; and (3) to determine the change law of ecosystem WUE, and analyze the influence mechanism of environmental factors on WUE by combining hydrometeorological elements.

Study site
The study site is located in Ruoqiang County (40°27′10.2′′N, 87°54′02.9′′E, 842 m above mean sea level, AMSL), Xinjiang, China. The EC station is located in the Populus euphratica forest on the bank of the Tarim River (Fig. 1). The climate in this region is temperate continental climate with annual precipitation between 27 and 58 mm. The annual potential evaporation amounts to 2671-2902 mm. The temperature in January and July is −14.9°C and 33.6°C, respectively. Annual sunshine hours are 3082.7-3121.3 h (Wang and Guo 2018). The water table is roughly between 2.8 and 4.5 m. The soil is dominated by sand (0-100 cm depth) and sandy loam soil (100-200 cm depth). The zonal vegetation in this area is temperate semi-arbor desert. Due to the accumulation of river water and groundwater, the floodplain develops a certain area of non-zonal meadow vegetation, forming arid riparian forest composed of trees, shrubs, and herbaceous plants. Trees and shrubs are distributed along the river, and herbaceous plants are aggregated in the shrub community. Main dominant plants are Populus euphratica (the only group of trees), Tamarix, Karelinia caspica, Halimodendron halodendron, and other shrubs. Average height of Populus euphratica is 6-14 m, and the average forest age is about 50a.

Data collection and processing
Water vapor and CO 2 fluxes were measured at a height of 15 m above the ground surface by an eddy covariance (EC) system consisting of an open-path infrared gas analyzer (EC150) and a three-dimensional sonic anemometer (CSAT3). The EC system performed real-time monitoring for 24 h and automatically acquired raw data at a frequency of 10 Hz. It was further equipped with meteorological element monitoring sensors, including air temperature and humidity sensor (HMP155A), four-component radiometer (NR01), rain gauge (TE525MM), soil heat flux sensor (HFP01), photosynthetically active radiation sensor (LI190R), and soil moisture meter (SDI-12). The soil moisture meter (SDI-12) monitored the soil water content (SWC) at 20 cm, 40 cm, and 60 cm below the surface. We used SWC data at 20 cm for calculation in this paper.
The datalogger (CR3000) collected horizontal wind speed (u, v), vertical wind speed (w), supersonic temperature (Ts), H 2 O concentration (q), and CO 2 concentration (CCO 2 ) at the sensor height. The final flux data, including CO 2 flux (Fc), net radiation (RN), latent heat flux (LE), sensible heat flux (H), and soil heat flux (G), were calculated and recorded by CR3000 at a frequency of 30 min. In the flux calculation process, double rotation, humidity correction, and WPL density adjustment were performed (Foken et al. 2004). The stationary, integral turbulence characteristics of each 30 min flux data were calculated to filter out the periods of poor turbulent development (Zhao and Liu 2018).
Due to the adverse weather, the flux data inevitably had outliers and missing values. Two methods were used to interpolate the missing and culled fluxes data: (1) mean diurnal variation method for gaps less than 7 days-missing values can be replaced by the average of adjacent days in the same time.
(2) Nonlinear regression method for gaps more than 7 days: the principle of this method is to establish a regression relationship between environmental factors and fluxes using observation data, and then use the regression function to estimate the missing data.

Energy balance closure
Energy balance closure is an indicator of flux data quality assessment (Meyers and Hollinger 2004). Conventional methods for evaluating energy closure are as follows: ordinary least squares (OLS) linear regression, reduced major axis (RMA) linear regression, energy balance ratio (EBR), and energy balance relative residual (Yang et al. 2018). Among them, the EBR is the most widely used method for analyzing the energy closure.
According to the first law of thermodynamics, the surface energy balance equation can be described by the following formula: where LE is the latent heat flux, H is the sensible heat flux, RN is the net radiation, G is the soil heat flux, S is the heat stored in the canopy, and Q is the sum of additional energy source/sink. For bare land or ecosystems with low vegetation, the values of S and Q are usually small (Helbig et al. 2016). So, the energy balance equation can be simplified to: In the above formula, the left side of the equal sign is turbulent energy, and the right is available energy. When turbulent energy (LE + H) is equal to available energy (RN-G), it is called energy closure. This shows that the EC system can accurately observe the turbulent energy. Otherwise, the EC system overestimates or underestimates the turbulent energy. In this study, EBR was used to evaluate the energy closure condition, defined as: 2.4 Calculation of WUE is the ratio of carbon flux to water vapor flux, that is, the ratio of the CO 2 fixation to the evapotranspiration (ET) in a certain period of time. We used the ratio of gross primary productivity (GPP) to ET to define ecosystem WUE, expressed as: GPP was calculated with the following formulas: where NEP is net ecosystem productivity, which is equal to the negative value of the net ecosystem CO 2 exchange (NEE). CO 2 flux (Fc) measured by EC system is the net ecosystem CO 2 exchange; Re is total respiration of the ecosystem. Re was calculated with the following formulas: where R n is respiration at night. The net CO 2 exchange (Fc(night)) can be considered R n of the ecosystem (Fei et al. 2018 , and B = 0.044. We found that the turbulence at night was weak, and the F c data at 30-min frequency was noisy. Therefore, we used the average value of F c and Ta data at night to study the relationship between the two. As shown in Fig. 2, the correlation coefficient R 2 = 0.61.
ET was calculated with the following formulas: where LE is latent heat flux measured by EC system; λ is latent heat of vaporization with a fixed value of 2.454 × 10 6 W·m −2 ·mm −1 .

Dynamics of environmental factors
We used auxiliary sensors in EC system to measure the environmental factors. The changes of these environmental factors during the study period (2018.1.1-2018.12.31) are shown in Fig. 3. The seasonal variation characteristics of Ta ( Fig. 3(a)), PAR ( Fig. 3(b)), and SWC ( Fig. 3(c)) were basically the same, with the lowest in winter at −14.6°C, 95μmol/(m 2 ·s) and 6.7%, and the highest in summer at 34°C, 651μmol/(m 2 ·s) and 10.6%. PAR showed low values for a few days in July due to the effects of rainy days. The amount of soil water replenished by rainfall was expected to be very limited since precipitation in the study site was scarce. There were two reasons for the high SWC in summer. One was the incremental water from the upper reaches of the Tarim River. The second was ecological water supply (Xue et al. 2019). Seasonal variation characteristics of CO 2 concentration ( Fig. 3(d)) and RH (Fig. 3( y=0.73exp(0.044(x-2.2)) R² = 0.61 Fig. 2 The relationship between CO 2 flux at night (Fc(night)) and air temperature (Ta). The downward shortwave radiation (DR) is used as an indicator to distinguish daytime and nighttime. It is generally considered that when DR < 1 W/m 2 , the corresponding period is nighttime.
concentrated in summer, so the CO 2 concentration in the air was relatively low. Due to the high temperature in summer, the air in the study site was dry, which was contrary to the conclusion that RH was superior in the south of China. In general, the water resource in the arid region is limited and the actual evapotranspiration is small (Zhu et al. 2018).

Evaluation results of EBR
The turbulent energy (LE + H) observed by the EC system usually cannot balance the available energy (RN-G). Wilson et al. (2002) analyzed the observation data from 22 stations in the Global Flux Network, indicating that 10-30% fluxes unclosed in all stations. The energy fluxes obtained by the EC system were lower than that from the radiometer. Some scholars (Stoy et al. 2013) have therefore started to suspect CO 2 flux measurement result. The phenomenon of non-closing and its mechanism have also become a hot spot in the study of turbulent flux. The half-hourly fluxes had mean EBR of 0.74 in this study (Fig. 4a), within the scope of international similar observations (0.55-0.99) (Wilson et al. 2002). Figure 4 b was an ordinary linear regression of turbulent energy and available energy. The regression coefficient was 0.74, which was equal to the mean EBR. The EBR pattern presented a general decrease from 9:30 to 20:00 in every season (Fig. 4c). During the period of sufficient radiation (13:00-18:00), the EBR was very stable, close to 1.
However, compared with the available energy measured by the radiation system and the soil heat flux measuring instrument, LE and H were underestimated. This was primarily due to the weak turbulence at night. Moreover, the canopy heat storage term was not considered, and the source region of the EC system was different from the radiation observation system (Ding et al. 2010). Figure 5(a) showed the monthly variation of energy fluxes. RN, as the outward driving force of the soil-vegetationatmosphere system, heats the atmospheric boundary layer mainly in the form of H and LE, and some energy in the form of G enters soil (Sánchez et al. 2009). The seasonal variation of LE, H, and G was basically the same as RN. After October, the air temperature decreased with RN. However, due to the coverage of fallen leaves, the soil temperature did not decrease as rapidly as the air temperature. The soil temperature was higher than the air temperature, indicating more heat stored in the soil, and the soil gradually released heat to the air. Since April, with the rapid increase of RN, the heat stored in the soil became lower than that of air. So, the atmosphere started to transmit heat to the soil. There were considerable differences in seasonal variations of LE and H. LE was greater than H in June-September and lower than H in other months. LE varied between 10 and 113 W/m 2 , reaching its maximum in July and close to 10 W/m 2 in December, January, and February. LE characterizes the amount of surface ET, whose energy exchange is linked to the phase transition of water. In summer, the temperature was high and the ET was concentrated in the study area (Yan et al. 2006). There was a significant seasonal variation in the ratio of every energy flux to RN, indicating that the proportion of energy consumed by vegetation life activities increased during the growing season, while the proportion of energy consumed in the physical environment decreased (Chan et al. 2018). From the perspective of diurnal variation in energy fluxes (Fig. 4b-e), each energy flux showed a unimodal curve and peaked around 14:00. The positive and negative changes in energy fluxes were linked to the sunrise and sunset. From January to March, RN was positive within 1 day for 10.5 h, with an average of 255 W/m 2 . From April to June, RN was positive for 12.5 h, with an average of 357 W/m 2 . From July to September, RN was positive for 10 h, with an average of 184 W/m 2 . From October to December, RN was positive for 8 h, with an average of 121 W/m 2 . LE and H had similar daily trends with RN. The peak time was basically the same as RN. One hour after sunrise, the soil-vegetation system began to assimilate energy. LE was on the brink of zero from October to March. This was explained by the fact that the ground was mostly frozen, and the water vapor transmission was very weak during this time. The peak value of LE in July was 215 W/m 2 . From the perspective of H, it reached a peak (254 W/m 2 ) in April. The positive and negative changes in G were affected not only by the sunrise time, but also by the soil. Owing to different soil physical and chemical properties in different seasons, soil thermal conductivity was different ). Therefore, the delay time of soil heat absorption and heat dissipation was also altered. Negative value of G indicated that heat radiated from the soil to the atmosphere and the soil was the heat source. Conversely, the soil was heat sink ).

Seasonal and diurnal variations of GPP, NEP, and Re
Carbon flux data was used to calculate the gross primary productivity (GPP), net ecosystem productivity (NEP), and respiration (Re) of the Populus euphratica in 2018. The seasonal variation characteristics are shown in Fig. 6(a). GPP was generally larger during the growing season (from April to October). During this period, the photosynthesis and respiration of the ecosystem were relatively active. During the nongrowth season from November to March of the following year, the temperature drop caused the vegetation to dry up, and the respiration of the entire ecosystem was much greater than photosynthesis. GPP of the Populus euphratica ecosystem was 877 g C/m 2 /year and the maximum rate was 3.40 μmol/(m 2 ·s), which appeared in June. The minimum production rate of GPP was 0.192 μmol/(m 2 ·s), which appeared in December. (c) Time Fig. 4 a Daily mean EBR in 2018, b relationship between the turbulent flux (H + LE) and available energy (RN-G), c diurnal changes of EBR in different seasons. The EBR was not meaningful during the morning and evening transition periods when the mean value of RN-G is close to zero NEP was positive from May to September, indicating that photosynthesis in the ecosystem was greater than respiration during this period, and the entire ecosystem behaved as a carbon sink. NEP of the Populus euphratica ecosystem was 345 g C/m 2 /year. The maximum production rate of NEP was 1.28μmol/(m 2 ·s), which appeared in June. The minimum production rate of NEP was −0.68μmol/(m 2 ·s), which appeared in March. The average production rate of NEP is 0.086μmol/(m 2 ·s).
The respiration intensity of Populus euphratica ecosystem was 532 g C/m 2 /year, accounting for 61% of GPP. The respiration intensity of ecosystem was greater than NEP. The seasonal change of respiration showed a single peak curve, which increased from February to August. The maximum rate of respiration rate was 2.51μmol/(m 2 ·s), which occurred in August, and the minimum value was 0.04 μmol/(m 2 ·s), which occurred in February. The average respiration rate of Populus euphratica ecosystem was 0.97μmol/(m 2 ·s).
The diurnal variation characteristics of GPP, NEP, and Re are shown in Fig. 6(b-m). Similar to the diurnal variation of energy flux (Fig. 5(b-m)), their variations also showed a single-peak curve. During January-March, April-June, July-September, and October-December, GPP and NEP from 12:00 to 19:00, 9:30 to 20:30, 10:30 to 21:00, and 10: 30 to 18:30 were a positive value. It indicated that the ecosystems absorbed carbon and the duration was 7, 11, 10.5, and 8.5 h. However, the peak time (15:00) was about an hour later than the energy flux (14:00).

Seasonal and diurnal variations of WUE
In the Populus euphratica ecosystem, ET was close to zero in winter (Fig. 7), so only the WUE of the growing seasons was calculated. During this period, the mean WUE was 2.2 g C/kg H 2 O and the maximum WUE was 5.5 g C/kg H 2 O. Due to the combined effects of photosynthesis and evapotranspiration, WUE did not peak when GPP and ET reached their highest values. WUE was opposite to the trend of photosynthetic and evapotranspiration. This showed that WUE of the Populus euphratica ecosystem was not necessarily high when GPP and ET were high. Under the conditions of high temperature, strong radiation, and low humidity in summer, the growth rate of ET was much larger than that of GPP.
The instantaneous WUE in daytime (8:30-20:00) was calculated using fluxes data at a frequency of 30 min (Fig. 8). We eliminated the data from 20:30 to 8:00 because WUE at night was of little significance, and the change rate of WUE was larger when the illumination was low. The daily variation of WUE in each month of the growing season was basically the same. WUE gradually increased with the enhancement of radiation (Fig. 5), reached a maximum at 14:00, and then gradually decreased.

Effects of climatic variables on WUE
Various environmental factors influenced WUE by controlling vegetation photosynthetic and transpiration (Kuglitsch et al. 2008). Both photosynthesis and evapotranspiration will increase with temperature. However, the growth rate is different, and the value of the ecosystem WUE determined by the two is also different (Wolf et al. 2013;Wu et al. 2017). In Populus euphratica forest, WUE became smaller as Ta increased ( Fig. 9(a)). This is because ET increased greater than photosynthesis with increasing temperature. It can be seen from Fig. 9(b) that WUE decreased as PAR increased. Solar radiation first affects the stomatal conductance of vegetation leaves, and then photosynthesis and transpiration change accordingly. Although PAR is the source of photosynthesis, the impact on transpiration is different. Plant photosynthesis only takes place within a certain range of radiation. Photosynthesis is limited with the increase of PAR, and there will be light saturation phenomenon. However, there is no upper limit for transpiration. Under sufficient water conditions, light will cause the temperature to rise, and transpiration will continue to strengthen.
The water for vegetation growth in Populus euphratica forest ecosystem mainly comes from soil water. Therefore, soil water is an important factor influencing WUE. Figure 9(c) showed that WUE increased as SWC decreased. The study area is in an arid area, so the vegetation often closes the leaf stomata due to lacking water. As a result, evapotranspiration decreased, and WUE increased accordingly. Moderate drought is beneficial to improve WUE (Krishnan et al. 2006;Yang et al. 2016), but if the drought continues to increase, the intensity of photosynthesis will decrease (Zhao et al. 2007). It is another matter to reduce or continue to maintain WUE under this result.
The increase of CO 2 concentration promoted the increase of WUE (Fig. 9(d)). There are two reasons. For one thing, increased CO 2 concentration will directly promote vegetation photosynthesis. For another, increased CO 2 concentration will cause the vegetation leaf stomata to close, and the water consumption will decrease. Thus, vegetation can convert more water into productivity (Forner et al. 2018).
RH was the main environmental factor affecting WUE ( Fig. 9(e)). Most studies have found a linear positive correlation between RH and WUE, but some studies have reached the opposite conclusion (Li et al. 2008). The reasons for the two diametrically opposite research results may be related to the soil water content. Figure 9(f) showed that WUE decreased sharply with the increase of P and then slowly decreased. Although the increase in precipitation could improve the GPP of the ecosystem, ET increased faster.

Evapotranspiration characteristics at different time scales
The main form of LE is water phase change, so the variation of ET is similar to the LE (Fig. 5(a)). The annual ET of the Populus euphratica ecosystem was about 345 mm. The evapotranspiration process mainly occurred from May to October at 286 mm, accounting for 93% of the annual evapotranspiration, indicating that the ET in this area was relatively concentrated. Vegetation of the Populus euphratica ecosystem has grown rapidly since April. As the temperature rises, the melting snow at the source of the Tarim River supplements the amount of water in the channel. In addition, summer rainfall is relatively high. River water and groundwater are abundant. These conditions are conducive to the growth of vegetation while also accelerating the evaporation and transpiration of the ecosystem.
The daily evapotranspiration of the Populus euphratica ecosystem was basically close to zero from November to March ( Fig. 10(b-m)). At the same time as the temperature decreased, it became difficult for vegetation to grow, and the water consumption of the ecosystem also decreased sharply. The maximum value of daily evapotranspiration occurred in July at 3.82 mm. The daily changes of evapotranspiration in July and August were basically similar, and both reached their maximum around 14:30 pm. Evapotranspiration lasts from 11:30 to 19:30 at a high value.

Comparison of WUE with other ecosystems
The daily average WUE (2.2 g C/kg H 2 O) of the Populus euphratica ecosystem was lower than other temperate forest ecosystems (2.57-6.07 g C/kg H 2 O) (Tan et al. 2015). When compared with forests of the same latitude, WUE of Populus euphratica forest was not as high as other forests in non-arid regions, but higher than grassland, wetland, and crop land (Fig. 11), because long-term drought enhanced the ecosystem's resistance and resilience to water scarcity in order to mitigate the effects of water loss (Gang et al. 2016).
In addition, we found that the WUE of Populus euphratica was relatively low, which was closely related to the local climatic conditions. During the growing season, although the ecosystem had the highest GPP, intense ET led to more water consumption while absorbing the same amount of carbon.

Latitudinal trends of WUE in different vegetation types
It is widely known that the vegetation types, stand structure, soil type, stand age, and geological evolution of terrestrial ecosystems vary widely. WUE had a clear zonal change trend (Fig. 10). As the latitude increased from tropical to high latitudes in the north, the average WUE of different ecosystems for many years has shown an upward trend, peaking at about 51°N, and then tending to decline at high latitudes. Although the WUE of different plant type ecosystems was different, all the peaks occurred within approximately the same latitude range (45°N-55°N). This indicated that there was a key zone of differentiation driven by radiation and moisture. In fact, latitude itself is not a driving variable, but a complex proxy factor under the combined action of abiotic and biological factors, which determines the main ecological zone (Tang et al. 2014).
However, in the case of the same latitude range, there were also differences in the size of WUE among different vegetation groups (Fig. 10). The WUE of evergreen forests was higher than that of deciduous vegetation at the same latitude. Broad-leaved forests had higher WUE than coniferous forests. The overall WUE of the forest ecosystem was higher than that of grasslands, wetlands, and crop land. Except for the wetland ecosystem, WUE of crop land was the lowest, which might be affected by the unproductive stage, and the soil evaporation was higher than that of permanent vegetation.   Fig. 9 Relationship between water use efficiency (WUE) and (a) air temperature (Ta), (b) photosynthetically active radiation (PAR), (c) soil water content (SWC), (d) carbon dioxide concentration (CO 2 ), (e) air relative Patterns in ecosystem carbon, water, energy fluxes, and WUE of Populus euphratica forest in northwest China were studied in light of combined influences of environmental variables and plant biological processes. The main conclusions are as follows: (1) The energy fluxes of the Populus euphratica ecosystem were not closed. The average EBR was 0.74, which was within the range of international comparable observations (0.55-0.99).  (3) The carbon exchange in the Populus euphratica ecosystem mainly occurred from April to October, and the whole ecosystem showed a strong carbon sink. From December to February, the ecosystem became a carbon source. Annual GPP, NPP, and Re were 877, 345, 532 g C/m 2 , respectively. The daily maximum ET appeared in July (3.82 mm) and the annual evaporation was about 345 mm. (4) During the growing seasons, the mean and maximum daily WUE was 2.2 and 5.5 g C/kg H 2 O. WUE decreased with increasing Ta, PAR, and NDVI. CCO 2 and RH could promote the growth of WUE. WUE increased under moderate SWC condition. Under the conditions of high temperature, strong radiation, and low humidity in summer, the growth rate of ET was much larger than that of GPP, causing WUE to drop to the lowest point. (5) WUE of the Populus euphratica ecosystem was lower than other temperate forest ecosystems, but higher than grassland, wetland, and crop land. By comparing WUE of various types of terrestrial ecosystems, we found that WUE exhibited significant spatial variability with geographical location and climate, but the peaks of different plant functional types occurred within approximately the same latitude range (45°N-55°N).
Availability of data and material All data generated or analyzed during this study are included in this article.
Code availability Not applicable.

Declarations
Ethics approval Not applicable.
Consent to participate All the authors read and approved the final version of the manuscript. The authors agree to publish the article.

Consent for publication
The authors confirm that the work described has not been published before, and it is not under consideration for publication elsewhere.

Conflict of interest
The authors declare no competing interests.