Partitioning daily evapotranspiration from a marsh wetland using stable isotopes in a semiarid region

Wetlands are important ecosystems that provide many valuable ecological services, including habitat for species, flood peak attenuation, water purification and climate change mitigation, among others (Woodward & Wui 2001). Hydrological properties are the driving factors of a wetland ecosystem (Mitsch & Gosselink 2007), and particularly, evapotranspiration (ET) plays an important role in the ecosystem water budget and energy balance (Hu et al. 2009) and affects the composition, species diversity and succession of wetland vegetation communities. In wetlands, ET causes major losses of water from the open water surface by evaporation and from the transpiration of emergent plants, which influence regional water cycling (Pauliukonis & Schneider 2001; Chen et al. 2002); therefore, accurate estimates of ET rates for wetland management and water resource assessment are essential. In the past, accurate water balance estimates could not be obtained easily because ET information from vegetation was lacking (Fermor et al. 2001). As a result, the effectiveness of many wetland management practices was low (Peacock & Hess 2004). Currently, worldwide, much research is focused on the contribution of Phragmites australis to ET, including in Australia (Headley et al. 2012), America (Drexler et al. 2008), Spain (Sánchez-Carrillo et al. 2004), Italy (Borin et al. 2011), England (Fermor et al. 2001; Peacock & Hess 2004), Germany (Herbst & Kappen 1999), Canada (Lafleur 1990) and China (Zhou & Zhou 2009; Yao et al. 2010; Xu et al. 2011). Multiple technologies have been used and include the Bowen Ratio Energy Budget (BREB) method (Peacock & Hess 2004), measurements of sap flow (Moro et al. 2004), the eddy covariance (EC) method (Zhou & Zhou 2009), remote sensing (Yao et al. 2010) and modelling (Herbst & Kappen 1999). Nevertheless, little research has focused on the hourly partitioning of ET using a stable isotope method in a P. australis marsh wetland.

Motivated by this gap in research, the daily consumptive use of water by P. australis and the partitioning of ET in the Momoge National Nature Reserve, a semiarid marsh wetland in northeastern China, were determined using a stable isotope method. The objectives in the current study were to (1) investigate the isotopic characteristics of the water pools and fluxes at the site, (2) determine the relative importance of different components to ET and the daily water use by P. australis and (3) estimate the daily ET distribution on an hourly scale.

The Momoge National Nature Reserve (45.9359° N, 123.6839° E) is located in the northern Songliao Plain and at the western edge of the Songnen Plain in the northwest of Jilin Province, China. The total area is approximately 1,440 km², of which 90% is wetland. The climate is semiarid with an average annual rainfall of 392 mm. The annual average temperature is approximately 4.2°C. The summers are hot and rainy, and the winters are extremely cold. Eighty per cent of the reserve is composed of the inland saline wetlands that are typical of this region of China (Figure 1) (Ming et al. 2007). The sampling site was dominated by a dense stand of P. australis of approximately 10 ha (50 m width and 2,000 m length), within which all vegetation, leaf area index (LAI) and water depth were measured in a vegetation survey to determine species composition, plant height and plant density.

In this study, many biological and meteorological factors were measured, and the parameters of P. australis included the leaf water content (W), specific leaf area (SLA), LAI and leaf-level gas exchange properties. The meteorological factors included water level fluctuation, in addition to air temperature and relative humidity and water surface temperature, which were monitored using an HOBO series temperature and humidity recorder (ProV2, America) and water thermometer, respectively.

The SLA was calculated from the dry weights of 10 pieces of 0.9 cm² punched from leaves during the day of sampling. The W was determined from the difference between the fresh and the dry weights of the leaves from five plants and the measured mean SLA of foliage. The LAI was determined by counting the number of individual plants within three replicate 1 × 1 m plots, sampling 20 individual plants, determining the dry weight and then calculating the leaf areas using dry weights and SLAs. Leaf-level gas exchange properties, including stomatal conductance, transpiration and leaf temperature, were measured using a LI-6400 analyser (Li-Cor Inc., Lincoln, NE, USA) on three leaves of individual plants. Table 1 shows the results.

In the wetland ecosystem, samples of different water pools were collected five times on two consecutive days, 20th and 21st June 2013, which included surface (0–5 cm) standing water, stem water below the water level and leaf water for P. australis, and atmospheric water vapour at three heights (20, 150 and 210 cm) above standing water. Additionally, the atmospheric background water vapour was collected at the height of 4 m above the ground with no vegetation approximately 100 m from the sampling site.

Standing water samples were collected directly using 2-mL glass vials 0–5 cm below the water level. Stems of P. australis were sampled near the roots and below the surface of the standing water. For the collection of leaf samples, the upper and lower leaves of the plants were mixed. Standing water, stem water and leaf water samples were all collected in triplicate, and all samples were sealed immediately with Parafilm in 2-mL glass vials and then frozen below 4°C in the laboratory.

Based on the methods of Helliker et al. (2002), atmospheric water vapour was cryogenically captured and analysed for δ¹⁸O. Water vapour was collected at three levels above the standing water through a low absorption plastic tube at a flow rate of 0.5 L min^–1 for 30 min and condensed in a 15-cm-long glass tube placed in a −80°C ethanol/liquid nitrogen bath. The melted water was poured into 2-ml glass vials, which were sealed with Parafilm and stored in a refrigerator until analysis. Each sample was approximately 0.5–1 mL, which is sufficient for δ¹⁸O analysis.

Stem and leaf water samples were extracted in the laboratory using a cryogenic vacuum distillation apparatus (Ehleringer et al. 2000). Then, all samples, including water vapour samples, were analysed for δ¹⁸O composition using a Liquid Water Isotope Analyser (Model DLT-100; Los Gatos Research Inc., USA). The precision for δ¹⁸O and δD analyses was better than 0.3‰ and 2‰, respectively (Sturm & Knohl 2010).

The scatter plot of δD versus δ¹⁸O is shown in Figure 2, and the correlation between δD and δ¹⁸O was described by δD = 3.53 δ¹⁸O−53.03 (r² = 0.88, n = 70) for the water pools and fluxes in the marsh wetland ecosystem. The data for the liquid water (standing water, stem water and leaf water) were all below the local meteoric water line (LMWL), and the data of vapour were all above the LMWL. These results suggested that the kinetic fractionation of H¹⁸O was higher than that of HDO (hydrogen + deuterium + oxygen) and that strong evaporation occurred at the study site because of the dry weather conditions, as indicated by the low slope of the scatter plot. The results for δ¹⁸O stable isotope were focused in this study to avoid redundancy. The primary biotic and environmental variables for the two sampling days are shown in Table 2, and the temporal variations of δ¹⁸O values of the water pools and fluxes are presented in Figure 3. The δ¹⁸O values of the surface standing water were nearly constant, with an arithmetic average (±standard deviation) of −4.5 ± 0.73‰. The stem water δ¹⁸O values (−5.50 ± 0.61‰) of P. australis were similar to those of the standing water. However, the leaf water showed a clear isotopic enrichment relative to the source (stem) water (Figure 3).

Using the measured or modelled isotopic end members (i.e., δ_c, δ_b, δ_E and δ_T; Table 2) and the measured leaf-level transpiration, T/ET was estimated based on the mass balance equations (Table 3). To evaluate the results of ET partitioning, two different methods were used to estimate the ratio of transpiration to ET: (1) based on the steady state estimates of δT and the P&G method (ISS method) and (2) based on the NSS estimates δT and the P&G method (NSS method). Using the NSS method, T/ET ranged from 0.19 ± 0.12 to 0.67 ± 0.39 with a mean of 0.38 ± 0.18 (Table 4). The ISS and NSS methods produced very similar results, and the T/ET estimate from the NSS method was approximately 0.05-fold higher than that from the ISS method. Using the estimated T/ET from the NSS method, the average estimated total water loss from E, T and ET was 6.29, 2.46 and 8.76 mm d⁻¹, respectively (Figure 4).

The hourly ET variation was expressed as a single-peak curve that was relatively low in the morning and evening and high at approximately noon (Figure 4). With the increase in net radiation, air temperature and water surface temperature, the ET amount dramatically increased from 0.41 mm in the morning to 3.75 mm at noon, an increase of approximately nine-fold. The T/ET ratio change appeared as an ‘N’ line, with the peak value of approximately 0.53 before noon, the second highest value of approximately 0.43 in the morning, and the minimum of approximately 0.21 at noon. The phenomenon indicated that the transpiration of P. australis increased with the increase of environmental factors (such as net radiation, air temperature and water surface temperature) faster than water evaporation. After noon, the water evaporation decreased faster than P. australis transpiration with the descendant coefficients of 1.287 and 0.317, respectively.

Previous studies suggest that not accounting for the NSS of δ_T can introduce large errors in partitioning of ET (Yepez et al. 2005; Lai et al. 2006). In this study, comparisons between the methods for partitioning ET showed that the results for ISS and NSS methods were very similar in estimating fractional contributions from E and T_ph (Table 4). Therefore, although the transpiration was not at an ISS (Table 3; Figure 4), the ISS method also gave reasonable estimations of contributions from E and T_ph. However, slight discrepancies (less than 5%) were detected for the early morning and evening samplings between the two methods; therefore, consideration of a NSS δ_T in modelling of leaf water for the transitional periods of the day would be beneficial.

Before modelling the NSS δ_T, the leaf water isotopic enrichment model should be validated using the observed values. The temporal variations of the modelled and measured bulk leaf water ¹⁸O enrichments (Δ_Lb) and the scatter plots of modelled versus measured values showed that in spite of a somewhat underestimate, the modelled Δ_Lb agreed well with the observed Δ_Lb (Figure 5). Two thresholds can be used to evaluate the departure of leaf water enrichment from a steady state: (1) Δ_Le−Δ_Les, the difference between NSS leaf water enrichment at the evaporating sites (Δ_Le) and the steady state values (Δ_Les) (Welp et al. 2008); (2) δT−δs, the difference between transpiration δ_T and the source water δ_s (Harwood et al. 1998). The results showed that using the first threshold, either from the forward prediction using Equation (21) in the Farquhar & Cernusak (2005) (herein F&C) model (Δ_{Le, F&C}) or from the back calculation using the observed bulk leaf water Δ¹⁸O based on Equation (16) in Farquhar & Cernusak (2005) (Δ_{Le, back}), leaf water enrichments were near but deviated from a steady state on both the 20th and 21st June (Figure 5). When using the second threshold, the modelled δ_T always deviated from the observed δ_s, indicating the necessity of modelling the NSS δ_T.

The isotopic composition of leaf water was calculated using the F&C model, which considers both the NSS and Péclet effect, and then compared with that of the observed leaf water. Previous studies show that considering the Péclet effect can improve the agreement between modelled and observed leaf water (Farquhar & Cernusak 2005; Welp et al. 2008). In this study, the modelled dΔ_Lb values were very consistent with the measured dΔ_Lb values, in spite of the underestimations, particularly in the transition periods of evening and morning (Figure 5). In such transitional periods, the observed Δ_Lb values were higher than those of Δ_Le, providing support for the argument that isotopic exchanges of H₂¹⁸O occur between leaves and the air (Farquhar et al. 2007; Welp et al. 2008).

In this study, both the modelled Δ_Le (Figure 5) and the calculated δ_T (Table 3) showed that the leaf transpiration was not at an ISS. The diurnal changes of δ_T values during the study time increased more rapidly than those of δ_x, which is similar to the pattern documented in earlier studies (Harwood et al. 1998; Yakir & Sternberg 2000; Welp et al. 2008). However, notably, the calculated δ_T on the early morning of the 21st June showed very high values during the day, which was caused by the high air humidity (RH= 77.1%) at the time of sampling, because δ_T can be positively correlated with air humidity instantaneously (Lee et al. 2007). Therefore, at this humid site at which the air was prone to saturation in the early morning and evening, the NSS in the modelling of leaf water and transpiration must be considered.

Many factors affect ET, including meteorological factors, plant physiological and ecological characteristics, the underlying surface conditions and other factors, with complex variations (Li et al. 2000). In practice, the possible factors are net radiation, air temperature, surface temperature, relative humidity, wind speed and some others (Si et al. 2005). Because of the short experimental time, the results might include some deviation. Therefore, we compared the meteorological data for 9 consecutive sunny days (Table 5). From the contrast of meteorological factors, the relative humidity in the study period was higher than that for the consecutive days in late June and the air temperature was lower by approximately 2.1°C, which resulted in the evaporation difference. The evaporation difference between the experimental period and late June was approximately 3.3%, less than 5%. Thus, the results for the study period were reasonable and credible.

To determine the factors influencing ET in wetlands in northeast China and to further validate the results of this study, we contrasted the ET amount and meteorological factors at wetlands of international importance that are distributed in northeast China (Table 6). Based on the amount of ET, these wetlands were divided into two groups: one group (first group), including the Zhalong and Momoge Wetlands, was distributed in a semiarid area, and the other group (second group) contained the Sanjiang and Panjin Wetlands located in a humid region.

ET is highly dependent upon net radiation (R_n) and air temperature, and peak values of ET often mirror the peaks in R_n or air temperature (Zhou & Zhou 2009). Moreover, the R_n is significantly positively correlated with the sunshine duration. From Table 6, no clear difference is observed in sunshine duration between the two groups; therefore, R_n may not be the primary reason for the difference in ET values. However, the air temperature of the first group was dramatically higher than that in the second group. In previous studies, the ET increased with increasing air temperatures, and the correlation coefficient was approximately 0.693. Thus, the air temperature was likely one reason for the differences in ET. Additionally, the relative humidity in the first group was 16.8-fold lower than that in the second group, a clear difference. The relative humidity reflects the influence of the vapour pressure deficit (VPD), with the relative humidity expressed as the ratio between the actual water vapour pressure and saturated vapour pressure at the same temperature (Huang & Liu 2000). Therefore, the air temperature and VPD or relative humidity were likely primary factors that caused the difference in ET between the two groups. The wind speed is a factor that also simultaneously affects the ET amount (Wang et al. 2007; Drexler et al. 2008). In the first group, the ET of the Momoge Wetland was larger than that of the Zhalong Wetland by approximately 0.53 mm/d and the wind speed also increased by approximately 1.3 m/s. In the Panjin Wetland, wind speed promoted and accelerated ET, with a correlation coefficient of approximately 0.477. Thus, overall, air temperature, VPD (or relative humidity) and wind speed are likely the primary factors leading to the differences in ET of wetlands in northeast China.

The use of stable isotope techniques to partition ET fluxes into evaporation and transpiration over a marsh wetland was the focus of this work. An experiment was conducted in the growing season in the semiarid area of the Momoge Natural Reserve. Isotope turbulent mixing relationships, isotopic values of transpired water from plants and those of evaporating water vapour from the wetland were used to estimate the transpiration and evaporation fractions contributing to the total ET. Three important conclusions were derived from this work. (1) The slope of the fitted curve for delta D and delta ¹⁸O was approximately 3.53 in the marsh wetland, indicating strong evaporation of surface water in the arid climate (including plant transpiration and water surface evaporation). (2) Using the NSS method, for the proportions of ET, water surface evaporation accounted for approximately 62% and plant transpiration was approximately 38%. The NSS method effectively improved the accuracy of each, particularly in the morning and evening. (3) According to E and T relative contributions and leaf transpiration rate (T_r) measurements, the estimated ET flux was approximately 8.76 mm. In a comparison of wetlands in northeast China, the air temperature, VPD (or relative humidity) and wind speed were likely the primary factors affecting the differences in ET. The water loss from ET is a useful measurement for wetland restoration and wetland water management, and these findings will be helpful in organizing a strict wetland water management plan for semiarid areas.

The National Science Foundation of China (NSFC41101091, NSFC 41371193, NSFC 41471080 and NSFC 41301606) and the Science Foundation of Jilin Province (No. 20150101080JC) supported this study. We thank the anonymous referees for their valuable time.