Evaluating the influence of water table depth on transpiration of two vegetation communities in a lake floodplain wetland

Wetlands provide irreplaceable ecological values in supplying water resources, providing wildlife habitat, and maintaining biodiversity (Gilliam 1994; Baschuk et al. 2012). Despite the importance of these ecosystems worldwide, wetlands have been seriously threatened by droughts under the impacts of climate change and human exploitation (Mortsch 1998; Legesse et al. 2004; Candela et al. 2009). Wetland degeneration caused by water table drawdown, drying of rivers, and rainfall reduction has become an increasingly acute problem (Pattern et al. 2008; Zhang et al. 2008; Johansen et al. 2011). To protect these wetlands, better understanding is required of the mechanisms mediating hydrological processes and plant water use.

Water is fundamental for plant growth. It directly affects species germination, growth, and biomass accumulation (Fay & Schultz 2009). Rainfall, groundwater, and surface water are major water sources for vegetation transpiration. In areas with a shallow water table, upward movement of groundwater into the root zone plays an important role in supplying water to plants (Hurst et al. 2004; Satchithanantham et al. 2014). Previous studies focused on the effect of groundwater contributions to plant water use have been mainly conducted in arid regions (Yang et al. 2007; Zhu et al. 2009; Soylu et al. 2011); in humid areas, the role of groundwater has often been neglected, given the high level of rainfall. In fact, since the seasonal distribution of precipitation and transpiration are usually uneven, plants in humid areas may also experience water stress in the dry season when they have to depend heavily on other water sources for growth (Li et al. 2013; Blanken 2014).

Generally, wetland plants also experience periodic drought stress, especially in floodplains and river-connected lakes, as the habitat changes seasonally from aquatic to terrestrial environments due to variations of water table (Wilcox et al. 2002; Pagter et al. 2005). Water table depth (WTD) significantly influences vegetation water supply when wetlands change to terrestrial environments (Zhang & Schilling 2006; Xie et al. 2011), and WTD has been verified to be strongly related to vegetation characteristics at different scales. At the population scale, WTD significantly influences vegetation growth and morphology characteristics (Vretare et al. 2001). At the community scale, the distribution patterns of biomass and species richness have been found showing linear or unimodal relationships along WTD gradient (Dwire et al. 2006; Sorrell et al. 2012).

The aforementioned studies mainly focused on external plant characteristics in exploring the effect of WTD, while the internal processes, such as vegetation water use, have received less attention. As has been demonstrated, vegetation water use is one of the main processes resulting in adjustments in plant morphology and changes in distribution patterns (Chaves et al. 2003; Cooper et al. 2006; Muneepeerakul et al. 2008). In riparian and floodplain wetlands, groundwater is a significant water source for vegetation use (Mueller et al. 2005; Xie et al. 2011). Sanderson & Cooper (2008) reported that a wet meadow obtained about 75%–88% of its annual water requirement from groundwater, with mean daily depth ranging from 0 to 1.2 m in a floodplain wetland. In periodic flooded wetlands of the Laurentian Great Lakes, Mazur et al. (2014) found that groundwater varying within −0.1–0.6 m depth contributed 59%–75% of the annual evapotranspiration. A decline in the water table will directly reduce plant water use, and over the long term might result in vegetation succession. Zhang & Schilling (2006) reported that water use by Phalaris arundinacea decreased exponentially from 7.6 mm/d to zero as the water table declined from the ground surface to 1.42 m below ground in a floodplain wetland. In a riverine wetland in Colorado, Cooper et al. (2006) found that when a water table that ranged from 0.42 to 1.25 m depth was lowered to vary from 2.24 to 2.61 m in depth, a decrease of 32% of vegetation evapotranspiration resulted, leading to a consequent shift of dominant species from wetland grasses to shrubs.

The effects of groundwater availability on transpiration has been explored mostly in riverine wetlands where the water table varies typically within 3 m depth, and the effects occur primarily on grasses and meadow ecosystems (Kahlown et al. 2005; Yang et al. 2007; Sanderson & Cooper 2008), but knowledge is limited on the connections between groundwater and transpiration in wetlands with large variations in WTD, and little is known regarding the effects of these processes on mesophytes and hygrophytes. Upward fluxes of groundwater into the root zone depend on many factors, including WTD, soil hydraulic properties, vegetation types, evapotranspiration demand, and root development (Shah et al. 2007; Luo & Sophocleous 2010). Consequently, the response of different plants may differ in relation to various WTDs and soil conditions. The effect of water tables with large ranges of fluctuation on vegetation transpiration remains poorly understood.

Poyang Lake, the largest freshwater lake in China, is naturally connected to the Yangtze River (Liu et al. 2000). The lake water level fluctuates seasonally by up to 10 m under the influence of catchment inflows and discharge to/from the Yangtze River, which creates an approximately 3,000 km² grass-covered wetland (Li et al. 2014). Since over 55% of the precipitation in Poyang Lake falls from April to June, the upward flow of groundwater is crucial in satisfying plant water demand during the main growth period from July to October (Liu et al. 2000; Hu et al. 2010). However, over the past decade, drought events have occurred frequently in the lake (Lu et al. 2011; Zhang et al. 2012b). The high lake level (in July–August) is not as high as normal and the receding level (in September–October) is lower than normal (Min & Min 2010; Zhang et al. 2014).

Consequently, wetland vegetations are confronted with serious water stress, especially in autumn. Substantial degradation of upland communities was noted in Hu et al. (2010) and Zhong et al. (2014). The native area of Artemisia capillaris community showed some degree of desertification (Dronova et al. 2011; Zhou et al. 2011). In many of the alluvial delta wetlands of inflow rivers, the distribution of the Phragmites australis community has shrunk dramatically, and some dominant hygrophytes have been replaced by xerophytes (Yu et al. 2010; Wu et al. 2010). These ecological problems are greatly threatening the biodiversity of the Poyang Lake wetland system, and have motivated a number of studies exploring the relationship between water level and vegetation. Studies have shown that at the landscape scale, vegetation distribution in Poyang Lake wetland is mainly shaped by WTD and inundation duration (Hu et al. 2010; Zhang et al. 2012a; Tan et al. 2016). At the local scale, vegetation biomass was revealed to follow a Gaussian model along WTD gradient (Xu et al. 2015). Although the above studies have highlighted the influence of WTD on vegetation growth characteristics in Poyang Lake, questions remain regarding the effect of WTD on vegetation transpiration. In particular, exactly how the groundwater contribution to vegetation transpiration changes along with a decline in the water table is poorly understood.

Mathematical models are important tools that can be used to quantify wetland water movement under a wide range of conditions and scenarios (Schlegel et al. 2004; Xie et al. 2011). There are mainly two types of wetland models: conceptual models and mechanistic models. Conceptual models were popularly used for wetland water balance assessment by treating the study area as a whole and ignoring the internal physical processes (Hussey & Odum 1992; Deng et al. 2003). Comparatively, the mechanical models, based on the Darcy law and continuity equation, can describe the water movement processes more accurately. A number of specific wetland models, e.g., the WETLAND and MIKE-SHE, have been widely used to describe the processes of overland flow, subsurface flow, and channel flow in depressions and river riparian wetlands at regional scales (Refsgaard & Storm 1995; Mansell et al. 2000; Zhang et al. 2008; Hammersmark et al. 2009). Other mechanical models, such as SWAP and HYDRUS, are effective in simulating interface water movement through the soil–plant–atmosphere system at field scales (Deng & Hu 2003; Li et al. 2011). The HYDRUS-1D model (Šimůnek et al. 2008), for instance, is a widely used model for exploring water flow transport in variably saturated profiles. It has been successfully applied in many ecosystems, e.g., croplands, river riparian wetlands, and seasonally flooded wetlands, due to its adaptability and flexibility (Jimenez-Martinez et al. 2009; Xie et al. 2011).

Thus, in this study, the above-referenced HYDRUS-1D model was used to simulate water movement through the groundwater–soil–plant–atmosphere continuum and assess the effect of WTD on transpiration of two Poyang Lake wetland vegetation ecotypes: A. capillaris and P. australis communities. The main objectives of this study were: (1) to compare vegetation transpiration of two communities in two distinct hydrological years representing high and low water table conditions; and (2) to quantitatively explore the effect of WTD on the groundwater contribution to transpiration.

The water table in this site varies following the seasonal pattern of the water level in Poyang Lake (Xu et al. 2014). It generally rises above the ground surface in high water periods (July–August) and decreases to approximately 10 m below ground in low water periods (January–February) (Figure 2(b)). The water table in 2012 was significantly higher than in 2013. In the A. capillaris community, the highest water table level in 2012 almost reached the ground surface, but was at 2.2 m below ground in 2013. In the P. australis community, the land was flooded from May 17 to August 31 in 2012, during which time the groundwater table remained above the ground surface; however, in 2013 the highest water table never reached the ground surface. Therefore, the water level conditions in 2012 and 2013 represent high and low water table conditions, respectively.

The soil texture is mainly sand in the A. capillaris community and silt in the P. australis community, with mean soil bulk density of 1.33 g cm⁻³ at 0–100 cm depth. The A. capillaris community is mainly composed of mesophytic species of A. capillaris, Cynodon dactylon, Eriophorum angustifolium, and Hemarthria altissima. The P. australis community is composed of hygrophytes, including P. australis, Triarrhena lutarioriparia, and Artemisia selengensis. The upland wetland plants, e.g., the A. capillaris community, usually have one growing season. They germinate in March, thrive during April–August, blossom during September–October, and wither after the middle of December. For the middle–high wetland vegetation, e.g., the P. australis community, emergent plants usually grow fast out of the water when the land is flooded in July–August and have some re-germination after flooding again. Additionally, based on annual vegetation sampling results, the total biomass of the A. capillaris and P. australis communities both peaked in July–September (Xu et al. 2015). Thus, the periods during April–June and July–October are, respectively, defined as the initial and main growth stages for the upland wetland vegetation (Liu et al. 2000).

At the center of each community, volumetric soil moisture within the root zone was monitored at depths of 10, 50, and 100 cm by automatic probes (MP406 loggers, LSI-LASTEM, Italy). Next to the soil moisture sensors, one groundwater observation well was installed. Groundwater level was measured by a water pressure sensor (DQC001, LSI-LASTEM) installed inside the well. All the sensors were installed in December 2011 and the data were recorded daily from January 1, 2012 to July 31, 2014.

A micro-meteorology station was established to measure daily meteorological data at the study site, including rainfall, solar radiation, humidity, wind speed, and the maximum and minimum air temperature.

A Bowen ratio-energy balance (BREB) system (SP300, LSI-LASTEM) was constructed in December 2012 to estimate evapotranspiration in the A. capillaris community. The lower and upper probes for temperature and humidity (DMA672.1, LSI-LASTEM) were positioned 1.2 and 2.5 m above the ground surface, respectively. Net radiation was measured by a net radiometer (DPA240, LSI-LASTEM) positioned at 3 m above ground. Soil heat flux was measured by two soil heat flux plates (HFP01, LSI-LASTEM) buried 10 cm deep. Soil temperature was also measured using a thermocouple probe (TM10 K, LSI-LASTEM) at a depth of 10 cm. Measurements were made every 10 min from January 1, 2013 to July 31, 2014. Daily estimates of evapotranspiration were calculated at 1 h intervals and summed for the daylight period.

Distribution of root length density was investigated in August 2013 when roots were fully developed. During the sampling time, the site was exposed and the groundwater depth in both vegetation communities was greater than 2 m (Figure 2). Soil samples of 10 cm × 10 cm × 10 cm were excavated to a depth of 1 m until no obvious roots were discovered, and then carefully washed in 0.05 mm mesh screens to obtain all roots. The length of active roots (diameter <2 mm) at each soil layer was measured by a root scanner (WinRHIZO 2008, Canada).

Soil samples were taken from the ground surface down to 160 cm depth in each community using a soil auger. Well construction revealed that soils deeper than 160 cm were silt and thus were treated as homogenous. Soil mechanical composition and bulk density were analyzed in the laboratory and are summarized in Table 1.

In order to simulate water movement in the saturated–unsaturated zone, the lower boundary was selected below the water table level in the low stage period, at 10 and 8 m, respectively, in the A. capillaris and P. australis communities. The boundary was specified as a time-varying pressure head boundary and described by WTD measurements. The initial soil moisture profile was specified based on linear interpolation between soil moisture measured at January 1, 2012 and the saturated soil moisture below water level.

The calibration period was from January 1, 2012 to May 31, 2013, and the validation period was from June 1, 2013 to July 30, 2014. The simulated period included both wet and dry phases, allowing us to fit the model at a wide range of soil water conditions. Thus, the calibrated model can reproduce a soil–water environment that more accurately reflects actual field conditions. Moreover, when the land was flooded from May 17, 2012 to August 31, 2012, the soil profile was saturated and vegetation transpiration and evaporation were both assigned at the potential rate.

The initial soil hydraulic parameters, ⁠, ⁠, ⁠, n, and ⁠, were estimated by inputting the soil bulk density and percentage of sand, silt, and clay (Table 1) into the Rosetta pedotransfer function model (Schaap et al. 2001). The Marquardt–Levenberg inverse algorithm, incorporated in the HYDRUS-1D model, was subsequently used to optimize these parameters using field-measured data (Marquardt 1963). The calibrated soil hydraulic parameters are shown in Table 2.

The parameter values of and P in the S-shaped model ranged from −950 to −5,000 cm and 1.5 to 3, respectively, for different plants and soil types, as reported by previous studies (Skaggs et al. 2006a; Zhu et al. 2009). In the A. capillaris community, the soil texture in the root zone was mainly sand. Under this coarse soil, soil moisture usually decreased quickly. Consequently, relatively higher values of −950 cm and 3 were separately assigned to and P for the A. capillaris community to represent a steep curve in the soil moisture drainage process. In addition, the and P for the P. australis community were assigned values of −2,456 cm and 3, respectively, which were similar to the experimental results of Xie et al. (2011) for a P. australis community in a Yellow River delta wetland. In the calibration period, our results showed that simulated soil moisture varied slightly when these two parameters changed among the reported values. This agreed with the findings of Zhu et al. (2009) and Xie et al. (2011). Thus, the values of and P in this study were not further calibrated.

In the initial growth stage from April to June, the actual evapotranspiration of the A. capillaris and P. australis communities totaled 157–163 and 328–338 mm, respectively, and were equal to the potential level (Table 4). This period is also the wet season in Poyang Lake, and the rainfall infiltration amounted to 752–1,110 mm, accounting for approximately 55% of the annual amount. It is obvious that in this stage, rainfall can satisfy the evapotranspiration water demand and water was not a limiting factor for vegetation growth. With large amounts of rainfall infiltrating, the root zone bottom was mainly experiencing downward soil water movement (Figure 7). The cumulative amount of percolation in the A. capillaris and P. australis communities comprised about 74%–90% of the rainfall infiltration.

In the main growth stage from July to October, the actual evapotranspiration amounts of the A. capillaris and P. australis communities increased significantly to 257–430 and 599–759 mm, respectively. However, rainfall infiltration decreased sharply to less than 210 mm, which was much lower than evapotranspiration. During this period, upward water movement across the root zone bottom existed under the high water table condition (Figure 7). This indicates that rainfall in this stage cannot satisfy plant water demand, and vegetation transpiration must depend on the groundwater contribution.

According to variation in the depth of the water table, 2012 and 2013 represented high and low water table years, respectively (Figure 2(b)). Significant differences in vegetation transpiration in the main growth stage were detected between the two different hydrological years (Figure 7, Table 4). In the high water table year, the actual transpiration of A. capillaris and P. australis communities in the main growth stage totaled 334 and 677 mm, respectively, accounting for 96% and 92% of the potential transpiration totals. However, in the low water table year, the actual transpiration in this period added up to 203 and 510 mm, respectively, only accounting for approximately 55% of the potential transpiration totals due to there being insufficient water to support potential water uptake (according to Equation (7)). This revealed that in low water table conditions, vegetation transpiration in the main growth stage was constrained by the water deficit.

Two water sources contribute to vegetation transpiration when the wetlands are exposed: rainfall infiltration and upward fluxes of groundwater. By comparing the totals, we found that groundwater plays an important role in vegetation transpiration in the main growth stage.

In the main growth stage of the high water table year, the WTD in the A. capillaris community varied from 0 to 7.2 m, with a mean value of 2.6 m (Figure 2(b)), and a cumulative 439 mm of groundwater moved up into the root zone (Table 4). The groundwater contribution was able to satisfy the entire amount of water lost to transpiration (334 mm) in the A. capillaris community. Whereas, in the main growth stage of the low water table year, WTD varied within 2.2–9.6 m, with a mean value of 5.3 m, and the amount of groundwater supplied to the root zone was only 13 mm, which accounted for merely 6% of the transpiration totals (203 mm). This indicates that WTD directly determines the groundwater contribution to root zone soil moisture and affects the transpiration water consumption of the A. capillaris community.

In the P. australis community, a direct groundwater contribution to root zone soil moisture existed in the main growth stage when the land was exposed (Figure 7(b)). In the high water table year, the land was exposed from September, and the WTD ranged from 0 to 4.9 m with an average of 2.4 m during September–October (Figure 2(b)). The cumulative amount of groundwater moving upward into the root zone was 161 mm, which could account for a maximum of 74% of the water (219 mm) transpired by the P. australis community during September–October. However, in the low water table year, a mean water table of 3.0 m depth (varied within 0–7.1 m) in the main growth stage resulted in a groundwater contribution of 401 mm, which accounted for 79% of the transpiration totals (510 mm). These results indicated that most of the water transpired by the P. australis community in the main growth stage originated from groundwater when the wetland was exposed. In low water table years without inundation, groundwater recharge directly determined the available water for the P. australis community.

Based on the validated HYDRUS-1D model, simulation scenarios with constant WTD from July to October ranging from 0.5 to 3 m in 0.5 m increments were performed to further evaluate the influence of WTD on vegetation transpiration in the main growth stage. To make the results more representative of average meteorological conditions, scenario simulations were driven by daily averaged meteorological observations from 1955 to 2011 at the Duchang hydrological station (18 km from the study site). Soil texture and plant characteristics were the same as those used in the validated model.

The simulated scenarios of WTD = 0.5 and 1 m had the highest transpiration, with totals of 376 and 926 mm for the A. capillaris and P. australis communities, respectively. The water stress T_a/T_p ratio of the A. capillaris and P. australis communities was equal to 1 for both cases (Table 5). This suggested that when WTD was within 1 m, soil moisture was sufficient to sustain plants using water at the potential rate and transpiration was completely controlled by atmospheric demand.

Vegetation transpiration and the Ta/Tp ratio both decreased significantly with increasing WTD larger than 1 m, indicating that plant transpiration becomes limited by intensified water deficit. When the WTD was increased from 1 to 3 m, transpiration decreased to approximately 254 mm for the A. capillaris community and 364 mm for the P. australis community.

The simulated cumulative upward fluxes of groundwater to root zone soil moisture also decreased with increasing WTD (Table 5). With WTD within 1 m, approximately 209–213 and 711–715 mm of groundwater moved into the root zone of the A. capillaris and P. australis communities, respectively. Increasing the WTD from 1 to 2 m reduced the upward fluxes of groundwater to the A. capillaris and P. australis communities to 69 and 266 mm, respectively. When the WTD was further increased to 3 m, it was decreased by 92% and 86%, respectively.

The ratio of groundwater contributions (G) to vegetation transpiration (T_a) at different depths is shown in Table 5. It decreased with increasing WTD, which is in accordance with previous results (Luo & Sophocleous 2010; Xie et al. 2011). Under long-term annual average meteorological conditions, groundwater upward flow contributed as much as 56% and 77% of the vegetation transpiration of the A. capillaris and P. australis communities at WTD within 1 m, respectively. When the water table was lowered to 3 m depth, the groundwater contributions to the transpiration of the A. capillaris and P. australis communities decreased to approximately 7% and 28%, respectively.

Similar linear relationships were also reported by Grismer & Gates (1988) and Sepaskhah et al. (2003); Sepaskhah & Karimi-Goghari (2005). According to the similar slopes of Equations (15) and (16), we know that the contribution of groundwater to vegetation transpiration decreased by approximately 25% for every 1 m decline in the water table. In addition, the intercept for the P. australis community was higher than that for the A. capillaris community. This may indicate that groundwater plays a more important role in transpiration of the P. australis community than the A. capillaris community.

Water movement through the groundwater–soil–plant–atmosphere continuum is fundamental in the study of wetland eco-hydrological processes (Mazur et al. 2014). Numerical simulation has become an important method for water movement research. However, the complex wetland environment and limited monitoring methods made it still difficult to explore interface water transfer (Deng & Hu 2003; Xie et al. 2011). This study quantified the interfaces water flux dynamics in a typical river-connected lake wetland with large water level variations, and assessed the influence of WTD on vegetation transpiration. The results are helpful for a better understanding of the interactions of groundwater, soil water, and vegetation, and may provide reasonable references for wetland conservation and management.

In this study, continuous field observations of meteorology, hydrology, soil, and vegetation were conducted at an experimental site. The HYDRUS-1D model was well calibrated by soil water content and evapotranspiration with reasonable accuracy (Table 3). It should be noted that the root zone bottom water fluxes data are not available in the study, thus further calibration may be necessary by using the measurements of a lysimeter. As well, the assumption of 1D vertical soil water movement was applied to the study area. This assumption may be regarded as valid for the following reasons. First, the water table gradient is less than 0.002 in the site, implying a weak lateral water movement. Second, the soil texture is mainly composed of sand without a clay course, and thus no obvious lateral subsurface runoff is generated. However, portions of clay components usually increase significantly at low regions near the main lake area (Liu et al. 2000). As such, it would be more reasonable to extend the current model to a 2D vertical simulation to account for lateral water movement. Nonetheless, the present model is adequate as a first attempt to achieve the objectives of this study.

In river riparian and seasonally flooded wetlands, vegetation usually shows a zonal distribution pattern along the WTD gradient (Dwire et al. 2006). Many studies have indicated that plant composition and community construction are controlled by their ability to cope with the flooding that characterizes hypoxia environments (Ferreira 2000; Leyer 2005). Indeed, most wetland plants can reduce damage from low oxygen stress by developing aerenchyma and morphology adjustments (Coops et al. 1996; Qin et al. 2013). However, wetlands usually alternate between flooding and drought, especially under large variations in the depth of the water table, and consequently, plants typically suffer periodic drought stress (Sommer et al. 2003; Hammersmark et al. 2009; Xie et al. 2011; Li et al. 2013). Our results supported this point, finding serious water deficit for transpiration of the A. capillaris and P. australis communities during August–October in low water level years. As the water tables declines, seasonally flooded wetlands usually face a reduced water supply, which constrains plant water absorption. Some studies have suggested that flood-tolerant wetland plants usually have a low tolerance to drought, since a trade-off exists in species between tolerance to flood and tolerance to drought (Luo et al. 2008; Li et al. 2013). Clearly, drought is another severe stress influencing wetland plant growth, and it might impose more harm to plants than flooding.

In the Poyang Lake wetland, severe droughts have occurred frequently during the 21st century, particularly serious autumn dryness (Jiao 2009; Zhang et al. 2012b). Lake water levels from September to November in the 2000s were the lowest in nearly 60 years, with the minimum level even breaking the record low minimum value (Zhang et al. 2014). Consequently, the upland wetland habitats are exposed earlier and longer than normal under the decreasing water levels (Hu et al. 2010; Gan et al. 2011; Lai et al. 2012). Our results demonstrated that the low water table (e.g., in 2013) caused severe water stress for upland wetland vegetation transpiration during August–October. This might be one of the primary reasons for the upland wetland vegetation degradation and expansion towards the lake. Some studies have implied that the autumn dryness in the Poyang Lake wetland may be worsened in the future due to less catchment discharge in dry seasons and the construction of more dams on the upper reaches of the Yangtze River (Min & Min 2010; Ye et al. 2011; Zhang et al. 2014). This might cause the wetland water table to decline further and exacerbate the water shortages affecting vegetation transpiration in autumn. Consequently, interspecific competition for scare water may enhance among plants when the wetland is exposed. Over the long term, drought-tolerant plants might become dominant due to their stronger water competitiveness compared to flood-tolerant plants. Ultimately, the severe water stress is likely to accelerate the positive succession of Poyang Lake wetland vegetation.

In floodplain wetlands, understanding the effect of WTD on vegetation transpiration is important for revealing the interactions between soil water and vegetation. In this study, HYDRUS-1D models were used to explore the changes in groundwater contribution and vegetation transpiration for two vegetation communities under different WTD conditions in the Poyang Lake wetlands. The models were calibrated using field-measured soil moisture, and good agreements were achieved. The simulated results will be vitally helpful in predictions of the potential impacts of water table variation on wetland vegetation ecosystems under the background of climate change. The main conclusions are as follows:

1. In the initial growth stage (April–June), rainfall was higher than evapotranspiration, while in the main growth stage (July–October), rainfall could not satisfy plant water demand and transpiration was influenced by WTD. In the high water table year, vegetation transpiration comprised over 90% of the potential transpiration values; whereas in the low water table year, the transpiration totals accounted for only 55% of the potential transpiration totals, indicating severe water deficit. In the high water table year, upward fluxes of groundwater satisfied the transpiration water consumption of the A. capillaris community, while in the low water table year, it contributed merely 6% of the transpiration totals. In both years, groundwater contributed over 70% of the water transpired by the P. australis community in the main growth stage when the land was exposed.

2. Scenario simulations with constant WTDs in the main growth stage revealed significant decreases of vegetation transpiration with increasing WTD. When WTD was at 1 m, vegetation transpiration of the A. capillaris and P. australis communities could remain at potential levels, at which groundwater could contribute 56% and 77% of the transpiration totals, respectively. The ratio of groundwater contributions to vegetation transpiration decreased with WTD and could be described by linear equations for the A. capillaris and P. australis communities. Every 1 m decrease of water table resulted in approximately 25% reductions of the groundwater contribution ratio.

This study provides improved understanding of vertical water transportation in a groundwater–soil–plant–atmosphere system, exemplified in the Poyang Lake wetland. The results showed that WTD conditions significantly influenced vegetation transpiration. In addition, the findings of this study indicated that vegetation may also suffer periodic water stress in floodplain wetlands with large water level variation amplitudes. We recommend that further investigation of vegetation water use in humid regions should consider the influence of groundwater supply in dry seasons.

This study is funded by the National Natural Science Foundation of China (41371062), the National Basic Research Program of China (2012CB417003), and the Science Foundation of Nanjing Institute of Geography and Limnology, CAS (NIGLAS2012135001). We are also grateful to Mr Tan Zhiqiang, Miss Li Mengfan, and Miss Li Miao for assistance with field and laboratory work.