Introduction

Coastal wetlands are very efficient in capturing and storing carbon. The area of coastal wetlands is less than 0.5% that of the ocean, but the carbon (C) stored therein accounts for about 50% of the total organic C buried in marine sediments (Duarte et al. 2005). The atmospheric carbon sequestered in vegetated, tidally influenced coastal ecosystems has been termed “Blue Carbon” (Nellemann et al. 2009). Blue carbon wetlands represent one of the densest C sinks in the biosphere (Duarte et al. 2013) because of (1) the high efficiency with which wetlands trap suspended matter and associated organic C during tidal inundation and sedimentation induced by riverine flooding, (2) the appreciable rates of belowground plant root growth in wetlands, and (3) lower concentrations of oxygen in wetland soils than in upland counterparts. Metabolism dominated by anaerobic decomposition reduces the rate of soil organic matter mineralization in wetlands. Furthermore, the sulfate ions (SO42−) in seawater that inundates coastal wetlands inhibit the production of methane (CH4) (Capooci et al. 2019; Poffenbarger et al. 2011). The carbon captured by coastal wetlands can therefore be stored in soil or sediment for thousands of years (Lo Iacono et al. 2008; Macreadie et al. 2013; McKee et al. 2007). Methane has a greater sustained-flux global warming potential (SGWP) (Neubauer and Megonigal 2015) than CO2. The high C sequestration rates and low rates of CH4 emissions that distinguish coastal wetlands as C sinks limit the effect that Greenhouse Gas (GHG) emissions have on Earth’s radiative balance (Chmura et al. 2003; Neubauer and Megonigal 2021).

However, the lifetime of the organic C stored in wetland soils depends on the form and location of the stored C (Angst et al. 2017; Ravn et al. 2020; Rumpel and Kögel-Knabner 2011), the environmental conditions in the soil (Chapman et al. 2019), and the presence of microenvironments suitable for decomposers and detritivores (Barthod et al. 2020). Furthermore, the decomposition of C and the uptake of C in wetlands are interrelated processes. Lu et al. (2020) has revealed that the decomposition of organic matter in coastal wetland stimulates plant growth and enhances the absorption of atmospheric C through the acceleration of weathering and the subsequent release of mineral nutrients. The interaction between plant growth and mineral weathering, along with the underlying mechanisms that connect organic matter decomposition and carbon uptake, are susceptible to anthropogenic alterations such as climate warming, invasive plant species, and tidal limitation (Chen et al. 2015; Mueller et al. 2016; Noyce and Megonigal 2021). These changes can confound accurate prediction of future alterations of radiation balance in coastal wetlands.

An increase in temperature can alter soil biogeochemical reactions. When the temperature of a body of water rises, for example, the solubility of CO2 in the water decreases, and some of the CO2 dissolved in the water escapes into the atmosphere (Li 2016). Furthermore, temperature indirectly affects the exchange of GHGs in coastal wetlands by controlling the rates of microbial metabolism and plant growth (Chapman et al. 2021; Noyce and Megonigal 2021). Warming can enhance gross primary production (GPP) by prolonging the growing season, increasing plant biomass, and enhancing plant photosynthetic performance (Yu et al. 2018). However, warming can also increase the rate of ecosystem respiration (Reco) and therefore stimulate emissions of CO2 and CH4 associated with the mineralization of organic C by enhancing the respiration rates of both autotrophs and heterotrophs. Excessive warming may adversely affect rates of carbon fixation by wetland plants if temperatures become supraoptimal.

In contrast to native plant species, the impacts of invasive plants are difficult to predict because they are function of ecological and geological conditions. The impacts may lead to an increase of carbon sequestration (Ge et al. 2015; Liao et al. 2007; Liu et al. 2022) or a transition from sink to source (Bradley et al. 2006; Xu et al. 2022). The invasion of new and prolifically growing plant species within saltmarshes has led to enhanced C pools in some ecosystems via, inter alia, increases of aboveground biomass (AGB) (Bottollier-Curtet et al. 2013). However, other studies have indicated that plant invasion leads to a release of a large amount of C-based greenhouse gases (Tong et al. 2012; Zhou et al. 2015), and still other studies have suggested that a transition from native to non-native species may have a limited net influence (Wang et al. 2019). At least two lines of reasons account for these conflicting results. First, the feedback effects of invasive plants differ significantly as a function of the eco-geological conditions that characterize specific soil environments (Davidson et al. 2018). Second, measurements made in many studies have emphasized only a certain component of the C cycle and have failed to include all relevant fluxes, i.e., NEE, Reco, and CH4 fluxes. Integrating the impacts of biotic factors such as invasive plant species with abiotic changes like warming is essential to advancing our understanding of future coastal wetland C cycling.

Tidal restriction refers to the limitation or obstruction of ocean-land interaction of coastal wetlands caused by human activities that alter the structure or landform of the area. Tidal restriction, such as dams and roads, have seriously affected the carbon sink of coastal wetlands (Kroeger et al. 2017). This has resulted in a reduction of carbon sequestration and an increase of GHG emissions (USEPA 2020). For example, reclamation-induced tidal restriction can increase the composition of dissolved carbon in salt marsh creeks and make them sources of greenhouse gases (Tan et al. 2021). Based on the modeled of climate forcing, Kroeger et al. (2017) have concluded that restoring tidal connectivity has a greater impact on reducing GHG emissions and increasing carbon sequestration than wetland creation or conservation. The growth of wetland plants, the emissions of CO2 and CH4, and the balance between the production and decomposition of organic matter largely depend on water table depth (WTD) and salinity. On the one hand, tidal restriction will cause wetland drainage, which exposes organic matter that has accumulated over many years to oxygen. The aerobic decomposition of that organic matter leads to the production of CO2 (Drexler et al. 2013). On the other hand, tidal restriction can also result in a reduction of salinity and lead to increased CH4 release due to the lack of SO42− as an electron acceptor.

This study aimed to quantify the effects of climate warming, different plant species, and tidal restriction on the carbon cycle in coastal wetlands of Yancheng, China. To achieve this goal, we used Open Top Chambers (OTCs) and selected locations where there were differences between plant habitats and tidal influences. We addressed two key questions: (1) Did climate warming, plant species and tidal restriction change the exchange of GHGs and the radiation balance in these coastal wetlands? (2) Which of these three factors had a greater impact on the carbon cycle of these coastal wetlands?

Materials and methods

Site details

The Yancheng Wetland (32°32′–34°25′ N, 119°55′–121°50′ E) is located in a muddy coastline along the Jianghuai Plain, Jiangsu Province, China (Fig. 1), and is the largest subtropical tidal saltmarsh in China. It has been included in the Ramsar List of International Important Wetlands. The average annual tidal range in the study area is 3–5 m (Ding et al. 2014). Because of the influence of the East Asia Monsoon and the plain landform, Yancheng has a warm, humid climate. The mean air temperature is 15.5 °C, and the mean annual precipitation is 1015 mm (Appendix S1: Fig. S1). The dominant saltmarsh vegetation types are Phragmites australis and invasive Spartina alterniflora. The Yancheng Wetland is a migration hub for over 400 species of migratory birds; more than 50 million migratory birds visit it annually (Ye et al. 2021). Sustainable management priorities include provision of wetland services, which include enhanced CO2 sequestration and provision of a habitat for birds (Ye et al. 2021).

Fig. 1
figure 1

Experimental design (upper panel) and locations in the Spartina alterniflora and Phragmites australis wetlands in Yancheng (lower panel), Jiangsu Province, China

Full size image

Two experimental warming sites were studied, both of which are part of the Coastal-wetland Research On Warming Networks (CROWNs) collaboration. This collaboration was established in early 2018 on the northeastern coast of China with partner sites in the United States and Europe.

Experimental design

Table S1 summarizes conditions at two of the experimental sites. The CROWN-S site was the Simaoyou Spartina wetland in the south of Yancheng and was divided into two locations, a warming-experimental location (CROWN-SW) and an alternate control location (CROWN-SA). The CROWN-SA was located 2 km away from the CROWN-SW and provided an opportunity to examine the response of GHG fluxes to differences between environmental conditions other than temperature. The CROWN-SA was more strongly affected by tides than the CROWN-SW (Fig. 1). The CROWN-P site was vegetated by native Phragmites. It was 60 km to the north of the CROWN-S and also included two locations, a warming-experimental location (CROWN-PW), and an alternate control location (CROWN-PA). The CROWN-PA was located 2 km southwest of the CROWN-P (Fig. 1). Unlike the CROWN-PW, the CROWN-PA had no apparent exchange with the sea because of a road that blocked tidal exchange.

The experimental studies spanned 29 months from May 2018 to September 2020. Open-top chambers (OTCs) were installed to produce continuous, passively warmed environments for Spartina and Phragmites (Fig. 1). The OTCs were designed as large, octagonal structures for wind resistance and were 2.7 m tall with eight 1.07-m-wide, clear, side panels (maximum within-chamber diagonal distance of 2.8 m). Each chamber occupied a ground area of 5.53 m2. The frames of the OTCs were made of aluminum alloy. The tempered-glass panels were 4 mm thick with light transmittance > 92% to ensure sufficient oblique solar radiation. The tops were open. A 10-cm vertical space was reserved between the bottom of the OTC and the soil surface to accommodate the dynamic astronomical and wind tides, as well as flooding caused by rainfall. Next to each warming plot, a non-warming control plot with the same octagonal base was installed, but without OTCs. OTCs were also not installed at the two alternate control locations, CROWN-SA and CROWN-PA. Each warming-experimental location included 6 warming plots (wCROWN-SW or wCROWN-PW) and 6 non-warming control plots (cCROWN-SW or cCROWN-PW), and each alternate control location included 6 alternate control plots without OTCs (cCROWN-SA or cCROWN-PA) (Appendix S1: Table S1). Online monitoring systems (OMS) were installed within one of the 6 warming plots and one of the 6 non-warming control plots at each of CROWN-SW and CROWN-PW (Appendix S1: Table S1). Boardwalks (width > 30 cm) were installed at all four of the locations to prevent investigator trampling during monitoring (Yu et al. 2022).

Gas flux measurements

Measurements of NEE, Reco and CH4 fluxes were made using an Ultra-Portable Greenhouse Gas Analyzer (UGGA) (Los Gatos Research, Quebec, Canada) coupled with a static chamber. The cylindrical static chambers were made of clear plexiglass (50 cm in diameter, 50 cm in height). Vent holes were provided on the side of the chamber to connect the UGGA to monitoring instruments. Each chamber was set on a perennially installed stainless-steel observation base with a trough around the base for chambers to create a water seal. Holes were drilled on the observation base at the soil surface to allow for flood water exchange during non-monitoring periods but were plugged with rubber stoppers during sampling.

The 50-cm chambers were stackable to ensure that plants were overtopped during measurements. Chamber junctions were sealed, and volume adjustments were incorporated based on number of chamber segments needed. A small fan was installed in the chamber at the top to ensure that the gas in the chamber remained fully mixed during UGGA measurements. The UGGA was connected to the chamber, and changes of CO2 and CH4 gas concentrations in the chamber were recorded over multiple periods of 3 min. The NEE and CH4 fluxes were measured directly with light chambers; the measurement of Reco was made with chambers that were darkened by covering them with a blackout cloth. Flux was calculated as follows:

$$F=\frac{d\mathrm{c}}{d\mathrm{t}}\frac{M}{{V}_{0}}\frac{P}{{P}_{0}}\frac{{T}_{0}}{T}V$$
(1)

where F is the gas flux (mg/(m2·h)); dc/dt is the rate of change of gas concentration in the chamber (ppm/sec); M is the molecular weight of the gas (CO2: 44 g/mol, CH4: 16 g/mol); P is barometric pressure (hPa) in the chamber; T is the air temperature in the chamber (°C); V is the chamber volume (cm3); T0, P0 and V0 are gas temperature, barometric pressure and molar volume under standard conditions, respectively.

The measurements were conducted during three growth seasons and are summarized in Table S2: 2018 (May 16 to November 19), 2019 (April 8 to November 2) and 2020 (May 29 to September 16). We were not able to measure gas fluxes at the end of the 2020 growing season due to COVID-19 restrictions. We monitored both sites in the shortest possible time, usually within one week. The measurements were carried out on each plot in turn between 8:30 AM and 4:30 PM.

Environmental parameters

Environmental data were collected and recorded using an online monitoring system (OMS). Parameters measured by the OMS included air temperature at 100 cm above the ground, soil temperature at 5 cm below the soil surface, and water table depth (WTD) relative to the soil surface. The time step was 10 min. The following sensors were used: for air temperature and relative humidity, a temperature-and-relative-humidity probe (model HMP155A, Vaisala, Finland); for soil temperature, conductivity, and soil moisture content, a soil moisture–conductivity-temperature sensor (model CS655-L, Campbell Scientific, USA); for WTD, a pressure transducer (model CS456, Campbell Scientific, USA). All data were converted to daily averages for subsequent analyses.

At alternate control locations without any OMS installations, we collected environmental data manually at the time of each field monitoring. Briefly, the air temperature was obtained from UGGA when measured gas flux, salinity was tested by pen salinity meter (SA287, HAZFULL, China), and WTD was measured in a piezometer by using water level data logger (MX2001-01, HOBO, the U.S.).

Field sampling and analyses

Soil samples were collected for moisture content, determination of bulk density, and elemental analysis from all six types of two sites (Fig. 1 and Appendix S1: Table S1). Soil surface samples were collected at 3 randomly selected combinations out of 6 plots for each type. We sampled in April, July, September, and November of 2019. Before sampling soils, plant litter on the soil surface was gently scraped with stainless steel tools. Surface soil samples were collected in situ using a weighted metal ring knife (60.35 cm3). The samples were transported to the laboratory and dried to constant weight at 105 °C. After re-weighing, moisture content and bulk density were calculated as follows (Ye et al. 2015):

$$\omega =\frac{{m}_{w}}{{m}_{s}+{m}_{w}}\times 100 {\%}$$
(2)
$${\rho }_{b}=\frac{m}{V(1+\omega )}$$
(3)

where \(\omega\) is the moisture content (%); \({m}_{w}\) and \({m}_{s}\) are the mass of water (g) in the soil sample and the mass of dry soil (g) in the soil sample, respectively. \({\rho }_{b}\) is the bulk density; \(m\) is the wet sample mass inside the ring knife (g); and \(V\) is the volume of ring knife (cm3).

Soil samples for elemental analysis were randomly selected from 5 sub-samples at each plot, and then mixed into one sample. The samples were wrapped and labeled in tin foil and stored in the refrigerator until laboratory analysis. Total C (TC), soil organic C (SOC) and total nitrogen (TN) measurements were made with an elemental analyzer (Vario, Max CN, Elementar, Germany) (Liu et al. 2017). Major soil elements (P, K Mg, Ca, Fe and Mn) were analyzed using X-ray fluorescence spectroscopy (ZSX Primus II, Rigaku, Japan) (Zhao et al. 2018). Soil trace elements (Zn, Cu) were analyzed using an ICP-MS (820-MS, Varian, the U.S.) (Ding et al. 2019).

The pH and salinity of pore water were measured in the field with an ultra-portable pH meter (6010 M, JENO, the U.S.) and pen salinity meter (SA287, HAZFULL, China), respectively. The concentration of SO42− was measured using ion chromatography (ICS-600, Thermo Scientific, the U.S.). Pore water salinity and pH measurement were synchronized with GHG monitoring over three years. Pore water required for testing SO42− was collected in April and October 2021.

The chemical index of alteration (CIA) has been used to reflect the extent to which feldspar has weathered to clay minerals (Nesbitt and Young 1982). We used the following formula to determine CIA values:

$$\mathrm{CIA=[n(}{\mathrm{Al}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\mathrm{)/(n}\left({\mathrm{Al}}_{\mathrm{2}}{\mathrm{O}}_{\mathrm{3}}\right)\mathrm{+n}\left({\mathrm{Na}}_{\mathrm{2}}\mathrm{O}\right)\mathrm{+n}\left({\mathrm{K}}_{\mathrm{2}}\mathrm{O}\right)\mathrm{+n}\left({\mathrm{CaO}}^\ast \right))] \times100$$
(4)

where CaO* is the CaO incorporated into silicate minerals. The CaO mole numbers were corrected to eliminate the bias caused by the presence of calcium carbonate: n(CaO’) = n(CaO) − 10 × n(P2O5)/3; If n(CaO′) < n (Na2O), then n(CaO*) = n(CaO′), otherwise n(CaO*) = n(Na2O), where “n” stands for molecular content (McLennan 1993). The CIA was used to classify weathering levels as follows: unweathered (CIA < 50), primary weathering (CIA = 50–64), moderate weathering (CIA = 65–85), and intense weathering (CIA > 85).

Above-ground plant biomass measurements

Above-ground biomass was estimated using a non-destructive method (Ye et al. 2016). The heights and basal diameters of all shoots within the observation bases inside the OTC and control frames were measured using a tape measure and a vernier caliper, respectively. To establish a regression equation for biomass determination within the observation bases, we harvested 30 shoots encompassing the range of heights, diameters and weight outside the observation bases inside the OTC and control frames. Six shoots were dried at 65 °C for 24 h to obtain plant moisture content, which was used to calculate the dry weight of 30 shoots. We used a deformed combination form of the logistic curve to describe the relationship between the dry weight of shoots, heights and basal diameter as follows:

$${W}_{\mathrm{f}}=\frac{{D}_{\mathrm{m}\mathrm{a}\mathrm{x}}}{{1+\mathrm{e}}^{-{k}_{1}(D-{d}_{\mathrm{m}})}}+\frac{{H}_{\mathrm{m}\mathrm{a}\mathrm{x}}}{{1+\mathrm{e}}^{-{k}_{2}(H-{h}_{\mathrm{m}})}}$$
(5)

where Wf is the dry weight (g) of each shoot, D is its basal diameter (mm), and H is its height (cm), The parameters Dmax, k1, dm, Hmax, k2, and hm were determined by least squares. We estimated the areal biomass (g/m2) inside the frames by substituting the base diameter and height of all the shoots within a frame into Eq. (5), summing the AGB of all shoots, and dividing this sum by the base area.

The period of monitoring the heights, diameters, and weights used to estimate AGB were synchronized with the period of gas monitoring. The heights and basal diameters of the 28 tallest plants were selected only in May and September 2020 to represent the plant traits at the beginning and end of the growing season.

Annual greenhouse gas flux estimation and radiative balance calculation

To estimate daily GHG exchange fluxes, we assumed NEE to be constant throughout the daytime (12 h) and the fluxes of Reco and CH4 fluxes to be constant throughout the day and night (24 h). The seasonal variation of GHG exchange fluxes was assumed to have a Gaussian distribution. The interannual climate conditions in 2018, 2019 and 2020 were similar (Appendix S1: Fig. S1). The daily GHG exchange fluxes of the years 2018 (5 measurements), 2019 (4 measurements), and 2020 (2 measurements) were combined as one year in the Gaussian model equation:

$$f\left(x\right)=a1{e}^{\left(-{\left(\left(x-b1\right)/c1\right)}^{2}\right)}$$
(6)

where \(f\left(x\right)\) is GHG exchange flux (mg/(m2·d)) on day x of the year, and a1, b1, and c1 are parameters. The parameters a1, b1, c1 were determined by least squares after eliminating the abnormal GHG exchange fluxes that were outside the 95% prediction bands of the fitted curve. According to Eq. (6), the flux values of GHG exchange were computed on a daily basis and subsequently aggregated to derive the annual GHG exchange flux.

The CO2-equivalent GHG fluxes of CO2 and CH4 were quantified in terms of their sustained-flux global warming potential (SGWP) if the fluxes were positive (soil/plant-to-atmosphere) or their sustained-flux global cooling potential (SGCP) if the fluxes were negative (atmosphere-to-soil/plant) (Neubauer and Megonigal 2015). Over 100-year time horizons, the SGWP and SGCP of CH4 is 45 times that of CO2 on a mass basis (Neubauer and Megonigal 2019). We used NEE and Reco to represent daytime and night CO2 exchange, respectively. However, Reco was measured during the daytime. We therefore assumed that the actual Reco at night was between zero and the Reco at the daytime (0.5\(\times\)total Reco 24-hour estimated value), because Reco at night is generally lower than that at daytime (Xu et al. 2017; Yang et al. 2018). The net CO2-equivalent GHG flux is represented by radiative balance (Neubauer 2021) as follows: where a is a parametric variable. The value of a is zero if Reco is assumed to be zero at night, and the value of a is 0.5 if Reco is constant throughout the day (24 h). When the radiation balance exceeds zero, the type is a C source; when the radiation balance is negative, the type is a C sink.

Statistical analysis

If the data were heteroscedastic, Kruskal-Wallis tests were used to test for differences among types. For AGB and GHG, the effects of experimental warming (wCROWN-SW vs. cCROWN-SW or wCROWN-PW vs. cCROWN-PW), plant species (cCROWN-SA vs. cCROWN-PW) and tidal restriction (cCROWN-SW vs. cCROWN-SA or cCROWN-PW vs. cCROWN-PA), were analyzed by one-way analysis of variance (ANOVA). Paired t-tests were used to analyze whether warming had a significant effect on GHG fluxes. Pearson correlation coefficients were used to analyze whether the correlation between greenhouse gas fluxes and environmental factors were significant.

Linear mixed effects models were used to assess the relations among NEE, and Reco and CH4 fluxes and the environmental factors. The fixed effects were plant species (categorical variable), air temperature, salinity, WTD and AGB (continuous variables). The random effects were site and plot. The effect of each variable or interaction was evaluated by removing the variable/interaction from the original model and using a likelihood ratio chi-square test for significant differences at the 5% significance level between the original model and the model without the variable/interaction. The original mixed effects model for GHG fluxes took the form:

$$GHG {fluxes}_{i}={\beta}_{0i} \cdot {AGB}_{i}+{\beta }_{1}\cdot Air temperature+{\beta }_{2}\cdot Salinity+{\beta }_{3}\cdot WTD+{b}_{1}\cdot \left(Site\right)+{b}_{2}\cdot \left(Plot\right)+{\varepsilon}_{i}$$
(7)

Where β, b and εi are the fixed variable coefficient, random variable coefficient and residuals for plant species, respectively. Statistical analyses were performed in SPSS (version 25, IBM Corporation, 1 New Orchard Rd, Armonk, NY 10,504, USS).

Results

Temperature manipulation

OTCs significantly increased average air temperature by 0.69 °C at CROWN-SW (p < 0.05) and 0.54 °C at CROWN-PW (p < 0.05) compared to control plots over a day-night cycle. The response of air temperature to warming was particularly apparent during the day, when the difference between warmed and control plots averaged 1.55 °C at CROWN-SW and 1.11 °C at the CROWN-PW. However, there was a slight cooling phenomenon at night, when OTCs decreased average air temperature by 0.18 °C at CROWN-SW and 0.04 °C at CROWN-PW (Appendix S1: Table S3, Figs. S2 and S3).

Environmental parameters and plant characteristics

Table 1 summarizes the environmental parameters and plant characteristics at 6 types. Except for the air temperature at 2 warming-experimental sites (Appendix S1: Table S3, Figs. S2 and S3) and salinity and SO42– concentration of CROWN-SW (Table 1), warming did not have a significant impact on environmental parameters. Compared with CROWN-PW, CROWN-SW had lower WTD (p < 0.05), lower moisture content (p < 0.05), higher salinity (p < 0.05), and a higher SO42– concentration (p < 0.05). The environmental conditions were different at CROWN-PA, where the soil moisture content was 65.64%. The salinity and SO42– concentration at the CROWN-PA, which was unaffected by the tides, were the lowest. Salinities were generally in the range 10–18 PSU at the other three locations, all of which were influenced by tides. Pore water pH was not significantly different among locations (p > 0.05). Locations other than CROWN-PW were classified as primary weathering sites based on their CIA values (Table 1). The CIA values as well as the concentration of Iron (Fe), Zinc (Zn), Copper (Cu) were lower at alternate control locations than the warming-experimental locations (Table 1).

Table 1 Physical and chemical environmental parameters and plant characteristics (mean ± CI) for experimental studies in Yancheng, Jiangsu Province, China
Full size table

Across growing seasons, warming altered plant characteristics (height and base diameter) of Phragmites (p < 0.05). However, warming increased the height of only Spartina significantly (p < 0.05), and warming caused no significant changes of base diameters across the growing season (Table 1). Warming significantly increased the AGB of Phragmites but not of Spartina (Table 2 and Table S4). Plant characteristics were of course different among locations. Plant heights were significantly higher at cCROWN-SW and cCROWN-PW compared to their corresponding alternate control location in the early growing season (p < 0.05). The AGB of different plant species differed significantly (Table 2). Except for September 2018, the AGB was much lower for Spartina than Phragmites (Fig. 2). Tidal restriction significantly affected the AGB of Spartina but did not affect Phragmites (Table 2). Except for the year 2018, the AGB was greater for cCROWN-SA than for cCROWN-SW throughout all seasons (Fig. 2). In contrast, during more than half of the measurement periods, the AGB was lower for cCROWN-PA than for cCROWN-PW (Fig. 2).

Table 2 Statistical analyses for the effects of plant species, experimental warming, and tidal restriction on aboveground biomass (AGB), net ecosystem CO2 exchange (NEE), ecosystem respiration (Reco) and CH4 fluxes for experimental studies in Yancheng, Jiangsu Province, China
Full size table
Fig. 2
figure 2

Seasonal variation of Phragmites australis and Spartina alterniflora aboveground biomass (AGB) by treatment and growing season for experimental studies in Yancheng, Jiangsu Province, China

Full size image

NEE, R eco and CH4 fluxes

Warming tended to inhibit the magnitude of NEE (equivalent to the absolute value of NEE). Among 8 of 11 measurement periods, the magnitude of NEE was lower at wCROWN-SW than at cCROWN-SW. Three of those differences were significant (Fig. 3a; * indicates a significant effect of warming). Warming at CROWN-PW also reduced the magnitude of NEE in 8 of 11 periods, two of which were significant (Fig. 3b; * indicated a significant effect of warming). In the early growing season, warming tended to increase Reco. However, warming tended to inhibit Reco in the late growing season (Fig. 3c and d; * indicated a significant effect of warming). There were no consistent effects of warming on CH4 fluxes, but one measurement in September at peak vegetation biomass showed that warming significantly increased CH4 fluxes in CROWN-PW (Table 2; Fig. 3e; * indicated a significant effect of warming). However, the responses of NEE, Reco, and CH4 flux to warming were not significant if all the measured fluxes were included in the analyses (Table 2).

Fig. 3
figure 3

Seasonal variation in NEE, Reco, and CH4 fluxes for experimental studies in Yancheng, Jiangsu Province, China (* indicates the response of GHG fluxes to warming is significant, p < 0.05)

Full size image

The NEE, Reco, and CH4 flux varied significantly between plant species (p < 0.05, Table 2). The Reco and CH4 flux of cCROWN-SA were significantly higher than those of cCROWN-PW (Table 2, Table S4 and Fig. 3). However, the magnitude of NEE was higher in Phragmites wetlands (Table 2, Table S4 and Fig. 3). GHG exchange showed clear seasonal patterns, except for CH4 fluxes in CROWN-PA (Fig. 3).

Among 10 of 11 measurement periods, the magnitude of NEE was lower at cCROWN-PA than at cCROWN-PW (Fig. 3b), especially in the middle of the 2019 growing season. Remarkably, the CH4 fluxes were greater at both alternate control locations (CROWN-SA and CROWN-PA) and those CH4 fluxes was significant higher (p < 0.05) than the corresponding fluxes at the warming-experimental locations (CROWN-SW and CROWN-PW) throughout all seasons and years (Fig. 3e and f).

NEE, R eco and CH4 fluxes versus environmental parameters and plant characteristics

Net ecosystem CO2 exchange (NEE) peaked at 25 °C in the Phragmites wetland, but NEE did not peak until 35 °C in the Spartina wetland (Figs. 4a and 5). The NEE was inversely related to WTD, salinity and AGB (p < 0.018, Table 3). The NEE in both vegetation types peaked at a salinity of ~ 15 PSU (Fig. 4b). There was a direct relationship between Reco and air temperature (p < 0.01, Table 3; Fig. 4d). The temperature effects on CH4 fluxes were a little less straightforward. Generally, high values of CH4 fluxes corresponded to temperatures higher than ~ 18 °C. However, in the freshwater CROWN-PA, CH4 fluxes did not follow this pattern, and high values were found at temperatures below 18 °C (Fig. 4g). Methane fluxes tended to be negligible at air temperature lower than 18 °C if measurements were made at salinities more than 5 PSU (Fig. 4h). Methane flux was negatively related to salinity (p < 0.018, Table 3), and were extremely high when the WTD was above the soil surface (Fig. 4i).

Table 3 Results of experimental studies in Yancheng, Jiangsu Province, China from linear mixed-effects models, with NEE, Reco, and CH4 fluxes as response variables
Full size table
Fig. 4
figure 4

Air temperature, salinity and water table depth versus NEE, Reco and CH4 fluxes for experimental studies in Yancheng, Jiangsu Province, China. (Positive water table depth values indicate that the water table was above the soil surface (flooded), while negative values indicate that the water table was below the soil surface)

Full size image
Fig. 5
figure 5

Relationships between NEE and temperature in CROWN-S (a) and between NEE and temperature in CROWN-P (b) for experimental studies in Yancheng, Jiangsu Province, China. Runs tests were based on sequences of values above or below the horizontal lines (the average values)

Full size image

Annual greenhouse gas flux and radiative balance

Annual GHG fluxes were estimated by integration of smooth curves fit to the daily fluxes. The methodology is illustrated in Fig. 6 in the case of the Reco data at cCROWN-SW and cCROWN-PW. The red crosse in Fig. 6b is the point that was ignored because it was outside the 95% prediction bands.

Fig. 6
figure 6

Daily Reco fluxes versus day of year for data collected at the cCROWN-SW and cCROWN-PW, respectively. The red cross indicates a datum that was ignored because it was outside the 95% prediction bands. The annual flux was estimated by integration of the area under the smooth curve

Full size image

Table 4 shows the annual greenhouse gas (GHG) flux estimates and radiative balance calculations from our experimental studies. Because the responses of the GHG exchange fluxes to warming were significant during some monitoring periods (Fig. 3), we estimated the annual GHG fluxes for the warming types (wCROWN-SW and wCROWN-PW). We believe that the warming effect would be very apparent if it were integrated throughout the year.

On the one hand, warming decreased the magnitude of annual NEE by 19% for the CROWN-SW and decreased the magnitude of annual NEE by 11% for the CROWN-PW. On the other hand, warming increased the annual CH4 fluxes by 28% for the CROWN-PW and generated no significant changes for the CROWN-SW. Moreover, warming had little influence on the annual Reco fluxes (Table 4). Warming increased the radiative balance (in kg CO2-eq/(m2·yr)) of the ecosystem from − 4.16 to − 0.74 to − 3.32 to 0.02 for CROWN-SW, and from − 4.97 to − 2.82 to − 4.13 to − 2.05 for CROWN-PW. The magnitude of the annual NEE of cCROWN-SA was significantly lower than that of cCROWN-PW. However, cCROWN-SA had a higher annual Reco and annual CH4 flux than cCROWN-PW. A significant finding was that the annual CH4 flux was particularly high (218.72 g/(m2·yr)) in the freshwater Phragmites wetland at cCROWN-PA, where it was ~ 58 times the lowest value measured at other types, a highly improbable result by random chance (Table 4). Moreover, the magnitude of annual NEE at CROWN-PA was also the lowest. The range of radiative balance at CROWN-PA was 7.11 to 9.64 kg CO2-eq/(m2·yr). The next highest annual CH4 flux was 33.52 g/(m2·yr) at CROWN-SA, which was ~ 9 times the lowest value measured at other locations. The range of radiative balance at CROWN-SA thus reached − 2.61 to 0.67 kg CO2-eq/(m2·yr).

Table 4 Annual greenhouse gas (GHG) flux and radiative balance range for experimental studies in Yancheng, Jiangsu Province, China
Full size table

Discussion

Responses of NEE, R eco and CH4 fluxes to warming

Our study area is in a subtropical zone where temperatures are high during the entire growing season. The highest daily average temperature reaching 30 °C. The additional warming that made the air temperature far exceed the optimum temperature for Spartina (30 °C) and Phragmites (25 °C) likely reduced the photosynthetic performance (Ge et al. 2014) and resulted in a reduction of the magnitude of NEE (Fig. 3a and b, * indicates a significant effect of warming). The cooling effect of plant transpiration was weakened because of the reduction in the air circulation within the OTC, and the resultant extremely high temperature of leaf surfaces further exacerbated the stress (Thongbai et al. 2010). Our previous study of leaf-scale photosynthetic performance (Jiang et al. 2022) also confirmed that warming does inhibit the photosynthetic performance of Phragmites. Spartina, as a C4 plant, has greater ability to acclimate to temperature. However, the main reason for the reduction of the magnitude of the NEE of Spartina by warming in this study (Fig. 3a) was likely the significantly higher salinity and lower WTD in the warming plots (Table 1) due to warming-enhanced evaporation. Overall, the response of NEE to this extent of warming was not significant (Table 2). Some studies have found that similar proportionate increases of Reco and GPP due to warming result in no significant change of NEEs (Chivers et al. 2009; Oberbauer et al. 2007). Other studies suggest that OTC failed to provide long-term effective warming to alter NEE, which may be related to the high humidity environment, where heat was lost in the form of latent heat (Johnson et al. 2013; Pearson et al. 2015). In addition, the response results of GHG exchange to warming is essentially a cumulative effect of the ecosystem over the entire experimental period (Oberbauer et al. 2007). In other words, the results now may depend on early responses from experiments a few years ago. And results in the middle and late growing seasons within a year may also depend on responses in the earlier growing season. This temporal asymmetry may also weaken the strength of the NEE-warming relationship.

The response of Reco to warming depended on the period of growing season (Fig. 3). Previous studies have shown that warming stimulates the activities of plants and microbes (Chivers et al. 2009; Fouche et al. 2014) and accelerated plant growth and microbial reproduction. Enhancement of biological respiration and soil organic matter decomposition will lead to an increase of Reco. This mechanism may account for the significant increase of Reco in the early growing season at CROWN-SW and CROWN-PW (Fig. 3c and d; * indicated a significant effect of warming). Unlike most previous warming studies that have focused on high latitudes, our study area was located in the subtropics, where the average annual temperature is higher and the plant growing season longer than at higher latitudes. A reversal in the response of Reco to warming was occurred in the late growing season, as there were a few cases of significant lower fluxes of Reco within warming plots, presumably due to warming stress (Fig. 3c and d; * indicated a significant effect of warming). Although Reco was highly correlated with variations of air temperature (p < 0.001, Table 3), the response of Reco to this magnitude of warming was not significant (p > 0.05, Table 2). From the perspective of temperature sensitivity (Q10) of Reco, even assuming that the Q10 is as large as 3 (Mahecha et al. 2010), the increases of Reco caused by warming were within the range of the noise of the data (Appendix S1: Section S1).

The response of CH4 fluxes to warming was insignificant (Table 2) with the exception of one measurement made on 11 September in the CROWN-PW. Previous studies have shown that CH4 fluxes do not respond significantly to OTC-induced warming, but rather to changes in WTD (Li et al. 2021; Munir and Strack 2014; Pearson et al. 2015). In other words, temperature is not the dominant factor affecting CH4 flux relative to the hydrological factors that control redox conditions. Moreover, the changes of the soil temperature of the OTC in our study were apparently not consistent and positive enough to stimulate changes in microbial mineralization and decomposition (Appendix S1: Figs. S2 and S3). A previous active warming experiment showed that the response of CH4 flux to warming was significantly increased by an increase of soil temperature greater than 5 °C (Hopple et al. 2020; Noyce and Megonigal 2021). Our study also showed that if the magnitude of the temperature change was large enough (from 25 to 30 °C), the response of the CH4 flux to the temperature increase was obvious (Fig. 4g, excluding data from the cCROWN-PA). In addition, there may be lag in the response of CH4 flux to warming, and top-down warming (OTC-induced) presumably prolong this lag period (Hopple et al. 2020).

Responses of NEE, R eco and CH4 fluxes to plant species

The significant difference in NEE between native Phragmites and invasive Spartina was mainly due to the difference in AGB (Table 2). The significant increase of photosynthetic leaf area with an increase of AGB and vegetation cover (Guo et al. 2021), explains the negative correlation between NEE and AGB (p = 0.018, Table 3). The advantage of Phragmites in AGB was particularly manifested in plant height (Table 1), which undoubtedly contributes to its capacity to absorb solar energy and CO2. The fact that the temperature at which the peak of NEE appeared was lower at CROWN-P than at CROWN-S indicated that the optimum temperature for photosynthesis was lower for Phragmites than for Spartina (Figs. 4a and 5). However, it cannot be concluded that Spartina sequesters more carbon than Phragmites as the climate warms. When the temperature exceeded 25 °C, the magnitude of NEE decreased at CROWN-P, but it was still not lower than that at CROWN-S. The effect of changes in WTD and salinity on NEE should be attributed to tidal restriction rather than plant species.

The fact that Reco was higher in invasive Spartina than in native Phragmites (p < 0.05, Table 2 and Table S4), was consistent with previous studies (Liao et al. 2007; Zhang et al. 2010b; Zhou et al. 2015). Spartina may affect Reco by altering the quantity and quality of input organic matter (Inglett et al. 2012; Xu et al. 2014). However, in our study, the AGB and SOC of CROWN-SA were significantly lower than those of CROWN-PW (Table 1; Fig. 2). And high Reco also occurred in CROWN-SW, which was less affected by tides (Fig. 3c). The results revealed that the reduction of AGB and changes in the redox conditions of the environment did not limit the occurrence of high Reco in the Spartina. There were hence other factors that may have stimulated the increase of Reco. We hypothesized that there were two reasons: first, the respiration rate of Spartina, a C4 plant, is higher than that of Phragmites, a C3 plant (Xu et al. 2014); Second, Spartina may transport more photosynthate underground than Phragmites to provide substrates for respiration (Chen et al. 2016).

In CROWN-SW, the CH4 fluxes were significantly lower in invasive Spartina than native Phragmites (Fig. 3). Compared with Phragmites, Spartina has a more well-developed root system, higher root density, and finer taproots. These characteristics facilitate the accumulation of oxygen in the rhizosphere for CH4 oxidation (Xu et al. 2014). In situ studies showed that Spartina oxidizes CH4 more rapidly, presumably because of its well-developed aerenchyma and gas transport capacity (Tong et al. 2012). However, we suggest that the low CH4 fluxes at CROWN-SW should be attributed to water deficit caused by tidal restriction rather than to differences of plant species. The CH4 fluxes of CROWN-PW were no greater or even lower than those of CROWN-SA, which was more strongly affected by tides and had a higher WTD (Table 1; Fig. 3). Yuan et al. (2015) also found that CH4 release from Spartina was higher than that from Phragmites because of its location in a more saturated soil environment. In other words, if sea level continues to rise, the CH4 fluxes of CROWN-SW will presumably increase substantially and even surpass those of Phragmites. This pattern has been observed in other experimental mesocosm studies (Cheng et al. 2007; Zhang et al. 2010a). The CH4 fluxes from CROWN-SA may therefore be indicative of the CH4 fluxes from CROWN-SW in the future.

Responses of NEE, R eco and CH4 fluxes to tidal restriction

Human activities, such as road building, damming, and crab trapping (via crab trapping trench and fences), could restrict the tidal exchange between the warming-experimental locations and the alternate control locations. Tidal restriction commonly causes differences of salinity and WTD that affect GHG exchanges in salt marsh wetlands (Portnoy 1999).

Several studies have found that increased salt stress can inhibit the magnitude of NEE in coastal wetland ecosystems, especially during the early growing season (Abdul-Aziz et al. 2018; Volik et al. 2020; Wei et al. 2020). Increasing salinity adversely affects lipid metabolism and the synthesis of chlorophyll and protein in leaves. The resultant reduction of the photosynthetic rates of plants will adversely affect plant growth and the magnitude of NEE (Parida and Das 2005; Pierfelice et al. 2017). However, in our study, elevated salinity enhanced the magnitude of NEE at CROWN-PW compared to CROWN-PA (Fig. 4b). The explanation may be a moderate increase of salinity has little effect on the chlorophyll content and photosynthetic parameters of Phragmites but improves its water use efficiency (Li et al. 2018; Pagter et al. 2009). For example, Gorai et al. (2010) found that Phragmites grows best in a low-oxygen, moderate-salt environment. We hypothesize that Phragmites may produce additional low-molecular-weight compounds via more rapid photosynthesis to acclimate the ion balance within the plant and assimilate more C in this process (Parida and Das 2005). WTD is the basic condition for controlling NEE (p < 0.001, Table 3). Due to tidal restriction, the WTD was significantly lower for CROWN-SW than for CROWN-SA. That difference in WTD may explain the positive NEE at CROWN-SW (Fig. 3). Water deficit can be a threat to wetland plants, and the role of wetlands as C sinks can be constrained. For example, water willow in the southeastern United States experienced limited growth, productivity, and survival after more than two weeks of drought (Touchette and Steudler 2009). Chu et al. (2019) have also highlighted the fact that a high soil moisture content promotes plant biomass and the magnitude of NEE early in the growing season. Unlike CROWN-PW (the WTD was below the soil surface at low tide), which is affected by tides, tidal restriction caused a persistent retention of freshwater at the CROWN-PA. The CIA of CROWN-PA was lower because of long-term anaerobic conditions at that location (Table 1). The reduced production via weathering of Fe, Zn, Cu, essential trace elements for plant growth (Kaur et al. 2022; Lehmann et al. 2014), may have reduced the magnitude of NEE.

In our study, the weak responsiveness to variations of salinity exhibited by Reco (Fig. 4e) was contrary to the findings of previous studies (Huang et al. 2020; Wei et al. 2020). Generally, salinity stress reduces microbial cell activity and adversely affects plant growth (Chambers et al. 2016; Li et al. 2018). However, in coastal wetland ecosystems, different plants acclimate to different salinity gradients, and the same plants can also acclimate to different salinity gradients because of their high plasticity and/or genetic factors (Brix 1999; Clevering and Lissner 1999). The fact that salinity gradient have not caused variations of microbial respiration (Yan and Marschner 2013) may explain why Reco responded weakly to salinity variations. In the case of flooding, both autotrophic and heterotrophic respiration would be inhibited (Chivers et al. 2009), and the diffusion of CO2 could be hindered and some of the CO2 captured by water (Han et al. 2015). The fact that the vascular structure of wetland plants helps alleviate the impact of flood stress on Reco (Brix et al. 2001) explains why the Reco exhibited no consistent pattern in response to WTD (Fig. 4f). Overall, Reco was not affected by tidal restriction because of the ability of plants and microbes in coastal wetlands to acclimate to different salinity gradients and flooding conditions.

Water table depth (WTD) is an important determinant of soil redox conditions, which are an important determinant of CH4 fluxes (Zhao et al. 2020). Tidal restriction exposed the soil to oxygen and resulted in a lower flux of CH4 at CROWN-SW (Fig. 4i). When the WTD rose to the soil surface, it was no longer a limiting condition for CH4 release, and the fluxes depended on salinity, temperature, or other environmental factors. Our result also showed a significant relationship between CH4 flux and salinity (p = 0.018, Table 3). Consistent with most previous studies, increased salinity suppressed CH4 production, mainly because sulfate reduction took precedence over methanogenesis (Capooci et al. 2019; Olsson et al. 2015; Poffenbarger et al. 2011). Tidal restriction lowers the salinity in landward wetlands (Emery and Fulweiler 2017) and therefore led to significant CH4 production at CROWN-PA (Fig. 4h). Under unique environmental conditions, the flux of CH4 can be high even at low temperatures. Strong evidence of this was the extremely high CH4 fluxes at temperatures below 18 °C at CROWN-PA (Fig. 4g), where there were high concentrations of organic C and nutrients in the soil, an elevated WTD, and low salinity (Table 1). There were hence sufficient available substrates for CH4 generation, good reduction conditions, and weak competition from sulfate reduction. It should be noted that the changes in salinity and SO42– concentration do not necessarily occur synchronously. Compared with CROWN-SW, CROWN-SA has higher salinity and lower SO42– concentration (Table 1). We infer that the SO42− concentration decreased due to partial reduction to H2S under better reducing conditions. This may be another reason for the higher CH4 flux in CROWN-SA.

Carbon source and sink judgment

Warming weakened the magnitude of the annual NEE and resulted in a more positive radiation balance at the two warming types. The wCROWN-SW possibly became a weak carbon source (–3.32 to 0.02 kg CO2-eq m− 2 yr− 1). Compared with the Phragmites in CROWN-PW, Spartina in CROWN-SA decreased the annual NEE and increased the annual Reco and CH4 flux. Spartina overall reduces the carbon sink capacity of coastal salt marshes, and this effect is further exacerbated by tidal restoration or future sea level rise. Although CROWN-SA and CROWN-SW were carbon sinks, both were occasionally weak carbon sources (i.e. positive radiation balances). Tidal restriction severely affected the radiative balance of the Phragmites wetland from a consistent carbon sink (–4.97 to − 2.82 kg CO2-eq m− 2 yr− 1) to a consistent carbon source (7.11 to 9.64 kg CO2-eq m− 2 yr− 1) by decreasing the magnitude of the annual NEE and increasing the annual CH4 flux.

In summary, tidal restriction may have a greater impact on the carbon sinks of coastal wetlands than climate warming and invasive plants. Tidal restriction in the form of dams and roads directly affected wetland WTD and salinity. Carbon uptake by plants and anaerobic decomposition of organic matter were thus affected. The result has been a decrease of carbon storage potential and extreme changes in CH4 emissions in the coastal wetlands studied.

Climate warming and plant species may also affect the carbon cycle of coastal wetlands, but the impact may be limited. The wCROWN-SW and cCROWN-SA were weak carbon sources because of warming and changes in plant species. However, the positive radiative balance occurred near the boundary condition (assuming that Reco was the same during the night and daytime). This result and the high positive correlation between Reco and temperature (p < 0.01) suggest that these C sources may be overestimates. In other words, if the future climate warming is controlled below 1.5 °C and the scale of Spartina invasion remains the same, climate warming and plant species may have less impact on the carbon sink function of coastal wetlands. Further studied could help improve understanding of the impact of tidal restriction on the carbon sink of coastal wetlands. We acknowledge that the annual flux of GHG exchange may be overestimated by the chamber method. But our objective in estimating annual flux was to examine the mechanisms by which warming, plant species and tidal restriction affect GHG exchange in coastal wetlands, rather than to obtain precise values of annual flux for carbon trading purposes.

Concluding remarks

If the future temperature increase can be controlled within 1.5 °C, the magnitude of the radiative balance of native Phragmites might be reduced, but the direction is unlikely to change, and Phragmites will remain a consistent carbon sink. However, there is a risk of invasive Spartina becoming a carbon source under the influence of climate warming or sea level rise due to lower in AGB and the magnitude of NEE and higher in Reco and CH4 flux compared with native Phragmites. Notably, freshwater Phragmites wetlands formed by tidal restriction reduced the magnitude of NEE and substantially increased CH4 emissions. the result was a consistent carbon source.