The response of China’s wetland vegetation to climate changes

： Wetland vegetation dynamics are of vital importance for comprehending changes in ecosystem 8 structure. Under the background of global climate change, it is still unclear the change trends of wetland 9 vegetation in China, and whether there are differences between the response of wetland vegetation and 10 non-wetland vegetation to climate change. Based on Global Inventory Modeling and Mapping Studies 11 (GIMMS) NDVI3g, NOAA Vegetation Health Products (VHP) and climate data, this study explored the 12 response of wetland vegetation to climate change in China from 1981 to 2015. The results show that: 1) 13 NDVI of wetland vegetation in China shows a downward trend on the whole after the year of 2004. 2) 14 In water-limited zones, wetland vegetation NDVI is positive correlated with precipitation; while in 15 temperature-limited zones, it is positive correlated with temperature. 3) El Nino and La Nina may affect 16 wetland vegetation NDVI. The greater impact of La Nina phenomenon than El Nino phenomenon is the 17 possible reason for the upward trend of wetland vegetation NDVI, while the greater impact of El Nino 18 phenomenon than La Nina phenomenon may be the reason for the downward trend of NDVI. 4) The 19 response of wetland vegetation and non-wetland vegetation to climate change is significantly different. 20 Non-wetland vegetation responds more significantly to climate change than wetland vegetation.


25
Wetland vegetation is the totality of all plant communities in a wetland area (Cui and Yang, 2006).

26
It plays an important role in water storage, regulation of flood flows and maintenance of regional water 27 balance. Vegetation grown in wetlands can also purify environment by integrating environmental factors 28 (Powicki, 1998). In the past few decades, climate change has affected all continental ecosystems (Gao et

89
In addition, we use national wetland classification data to extract wetland vegetation samples. This

92
2012). They were synthesized from a spatial resolution of 30m to 1 km. We extracted coastal marshes 93 and inland marshes in the national wetland classification data for selecting samples. In details, the coastal 4 marshes include mangrove plants, reeds and so on, which are intertidal saltwater swamps with vegetation 95 coverage of 30% or more; inland swamps are permanent or seasonal swamps, including apes, herbs, 96 shrubs, forest swamps, oasis wetlands and spring wetlands with water surface exposed area less than 97 30%.

98
At last, we use the latest China climate zoning dataset (Zheng et al., 2010)

110
In this way, a total of 117 stable wetland vegetation (SWeV) samples were obtained nationwide ( Figure   111 2). Detailed information on the specific climatic zoning of SWeV samples can be found in Table S1.  September. We use the two-tailed T test to test the trend significance level of the regression equation. If 119 the obtained P is no more than 0.01, the trend is extremely significant.

120
The slope of the indicator trend line can by calculated by: where n is the number of years studied, i is the serial number of the year, and Aveseasonali is the average

134
Wavelet transform can decompose the time series into a high-frequency part and a low-frequency part.

135
High frequency can be used to extract the random item and the periodic item of time series, while the 136 remaining low-frequency part can be used to extract trend items of time series (Zhang et al., 2018a).

137
The wavelet transform is to shift (b) the wavelet basis function ψ(t) and then inner product with the Where Rxy is the correlation coefficient of the variables x and y, xi is the value of NDVI of i-year growing 158 season,īs the average of NDVI for the growing season of all study years, yi is the value of climate factors 159 of i-year growing season,īs the average of climate factors for the growing season of all study years.

160
We used partial correlation analysis to study the correlation between NDVI, near-surface 161 temperature and ground precipitation rate. When studying the correlation between NDVI and temperature, 162 the precipitation rate is set as the control variable. Moreover, we control temperature when studying the 163 correlation between NDVI and precipitation rate. The formula for calculating the partial correlation 164 coefficient is: Where Rxy.z is the partial correlation coefficient of x and y under the control factor z.

176
Through statistics (Table 1), we found that 82% of SWeV samples with NDVI ≥ 0.6 were located

209
Among the 117 research points, 40 of them whose VCI change trend is Type 2, which have a 210 significantly better wetland environment. The wetland environment of 18 samples whose VCI change 7 trend is Type 3 is significantly worse. There is no significant trend in VCI in 16 samples. Generally,

212
SWeV whose VCI change trend is Type 2 and Type1 accounts for 63% of the total.

223
We use the ground precipitation rate of China's regional ground meteorological elements dataset to   Table 2.

234
The response of wetland vegetation NDVI to climate variables is obviously different among the

251
Wetland vegetation with high NDVI is negative correlated with precipitation, while wetland 252 vegetation with low NDVI is positive correlated with precipitation (Table 1, Table 2). This is because the

269
The high frequency part of time series obtained by wavelet transform corresponds to the period term

303
El Nino and La Nina have opposite effects on climate. El Nino will disturb the consistent climate 304 characteristics of the region, that is, the precipitation in rainy season (or region) will decrease obviously, 305 and the temperature in low temperature season (or region) will increase abnormally. La Nina, on the other 306 hand, will significantly enhance the climate characteristics of the region, that is, the humid land will be 307 wetter and the arid land will be drier (Luo, 2000 (Table   317 3). The positive correlation (negative correlation) in Table 3 indicates that there is a positive correlation 318 (negative correlation) between indicators, but it has not passed the Two-tailed T test as a whole. The

344
Plateau has very dry climate. It is inland, so humid airflow that comes from ocean is difficult to reach.

345
The same situation occurs in S-AA and AA of PT. The increase of wetland vegetation NDVI in these 346 areas is likely to be affected by the increase of precipitation.

347
In dry climates, temperature rise will have a negative impact on wetland vegetation growth. In S-

356
In humid climates, the increase of precipitation will inhibit the growth of wetland vegetation.    Note: "PCorr_NDVI&T" is the partial correlation between NDVI and Temperature; "PCorr_NDVI&P" is the 598 partial correlation between NDVI and precipitation. TP is turning point; WA is Wet Area; S-WA is Semi-Wet Area;

599
S-AA is Semi-Arid Area; AA is Arid Area; CT is Cold temperature zone; MT is Medium temperature zone; WT is 600 Warm temperate zone; PS-f is Plateau sub-frigid zone; PT is Plateau temperate zone.

606
WA is Wet Area; S-WA is Semi-Wet Area; S-AA is Semi-Arid Area; AA is Arid Area; CT is Cold temperature 607 zone; MT is Medium temperature zone; WT is Warm temperate zone; NS is North subtropical; CS is Central 608 subtropical; PS-f is Plateau sub-frigid zone; PT is Plateau temperate zone.

609
"+" represents positive correlation; "+**" represents significantly positive correlation; "-" represents negative 610 correlation; "-**" represents significantly negative correlation; "x" represents no correlation between indicators.  The distribution of SWeV samples in wet-dry zones and temperature zones Figure 3 The average value and variance of NDVI of wetland vegetation samples in wet-dry zones (a) and temperature zones (b). (WA is Wet Area; S-WA is Semi-Wet Area; S-AA is Semi-Arid Area; AA is Arid Area.

612
CT is Cold Temperature Zone; MT is Medium Temperature Zone; WT is Warm Temperate Zone; NS is North Subtropical; CS is Central Subtropical; PS-f is Plateau Sub-frigid Zone; PT is Plateau Temperate Zone.)

Figure 4
The years in which the turning point (TP) is located (a) and spatial change trends of NDVI (b) of 117 SWeV samples Spatio-temporal change trends of near-surface temperature (a) and ground precipitation rate (b) Figure 7 One example of inter-annual variation trend of SWeV NDVI corresponds to El Nino and La Nina years