Spatiotemporal variation in vegetation phenology and its response to climate change in marshes of Sanjiang Plain, China

Abstract Sanjiang Plain is the largest marsh distribution area of China, and marshes in this region significantly affect regional carbon cycle and biodiversity protection. The vegetation phenology of marsh significantly affects the energy exchange and carbon cycle in that region. Under the influence of global climatic change, identifying the changes in phenology and their responses to climatic variation in marshes of Sanjiang Plain is essential for predicting the carbon stocks of marsh ecosystem in that region. Using climate and NDVI data, this paper analyzed the spatiotemporal variations in the start (SOS), end (EOS), and length (LOS) of vegetation growing season and explored the impacts of climatic variation on vegetation phenology in marshes of Sanjiang Plain. Results showed that the SOS advanced by 0.30 days/a, and EOS delayed by 0.23 days/a, causing LOS to increase significantly (p < .05) by 0.53 days/a over marshes of Sanjiang Plain. Spatially, the large SOS advance and EOS delay resulted in an obvious increasing trend for LOS in northern Sanjiang Plain. The rise of spring and winter temperatures advanced the SOS and increased the LOS, and the rise in temperature in autumn delayed the EOS in marshes of Sanjiang Plain. Our findings highlight the necessity of considering seasonal climatic conditions in simulating marsh vegetation phenology and indicate that the different influences of climatic variation on marsh vegetation phenology in different regions should be fully considered to assess the marsh ecosystem response to climatic change in Sanjiang Plain.

Vegetation is an important component of marsh ecosystem (Marani et al., 2013;Zeng et al., 2022). Marsh vegetation phenology significantly affects energy exchange, ecosystem functions, and carbon cycling (Vázquez-Lule & Vargas, 2021). Vegetation phenology of marsh is significantly susceptible to climatic variation, and it has altered significantly in recent decades due to climatic variation (Mo et al., 2015). Under the background of climatic variation, identifying the spatiotemporal variations of marsh phenology and understanding climatic effects on marsh phenology are essential for studying the regional carbon cycle (Luo, 2007;. Sanjiang Plain is the largest marsh distribution area of China, and these marshes affect the regional carbon cycle and biodiversity protection (Luo et al., 2022). Global climate change has significantly altered the marsh in this region (Wang et al., 2011), which may significantly impact the regional carbon sequestration and ecosystem functions . The effect of climatic variation on farmland vegetation phenology in Sanjiang Plain was analyzed, and it was observed that temperature promoted the advance of the vegetation growth period (Li et al., 2012). The effect of climate variation on forest phenology in Sanjiang Plain was studied, an increase in spring temperature advances the start of vegetation growing season (SOS), and an increase of autumn temperature delays the end of growing season (EOS) in Sanjiang Plain (Guo & Hu, 2022). Marsh ecosystems have unique climatic and environmental conditions compared with other ecosystems , which may lead to different effects of climatic variation on vegetation phenology.
The phenology in the marsh of Sanjiang Plain has been previously investigated. The SOS of Carex lasiocarpa was mainly concentrated in May (Sun & Song, 2008). The EOS of Carex lasiocarpa and Deyeuxia angustifolia were mainly concentrated in October (Hao et al., 2006). However, most researchers studied the plant phenology of few species in Sanjiang Plain, and no studies have investigated the vegetation phenological changes over marsh of Sanjiang Plain. Many researchers have indicated that climate change has different effects on the phenology of different vegetation types in one area or even on the same vegetation in diverse areas . Until recently, spatiotemporal variations of marsh vegetation phenology and climatic effects in Sanjiang Plain have not been clearly identified yet. Therefore, it is necessary to study spatiotemporal changes in phenology and climatic effects in marshes of Sanjiang Plain.

| Study area
Sanjiang Plain is situated in northeastern China with longitude and latitude of 129°11′E-135°05′ E, 43°49′N-48°27′N (Figure 1). The climate in study area is continental monsoon climate, with annual mean temperature from 1 to 5°C, and approximately 60% of the precipitation is concentrated in the months of July and August Wu et al., 2021). The common marsh species in Sanjiang Plain are Carex lasiacarpa, Calamagrostis anagustifolia, and Carex meyeriana .

| The climatic data
This research used monthly average temperature and precipitation data (2001-2020) from 21 climate stations in study region ( Figure 1). The meteorological data are downloaded from the China Meteorological Administration (http://www.nmic.cn/en), and underwent strict quality assurance .
To minimize errors and deviations, we performed initial quality control on these MODIS NDVI datasets based on pixel reliability parameters .

| The distribution of freshwater marsh data
Two marsh maps in 2000 and 2015 covering Sanjiang Plain (Mao et al., 2020) were used in this study. The spatial resolution of marsh distribution data was 30 m, and the accuracy has been verified in combination with field survey data (Mao et al., 2020).

| The extraction of phenology
For the purpose of avoiding the influences of land use/cover variation on our results, we extracted unchanged marshes (pertaining to marsh in both 2000 and 2015) during the study period and chose them as study area. Consistent with previous researches (Piao et al., 2006(Piao et al., , 2017Shen et al., 2018;Su et al., 2022), we used the polyfit-maximum approach to determine the phenology of marshes. First, we calculated the average of NDVI and analyzed the temporal variation in NDVI based on the formula (1): Here, t refers the Julian date (DOY), and NDVI (t) represents NDVI variation. When the NDVI rate reaches the largest increase/decrease date of the NDVI at this time is used as the threshold to determine average start and end date of growing season. We calculated the LOS using SOS and EOS of marsh vegetation in Sanjiang Plain.
As it is impacted by some nonvegetation effects of cloud, solar radiation angle, atmosphere and other factors, the NDVI detected by remote sensing usually has some outliers (Piao et al., 2006). We used the polynomial maximum method to better fit the NDVI time series. Its formula is as follows: where a 1 ,…a 6 refer the fitting coefficients of the least square regression.

| The calculation of climatic data
We applied the ordinary Kriging approach to interpolate the climatic data of weather stations into the marsh distribution of Sanjiang Plain, and then unified the spatial resolution of interpolated climate dataset into the same as NDVI dataset . Monthly climatic data were used to calculate mean values of temperature and precipitation in winter (previous December-February), autumn (September-November), summer (June-August), and spring (March-May).

| The trend analysis
For each variable, we calculated the regional average value based on the mean of all pixels distributed in this study region. This study used linear regression to calculate the change trends in variables (2001-2020) as following .
Here, M i refers to the variables value in the ith year; n means the period length and is 20 in this work; i refers the serial number of year; θ slope is the variation trend in variables of each pixel, if θ slope > 0, it means that the variables have an increasing trend, otherwise it is reduced.

| The relationship between phenology and climate factors
This study used a Pearson's correlation analysis to explore the impacts of climate change on marsh phenology .
Here, R ab is correlation coefficient; n refers to the period length and is 20, a i represents the mean value of the climatic factors in year i; a means the average climatic factors during the past 20 years; b i means the phenology in year i; b refers to the average phenology during the past 20 years.

| Distribution of marsh phenology in Sanjiang Plain
The (2) NDVI = a 0 + a 1 d + a 2 d 2 + … + a 6 d 6 in 285-305 DOY, and the region with later EOS was mainly found at eastern Sanjiang Plain ( Figure 2b). Accordingly, LOS was mainly concentrated in 155-190 days, and the region with a long LOS was mainly concentrated in the east of study area (Figure 2c).
Spatially, the region with the largest advance trend of SOS was distributed in the north Sanjiang Plain. The region with the delayed trend was mainly located in the middle and southeast of Sanjiang Plain ( Figure 2d). Furthermore, with regard to the changes in EOS, the region with the largest delayed trend of EOS was located in the north study area, and the region with the early trend was mainly located in the middle and southeast of study region (Figure 2e). The region with an increasing LOS trend was distributed in the north Sanjiang Plain. The region with a decreasing trend was located in the middle and southeast of study region (Figure 2f).

| Correlations between meteorological elements and marsh vegetation phenology in Sanjiang Plain
To explore the impacts of climatic variation on vegetation phenology in marshes of Sanjiang Plain, we analyzed the impacts of annual  (Figure 4). The region with the largest positive correlation between EOS and temperature in autumn was located east of Sanjiang Plain ( Figure 5); furthermore, the region with the largest positive correlations between LOS and temperature in spring and winter was located east of Sanjiang Plain ( Figure 6).

| The spatiotemporal variations of vegetation phenology in marshes of Sanjiang Plain
Spatial distribution was analyzed in terms of multiyear average vegetation phenology. The region with earlier SOS was mainly distributed in the east of study area (Figure 2a), which was similar to the result of

| Relationship of marsh phenology with climatic factors
The annual temperature showed significantly negative and positive relationship with the SOS and LOS, respectively ( It may be because that water is relatively abundant in marshes of Sanjiang Plain (Jin et al., 2016;Liu et al., 2022), and thus precipitation has no significant effects on marsh vegetation growth in Sanjiang Plain. By contrast, temperature plays an important role in many developmental biological processes of marsh vegetation (Badeck et al., 2004). The Sanjiang Plain is a temperate and relatively cold region, and the spring leaf onset of marsh plants generally require heat accumulation in this region (Geng et al., 2020).
The correlations between SOS and temperature from December to April were significantly (p < .05) negative, suggesting that increased temperature from December to April could advance the SOS in marshes of Sanjiang Plain . Our results suggest that increases in spring and winter temperatures may reduce frost and promote heat accumulation to initiate green-up (Luedeling et al., 2013;. The EOS had a significantly positive relationship with autumn temperature but had negligible correlation with precipitation, indicating that the rise in autumn temperature may delay the EOS in marshes of Sanjiang Plain. However, this result was different from a previous study (Liu et al., 2016), in which precipitation was the main determinant of grassland EOS and that increasing precipitation can alleviate water stress and delay EOS. The marsh ecosystem has sufficient water compared with the arid grassland ecosystem Shen, Liu, Zhang, et al., 2022  showed significantly positive correlations with temperature from January to April (Table 1), suggesting that increase in temperature from January to April can advance the SOS and could also prolong marsh LOS in study region. Our results showed that climate affects the LOS to a certain extent, specifically by affecting the SOS of marshes in study region.
In terms of multiyear changes in climatic factors, the average temperature from January to April showed increasing trends and the precipitation in June, August, September, and November showed significant increasing trends (  (Table 2), and temperature in July was positively correlated with the EOS (Table 1). Therefore, the increase in average temperature in July could account for the delayed EOS in study region. Spatially, the region with the largest increase in average temperature was located east of Sanjiang Plain in spring, summer, autumn, and winter (Figure 7). Spring and winter temperatures were significantly negatively correlated with marsh SOS (Figure 4), and positively correlated with marsh LOS in the east of study region ( Figure 6). Consequently, the increasing spring and winter temperatures may explain the advance of SOS and the increase in LOS in eastern Sanjiang Plain. Similarly, the highest positive correlation between EOS and temperature in autumn was located in eastern Sanjiang Plain (Figure 5c), indicating that the rising temperature in autumn may explain the delay of EOS in the east of Sanjiang Plain.

| Limitations
We should note that current work could have some limitations.

CO N FLI C T O F I NTE R E S T
The authors declare no conflict of interest.

DATA AVA I L A B I L I T Y S TAT E M E N T
The climatic data used in this study were downloaded from the China Meteorological Administration (http://www.nmic.cn/en), and the MODIS NDVI data were obtained from the Goddard Space Flight Center (https://ladsw eb.modaps.eosdis.nasa.gov/). The marsh distribution data were downloaded from National Earth System Science Data Center (http://www.geoda ta.cn/index.html).