Spatiotemporal variation in bamboo Solar-induced chlorophyll fluorescence (SIF) in China based on the global Orbiting Carbon Observatory-2 (OCO-2) carbon satellite and study on the response to climate and terrain

ABSTRACT

Bamboo forests have a high carbon sequestration potential. Solar-induced chlorophyll fluorescence (SIF) provides fresh insights into the temporal and geographical variations in bamboo development and its interactions with the environment. SIF can serve as a probe for photosynthesis. In this study, we used distribution information on Chinese bamboo forests as well as a global SIF product, the global Orbiting Carbon Observatory-2 (OCO-2) SIF (GOSIF), which is derived from the OCO-2 carbon satellite. Thus, we used SIF as an indicator of bamboo forest growth. We investigated the spatial and temporal variations as well as the trends of bamboo forest SIF in China from 2008 to 2019 and explained the response mechanisms of bamboo forest growth to climate change and terrain differences in the region. We also evaluated the combined effects of climate and terrain on bamboo forest growth using the partial least squares path model (PLS-PM). The results were as follows: (1) Low values were found in the west and north, while high values were found in the east and south, according to the average bamboo forest SIF values in China from 2008 to 2019. More than 88% of the pixels had significantly increasing values. In conclusion, bamboo forests in China are growing well. (2) Solar radiation (Rad) and minimum temperature (Tmin) promoted the growth of most bamboo stands, while precipitation (Pre) and maximum temperature (Tmax) did the opposite. (3) The terrain, with a digital elevation model (DEM) less than 1000 m and a slope less than 9°, had a positive effect on bamboo forest SIF. (4) Climate and terrain jointly explained 45% of the bamboo forest growth change, and the major factor was climate change.

KEYWORDS:

• Bamboo forest
• OCO-2
• SIF
• spatiotemporal variation
• climate change

1. Introduction

With a total area of approximately 30 million hectares, bamboo forests are the second-largest forest type worldwide and are extensively scattered throughout the subtropical areas of Asia, Africa, and South America (Du et al. 2018). The center of the world’s bamboo distribution is in China, which is rich in bamboo resources, with more than 6.4 million hectares of bamboo forest, accounting for approximately 20% of the global bamboo forest area (Kang et al. 2022). Recent studies have shown that bamboo forests, especially moso bamboo (Phyllostachys pubescens), offer excellent potential as carbon sinks and have a high photosynthetic carbon sequestration capability (Kang et al. 2022; Li et al. 2015; Mao et al. 2016; Yen and Lee 2011). Bamboo forests are crucial in reducing the effects of global climate change (Li et al. 2014, 2015; Mao et al. 2020; Song et al. 2017).

Bamboo forests are vulnerable to climate change (Janzen 1976), and bamboo forest growth is related to seasonal changes in precipitation and temperature as well as their long-term trends under climate change (Li et al. 2019). Zhang et al. (2022) discovered that the distribution and growth of moso bamboo forests were determined mainly by annual precipitation. Anderegg et al. (2015) indicated that a higher minimum temperature retarded the growth of bamboo forests. The distribution of bamboo is susceptible to accelerating climate change (Li et al. 2019; Zhang et al. 2022). Therefore, accurately assessing the impact of climate change on bamboo forests is a key issue in the rational management and development of bamboo forests. Terrain can cause spatial differences in vegetation hydrothermal distribution and nutrients (Daly, Conklin, and Unsworth 2010) and is another important factor affecting vegetation growth differences (Adams, Barnard, and Loomis 2014; Sun et al. 2021; Zhang et al. 2021). For example, Cheng et al. (2023) found that an appropriate increase in slope was beneficial to the growth of landscape vegetation. Han et al. (2019) showed a significant dependence of forest growth and cover trends on elevation. Shi et al. (2020) found that location variables, including latitude and elevation, also significantly influenced the distribution of moso bamboo. However, there are few studies on how climate and terrain affect the growth of bamboo forests. It is unknown whether the growth patterns of bamboo forests are consistent with topographical variations and climatic change.

Solar-induced chlorophyll fluorescence (SIF) is a signal emitted by vegetation in the range of 650–850 nm under sunlight irradiation (Rossini et al. 2015). SIF occurs in the pigment layer of the photosystem, and it is an indicator of the efficiency of photon transport to the photochemical reaction center (Meroni et al. 2009). Therefore, SIF is strongly associated with the photosynthesis of plants (Luis et al. 2014; Zhang et al. 2014). Because SIF can reflect the potential of the actual photosynthetic process, studying the spatial and temporal variations in SIF is scientifically important for evaluating the physiological conditions of vegetation growth, disease stress, and photosynthetic carbon sequestration (Song, Wang, and Wang 2020; Walther et al. 2016).

Remote sensing is an important technical means of obtaining global vegetation SIF data, and the satellite sensors that are currently available for SIF inversion products mainly include the Greenhouse gases Observing SATellite (GOSAT), the Global Ozone Monitoring Experiment-2 (GOME-2), the SCanning Imaging Absorption spectroMeter for Atmospheric CHartographY (SCIAMACHY), the Orbiting Carbon Observatory-2/3 (OCO-2/3), the first Chinese carbon dioxide observation satellite (TanSat), and the TROPOspheric Monitoring Instrument (TROPOMI) (Zhang et al. 2019). Since 2011, many researchers have used these satellite datasets to retrieve global SIF remote sensing products (Du et al. 2018; Frankenberg et al. 2011; Guanter et al. 2012; Joiner et al. 2011; Köhler, Guanter, and Joiner 2015; Yao et al. 2021). However, the SIF products of GOSAT, GOME-2, and SCIAMACHY have a low spatial resolution. The SIF products of OCO-2/3, TanSat, and TROPOMI have improved spatial resolution, but OCO-2/3’s image pixel distribution is sparse, and TanSat and TROPOMI’s SIF products have a shorter temporal coverage (Zhang et al. 2019). Furthermore, they are not easily matched with flux tower observations. To solve the problems related to these SIF products, Li and Xiao (2019b) generated a SIF product at a spatial resolution of 0.05°, i.e. GOSIF, using machine learning techniques based on data from the OCO-2 SIF, meteorological reanalysis and remote sensing data from the Moderate Resolution Imaging Spectroradiometer (MODIS). Compared with the discrete original distribution of OCO-2 SIF, GOSIF has spatially continuous data and will have wider and more flexible applications in studies.

Vegetation indices (VIs) are simple and effective parameters that reflect dynamic changes of vegetation (Phiri, Shiferaw, and Tesfamichael 2020; Xue and Su 2017). Since SIF is more directly related to photosynthesis, it has been found that SIF is more accurate than VIs in assessing the sensitivity of vegetation growth environments to climate change (He et al. 2017; Li et al. 2020; Terry et al. 2003). Compared with VIs, SIF is less influenced by background clouds and soils (Li and Xiao 2019b; Lichtenthaler et al. 1986; Sun et al. 2017). Therefore, SIF emerges as a new go-to indicator for identifying plant growth (Zhang et al. 2023) and its interactions with climate change. For example, Green et al. (2017) analyzed the global biosphere-atmosphere feedback pattern using GOME-2 SIF data and found that biosphere-atmosphere feedbacks may explain up to 30% of the variations in precipitation and surface radiation in regions where feedback occurred. Song et al. (2018) found that wheat’s reaction to heat stress may be captured using SIF on a large scale and in real time. Related studies have also shown that SIF can quickly predict how heat and water stresses affect plant development, and thus, SIF is an important indicator for characterizing vegetation under drought and heat waves (Guan et al. 2016; Sun et al. 2016; Wang et al. 2019). Terrain can cause spatial differences in vegetation water vapor and heat distribution and nutrients (Daly, Conklin, and Unsworth 2010), and it is another important factor that affects differences in vegetation growth (Adams, Barnard, and Loomis 2014; Sun et al. 2021; Zhang et al. 2021). For example, Leroux et al. (2017) evaluated the influence of terrain on vegetation growth trends. Shi and Xiao (2022) found that terrain is an important factor influencing SIF; however, how terrain affects SIF is still unclear.

Bamboo forests are scattered, and monitoring growth status is difficult. Currently, the growth status of bamboo forests in China is unknown. In addition, previous studies have focused mainly on the influence of climate on vegetation growth (Sun, Chen, and Su 2021; Ding, Li, and Peng 2020). However, studies addressing the response patterns of bamboo forest growth to climate and terrain are not available. Therefore, we aimed to better comprehend the growth tendencies of bamboo forests and the interactive response mechanisms to climate factors and terrain differences. In this study, GOSIF products and remote sensing information of the bamboo forest distribution in China were used. We first used the Theil-Sen Median method to analyze the growth status and spatial and temporal variability of bamboo forests in China; then, we used partial correlation analysis and regression analysis to explain the response mechanisms of bamboo forest growth to climate change and terrain differences in China. Finally, we used a partial least squares path model (PLS-PM) to assess the interaction of drivers on bamboo forest growth. This study aimed to provide new ideas for the evaluation of ecological functions such as bamboo forest management and carbon sinks.

2. Materials and methods

2.1. Research area

China is a large country with a variety of climates (Figure 1). From south to north, the country has tropical, subtropical, warm, medium, and cold temperate climate zones. China’s bamboo forests are mainly in subtropical zones, including Zhejiang, Fujian, Jiangxi, Hunan, Sichuan, Anhui, Hubei, Guangdong, Guangxi, and other subtropical provinces and regions.

Figure 1. Study area and data: bamboo forest distribution and meteorological stations (a), annual precipitation (pre) (b), annual mean solar radiation (Rad) (c), annual mean maximum temperature (Tmax) (d), annual mean minimum temperature (Tmin) (e), slope orientation (aspect) (f), digital elevation model (DEM) (g), slope gradient (slope) (h) and latitude (Lat) (i).

2.2. Data acquisition and processing

2.2.1. GOSIF data of bamboo forest distribution areas in China

The distribution data of bamboo forests in China was obtained by Cui et al. (2019), using MODIS normalized difference vegetation index (NDVI), reflectance products, and other data, based on a decision tree combined with the mixed pixel decomposition method, the distribution and abundance data of bamboo forests in China with a spatial resolution of 1 km × 1 km were extracted. The accuracy of bamboo forest classification reached 81.16%. The estimated bamboo forest area had a high correlation with the forest inventory data (R² = 0.95), and the RMSE of bamboo forest abundance was generally less than 0.6.

Li and Xiao (2019a) used a machine learning method to generate spatially continuous global GOSIF data using the OCO-2 SIF and meteorological data. GOSIF has an 8-day temporal resolution, a spatial resolution of 0.05°, and a strong correlation (R² = 0.73, p < 0.001) with the gross primary production (GPP) of eddy covariance (EC) flow locations (Cao et al. 2023; Li and Xiao 2019a). Therefore, this study extracted the monthly and annual SIF data of bamboo forests in China from 2008 to 2019 using GOSIF data as the basis and combined them with information on the distribution of bamboo forests in China. Monthly data were used only to analyze the average monthly trend of bamboo forests.

To match the distribution data of the bamboo forests in China with the spatial resolution of the GOSIF data, it is usually necessary to process the former by downsampling. However, the distribution of bamboo forests in China is relatively discrete. Therefore, to ensure the downsampling accuracy, the method shown in Figure 2 was used in this study to screen the SIF pixels of bamboo forests according to the methods of related studies in China and abroad (Gao et al. 2020; Kimm et al. 2021; Li and Xiao 2022; Wang et al. 2020; Wood et al. 2017). First, the area of bamboo forests in each bamboo forest pixel was calculated as the product of the area of each bamboo forest pixel and its abundance. Second, the sum of the area of bamboo forests within each GOSIF pixel was calculated. Finally, only the GOSIF pixels whose total area of bamboo forests accounted for more than 60% of the GOSIF pixel area were counted as bamboo forest SIF pixels. For the screening criteria of bamboo forest pixels, we refer to the area share used by previous scholars in screening pixels (50% (Wang et al. 2020), 60% (Gao et al. 2020; Kimm et al. 2021)). To make the screened pixels more representative, we chose 60% as the screening criterion.

Figure 2. Schematic diagram of global OCO-2 SIF (GOSIF) data coupled with bamboo forest classification and abundance data to extract the bamboo forest solar-induced chlorophyll fluorescence (SIF). S represents the GOSIF unit pixel area; S₀ represents the bamboo forest unit pixel area; S₁ represents the area of the bamboo forest in the bamboo forest pixel, and S₂ represents the total area of the bamboo forest in the GOSIF pixel.

2.2.2. Terrain data

Terrain data include DEM, Lat, slope, and aspect. Using the Advanced Space-borne Thermal Emission and Reflection Radiometer Global DEM Version-2 (ASTER GDEM V2), the DEM data were processed to create a digital elevation product with a 30 m resolution. Using the slope and aspect analysis tools in ArcGIS 10.6, the slope and aspect datasets were derived from the 30 m DEM data. To match the GOSIF data, the above data were resampled to a spatial resolution of 0.05°.

To obtain a better analysis and more accurate statistics of the relationship between SIF and terrain, this study classified aspect into eight categories, namely, north (N, (0–22.5° and 337.5° − 360°)), northeast (NE, (22.5° − 67.5°)), east (E, (67.5° − 112.5°)), southeast (SE, (112.5° − 157.5°)), south (S, (157.5° − 202.5°)), southwest (SW, (202.5° − 247.5°)), west (W, (247.5° − 292.5°)), and northwest (NW, (292.5° − 337.5°)), where the southeast, south, and southwest slopes are sunny and the northeast, north and northwest slopes are shady. The slope ranged from 0° to 30° and was categorized by 0.5° intervals. The DEM range was 0–4100 m and was categorized according to 100-m altitude intervals. For the analysis of other data and terrain data, multi-year average values were used.

2.2.3. Meteorological data

The daily ground-based meteorological data from 824 nationwide meteorological observation stations, provided by the China Meteorological Science Data Center (http://data.cma.cn/), were used to calculate the climate data for the years 2008 to 2019. These data included daily maximum and minimum temperatures, daily precipitation, and daily solar radiation. To investigate the effects of water, light, and heat on the growth of bamboo forests, we calculated the Pre, Rad, Tmax, and Tmin based on the sum or average of the daily meteorological data and obtained meteorological raster data with a spatial resolution of 1 km × 1 km using the data from the 824 meteorological stations and an inverse distance weighted interpolation algorithm. The temperature was corrected by the DEM, and it was assumed that the temperature decreased by 6.5°C for each per-kilometer increase in altitude. Solar radiation was simulated according to sunshine duration measurements at each station according to Ju et al. (2006). Similarly, the spatial resolution of the climate data was resampled to 0.05° in this study to match the GOSIF data.

2.3. Research methodology

In this study, the Theil-Sen median (TSM) method and Mann – Kendall (MK) test were employed to determine the trend and significance of SIF in Chinese bamboo forests from 2008–2019. Pixel-by-pixel analysis of the partial correlation between bamboo SIF and climate parameters was conducted using the partial correlation analysis method, and the significance of the findings was assessed using a t test. In addition, we used path analysis (PA) and a partial least squares path model (PLS-PM) to elucidate the effects of climate change and terrain differences on SIF in bamboo forests.

2.3.1. Analysis of temporal and spatial trends of bamboo forest SIF in China

The TSM method is a reliable, very computationally efficient nonparametric statistical trend calculation method (Thiel 1950; Yin 2021). This method is frequently used for trend analysis of long-term data series since it is immune to measurement errors and outlier data (Narayanan et al. 2013). This algorithm is shown in Formula (1):

(1) $\beta =Median\left(\frac{x_j-x_i}{j-i}\right),2008\le i\le j\le 2019$(1)

where $Median\left(\hspace{0.17em}\right)$ denotes the median value. $i$ and $j$ represent the year. $x_i$ and $x_j$ are the SIF values of bamboo forests in the corresponding years with $j$ > $i$. When $\beta >0$, it indicates an increasing trend of bamboo forest SIF; when $\beta =0$, there is no change in bamboo forest SIF; and when $\beta <0$, it indicates a decreasing trend of bamboo forest SIF.

The MK test has the advantages that it is unaffected by missing values and outliers and does not require that the measured values have a normal distribution or that the trend be linear (Harka, Jilo, and Behulu 2021; Kendall 1975; Mann 1945). Therefore, this study adopted the MK test to test whether the interannual variation in the SIF of China’s bamboo forests was significant. The detailed process is as follows:

Assuming that $x_1$,$x_2$, … … , $x_n$ are time series variables, the statistic shown in Eqs. (2)-(3) are constructed as follows:

(2) $S=\sum _{i=1}^{n-1}\sum _{j=i+1}^{n}sgn\left(x_j-x_i\right)$(2)

(3) $\mathrm{s}\mathrm{g}\mathrm{n}(x_j-x_i)=\left\{\begin{array}{c}1,x_j-x_i>0\\ 0,x_j-x_i=0\\ -1,x_j-x_i<0\end{array}\right.$(3)

where $S$ is the MK test statistic of the SIF time series, and $n$ is the number of years. Then, we constructed the standardized test statistic as shown in Eqs. (4–5):

(4) $Var(S)=\frac{n\left(n-1\right)\left(2n+5\right)}{18}$(4)

(5) $Z=\left\{\begin{array}{c}\frac{S-1}{\sqrt{Var(S)}},S>0\\ 0,S=0\\ \frac{S+1}{\sqrt{Var(S)}},S<0\end{array}\right.$(5)

where $Z$ is the normal distribution statistic, and $Var(S)$ is the variance. The test statistic Z was used to test the trend. In this study, the change trend was defined as a significant change at the 0.05 confidence level. The change trend of SIF can be divided into significant change ($\left|Z\right|\ge 1.96$) and insignificant change ($\hspace{0.17em}\left|Z\right|<1.96$).

The interannual variation trend of SIF in bamboo forest pixels in China was classified into three classes according to the TSM and MK methods, where $\beta <0$and $\left|Z\right|\ge 1.96$ indicate a significant decrease; $\hspace{0.17em}\left|\mathrm{Z}\right|<1.96$ indicates a nonsignificant change; and $\beta >0$and $\beta >0$ indicate a significant increase.

2.3.2. Analysis of the spatial and temporal responses of bamboo forest SIF to climate change in China

By excluding the impact of additional factors, partial correlation analysis can more accurately portray the correlation between two variables than can simple linear correlation analysis (Wang et al. 2021). The partial correlation coefficient (PCC) measures the partial correlation between two variables. The expression of the partial correlation coefficient is calculated as follows:

(6) ${\rho }_{12\cdot 345}=\frac{{\rho }_{12\cdot 45}-{\rho }_{13\cdot 45}\times {\rho }_{23\cdot 45}}{\sqrt{(1-{\rho }_{13\cdot 45}^2)\times (1-{\rho }_{23\cdot 45}^2)}}$(6)

(7) ${\rho }_{12\cdot 45}=\frac{{\rho }_{12\cdot 5}-{\rho }_{14\cdot 5}\times {\rho }_{24\cdot 5}}{\sqrt{(1-{\rho }_{14\cdot 5}^2)\times (1-{\rho }_{24\cdot 5}^2)}}$(7)

(8) ${\rho }_{12\cdot 5}=\frac{{\rho }_{1\cdot 2}-{\rho }_{1\cdot 5}\times {\rho }_{2\cdot 5}}{\sqrt{(1-{\rho }_{1\cdot 5}^2)\times (1-{\rho }_{2\cdot 5}^2)}}$(8)

where ${\rho }_{12\cdot 345}$ is the bias correlation coefficient between variables 1 and 2 at fixed variables 3, 4, and 5. ${\rho }_{12\cdot 45}$is the bias correlation coefficient between variables 1 and 2 at fixed variables 4 and 5, and ${\rho }_{13\cdot 45}$ and ${\rho }_{23\cdot 45}$ are calculated in the same way as${\rho }_{12\cdot 45}$. ${\rho }_{12\cdot 5}$ is the PCC between variables 1 and 2 at fixed variable 5, and ${\rho }_{14\cdot 5}$ and ${\rho }_{24\cdot 5}$ are calculated as${\rho }_{12\cdot 5}$.${\rho }_{1\cdot 2}$, ${\rho }_{1\cdot 5}$ and ${\rho }_{2\cdot 5}$ represent the correlation coefficients between variables 1 and 2, variables 1 and 5, and variables 2 and 5, respectively. If the PCC is positive, it indicates that the variable is positively correlated with SIF. Otherwise, the variable is negatively correlated with SIF. The larger the PCC value is, the greater the effect of the variable on the SIF. A smaller value indicates a weaker effect (Liu et al. 2021). The significance test for the partial correlation is shown in Formula (9):

(9) $t=\frac{r_{12\cdot 345}}{\sqrt{1-r_{12\cdot 345}^2}}\sqrt{n-m-1}$(9)

where $n$ is the number of samples, and $m$ is the number of independent variables. In this study, m = 4.

In this study, the partial correlations between the SIF and climate factors in China’s bamboo forests were classified as significant positive correlations (PCC >0, P < 0.05), nonsignificant positive correlations (PCC >0, P > 0.05), significant negative correlations (PCC <0, P < 0.05) and nonsignificant negative correlations (PCC <0, P > 0.05).

2.3.3. Impacts of climate and terrain on SIF’s spatial and temporal variations in Chinese bamboo forests

PA is an SEM without latent variables that can quantify the importance of multiple independent variables to dependent variables (Li et al. 2023). It is a method used to decompose the correlation coefficient into a direct path coefficient and an indirect path coefficient based on correlation analysis and multiple regression analysis (Stage, Carter, and Nora 2004). Suppose there are n random variables x₁, x₂, x₃, …, x_n and a dependent variable y. The correlation coefficient between any two factors is r_ij. The direct path coefficient from factor x_i to y is P_iy. The indirect path coefficient from factor x_i to y through factor x_j is r_ij*P_jy. Then the correlation coefficient r_iy between factor x_i and y can be decomposed as the sum of the direct and indirect path coefficients. The path coefficient is used to describe the relative importance of the independent variable relative to the dependent variable (Li et al. 2023). A positive or negative path coefficient indicates that the independent variable has a positive or negative influence on the dependent variable.

PLS-PM comprises structural models and measurement models (Benitez et al. 2020). The measurement model describes the process of using observable variables (manifest variables) to construct variables that cannot be directly observed (latent variables), such as combining Pre, Rad, Tmax, and Tmin to form climate factors. Structural models measure the relationships between latent variables, such as climate factors and terrain factors. Therefore, PLS-PM can accurately reflect the causal relationship between each manifest variable and between the latent and manifest variables (Raykov and Marcoulides 2000; Wang et al. 2021). PLS-PM can also describe the role of individual variables in the whole process and the relationship between different individual variables, which can effectively solve the problem of multicollinearity (Benitez et al. 2020; Chin and Newsted 1999). PLS-PM was analyzed by the smartPLS software.

The commonly used indicators for evaluating the fitting effect of SEMs are Goodness of Fit Index (GFI), Root-Mean-Square Error of Approximation (RMSEA), Normed Fit Index (NFI), and Standardized Root-Mean-Square Residual Index (SRMR) (Lai 2021; Mitchell, Anderson, and Ferguson 2017). Shi, Maydeu-Olivares, and Rosseel (2020) found that SRMR could be used to obtain more accurate confidence intervals and test of close fit than that of RMSEA. Lombardi and Pastore (2012) found that compared with GFI and RMSEA, NFI was a more reliable indicator for evaluating SEM fitting. Therefore, we used the standardized root mean square residual (SRMR) and normative fit index (NFI) metrics to assess the model fit (SRMR <0.09 and NFI > 0.90 indicate a good model fit) (Dirzyte, Perminas, and Biliuniene 2021; Henseler et al. 2014).

3. Results and analysis

3.1. Spatial and temporal distribution and evolutionary patterns of bamboo forest SIF in China

As shown in Figure 3, the SIF of bamboo forests in China had obvious spatial heterogeneity, showing low values in the west and north and high values in the east and south. In addition, the mean SIF of bamboo forests in China was 0.218 Wm⁻²μm⁻¹sr⁻¹, and pixels with higher and lower than average values accounted for 53.28% and 46.72% of the total pixels, respectively.

Figure 3. Spatial distribution of the mean solar-induced chlorophyll fluorescence (SIF) values of bamboo forests in China from 2008 to 2019 (a), mean SIF values with longitude (b) and latitude (c).

According to Figure 4a, bamboo forests in China have experienced an increase in SIF over the past 10 years with an annual growth value of approximately 0.0035 Wm⁻²μm⁻¹sr⁻¹. The largest SIF value of the bamboo forests in China was 0.236 Wm⁻²μm⁻¹sr⁻¹ in 2017, which was 21.03% higher than the lowest value of 0.195 Wm⁻²μm⁻¹sr⁻¹ value in 2008. The SIF of the bamboo forests in 2019 was 0.231 Wm⁻²μm⁻¹sr⁻¹, which also increased by 18.46% compared to that in 2008.

Figure 4. Trends in the annual mean (a) and monthly mean (b) bamboo forest solar-induced chlorophyll fluorescence (SIF) in China from 2008 to 2019. The black line in Figure 4 (a) shows the trend line of the interannual change; the error bars in Figure 4 (b) represent the error for each month over the 11 years.

Figure 4b shows that the monthly variation in the SIF in China’s bamboo forests was a single-peak distribution. The SIF was the lowest in January, at 0.072 Wm⁻²μm⁻¹sr⁻¹, and highest in June, at 0.421 Wm⁻²μm⁻¹sr⁻¹. This result indicated that the seasonal variation characteristics of SIF were basically consistent with the growth pattern of bamboo forests (Li et al. 2021).

Figure 5 shows the spatial distribution of the SIF trends of the bamboo forests in China over the last 10 years obtained using the TSM and MK methods. Figure 5 shows that the overall trend of bamboo forest SIF in China was increasing in the last 10 years. Additionally, the area of significant growth comprised 88.08% of the total area, the area of insignificant change comprised 11.92%, and there were few areas that experienced reductions. The bamboo forests with significant increases were widely distributed. The bamboo forests with significant increases were distributed mainly at the junction of Jiangxi and Hunan, in southwest and southeast of Fujian, and in western Zhejiang, while areas with no significant changes were mostly in central Hunan, northwest Jiangxi, and central Zhejiang.

Figure 5. Spatial distribution of the solar-induced chlorophyll fluorescence (SIF) trends of the bamboo forests in China from 2008 to 2019.

3.2. Spatial and temporal responses of bamboo forest SIF to climate change in China

Figs. 6a and 7a show that the overall average PCC between SIF and Pre in the bamboo forests was −0.43. The significant positive correlation accounted for 0.97% and no significant negative correlation. The SIF of bamboo forests distributed in central China was mainly positively affected by Pre, while the SIF of bamboo forests in the eastern coastal areas was mainly negatively affected by Pre.

Figure 6. Distribution of the partial correlation coefficients (PCCs) between solar-induced chlorophyll fluorescence (SIF) and annual precipitation (pre) (a), annual mean solar radiation (Rad) (b), annual mean maximum temperature (Tmax) (c), and annual mean minimum temperature (Tmin) (d) for bamboo forests in China from 2008 to 2019.

Figure 7. Sign and significance of partial correlations of solar-induced chlorophyll fluorescence (SIF) with annual precipitation (pre) (a), annual mean solar radiation (Rad) (b), annual mean maximum temperature (Tmax) (c), and annual mean minimum temperature (Tmin) (d) in bamboo forests in China from 2008 to 2019.

According to Figures 6b and 7b, the average PCC between the SIF and the Rad in the bamboo forests was 0.10. The significant positive correlations accounted for 13.50%, while there were no significant negative correlations detected in the bamboo forests. Rad had a positive effect on the SIF of bamboo forests in central and eastern China. The pixels with significant positive effects were concentrated in the central region. Rad had a negative effect on the SIF of bamboo forests in southern China.

Figs. 6c and 7c show that the average PCC between the SIF and Tmax of the bamboo forests was −0.11. The significant positive correlation accounted for 6.69% of these areas and there was no significant negative correlation. The SIF of bamboo forests in central and eastern China was negatively affected by Tmax, while the coastal areas in eastern and southern China were positively affected by Tmax.

Figs. 6d and 7d show that the average PCC between the SIF and Tmin of the bamboo forests was 0.48. The significant positive correlation accounting for 48.05%. There were no significant negative correlations detected. The area with significant positive correlation was mainly distributed in the central and eastern regions. Only bamboo forests in southern China were negatively affected by Tmin.

3.3. Effect of terrain on the SIF of bamboo forests

As shown in Figure 8a, the highest SIF was found on southwest-facing slopes, with a value of 0.224 Wm⁻²μm⁻¹sr⁻¹, followed by that on the south and southeast slopes, while the lowest SIF was found on north-facing slopes, with a value of 0.216 Wm⁻²μm⁻¹sr⁻¹. Overall, the SIF of the bamboo forests on the sunny slopes was significantly higher than that on the shady slopes. As shown in Figure 8b, the SIF increased with increasing slope when the slope was less than 9° and decreased with increasing slope when it was greater than 9°. As shown in Figure 8c, the SIF increased with increasing DEM when the DEM was 0 ~ 1000 m; the SIF decreased with increasing DEM when the DEM was 1000 ~ 2500 m; and when the DEM is above 2500 m, the relationship between SIF and DEM was not obvious. Figure 8 (d) shows that the SIF of the bamboo forests showed a linear decreasing trend as Lat increased.

Figure 8. Response of the multi-year average solar-induced chlorophyll fluorescence (SIF) to slope orientation (aspect) (a), slope gradient (slope) (b), digital elevation model (DEM) (c), and latitude (Lat) (d) in bamboo forests in China.

Figure 9 was used to describe: (1) how SIF changes with year and terrain factors; (2) How the growth value of SIF changes with the variation of terrain factors. Figure 9a shows that the interannual variation in the SIF characteristics of bamboo forests with different aspects all showed an increasing trend. The bamboo forest with the highest SIF value grew in the southwest direction in 2018, the growth value of the SIF in the southwest direction was the highest, and that on the north slope was the lowest. The SIF growth value of the bamboo forests on sunny slopes was significantly higher than that on shady slopes. As shown in Figure 9b, the SIF of bamboo forests at different slopes showed an increasing trend, and the maximum bamboo forest SIF occurred in 2018 when the slope was approximately 9°. The interannual growth value of the bamboo forest SIF with slope showed a trend of increasing and then decreasing, and at a slope of 9°, the growth value peaked. As shown in Figure 9c, the SIF of bamboo forests with different DEM values also indicated an increasing trend, and the SIF of bamboo forests with a DEM value less than or equal to approximately 2000 m was greater than the SIF of bamboo forests with a DEM higher than 2000 m. The maximum SIF of the bamboo forests occurred in 2018 at a DEM of approximately 1000 m. The interannual growth value of the bamboo forest SIF with DEM showed increasing and then decreasing trends, and the growth value reached a maximum when the DEM was approximately 1400 m. Figure 9d shows that the SIF of bamboo forests with different Lat values also showed an increasing trend, the interannual growth value of the bamboo forest SIF with Lat showed a linear growth trend, and the bamboo forest SIF and the bamboo forest SIF growth values both reached a maximum at a Lat of 23.75°. Figure 9d shows that the SIF of bamboo forests also showed an increasing trend with different Lat values and a linear growth trend with the decrease in Lat. Both the bamboo forest SIF and the bamboo forest SIF growth values reached a maximum at a Lat of 23.75°.

Figure 9. Variation characteristic in solar-induced chlorophyll fluorescence (SIF) response to terrain (slope orientation (aspect) (a), slope gradient (slope) (b), digital elevation model (DEM) (c), and latitude (Lat) (d)) in bamboo forests from 2008 to 2019. Each circle in Figure 9 (a) represents both the year and the gradient of SIF growth values; the color mapping in Figure 9 (a-d) represents the mean SIF value for a given terrain condition in a given year (e.g. the “+” in Figure 9 (b) represents the mean SIF value in 2009 when the slope was 8°); the scattered points in Figure 9 (b-d) represent the SIF growth value for 11 years under a certain terrain condition (e.g. the solid point in Figure 9 (b) represents the SIF growth value for 11 years at slope was 14°); the fitted line in Figure 9 (b-d) is the trend of the growth value based on the scattered points.

Therefore, for bamboo forests growing on sunny slopes, with DEM values below 1000 m and slopes less than 9°, the aspect, slope, and DEM promoted SIF, while Lat always inhibited an increase in SIF. In addition, the SIF of bamboo forests grew fastest at a Lat of 23.75°, a DEM of approximately 1400 m, and a southwest-facing slope of approximately 9°.

3.4. Integrated response analysis of bamboo forest SIF to climate and terrain in China

Figure 10 shows that climate change and terrain differences could jointly explain 45% of the SIF change in the bamboo forests in China. Among these factors, climate change had a positive direct effect on the SIF change in the bamboo forests, with a value of 0.64. Terrain differences had a significant negative direct effect on bamboo forest SIF, with an effect value of only −0.04. However, its indirect effect on the SIF change (from changing climatic conditions) reached a value of −0.56 (−0.87 × 0.64), and the total effect (direct and indirect effects) on the SIF change of bamboo forests was −0.60 (−0.56 + (−0.04)).

Figure 10. The partial least squares path model (PLS-PM) of bamboo forest solar-induced chlorophyll fluorescence (SIF) with climatic and terrain factors in China from 2008–2019. “*” hereafter represents P < 0.05.

In summary, climate change directly affected the spatial and temporal distribution of the bamboo forest SIF, but the indirect effects of terrain differences from changing climatic conditions should not be ignored.

We used PA to obtain the interaction of climatic and terrain factors. Figure 11a shows the impact of climate and terrain factors on the SIF of bamboo forests in China and the impact of terrain factors on climate factors. The degree of influence is expressed as a path coefficient. The figure shows that the normalized path coefficients (absolute values) of the eight elements that influence bamboo SIF were in the following order: Tmax (0.28) > Tmin (0.27) > Pre (0.25) > Rad (0.20) > Lat (0.11) > DEM (0.05) > slope (0.04) > aspect (0.03). In the relationship between terrain and climate, Lat, DEM, and slope had the greatest direct effects on Pre (−0.81), Rad (−0.47), and Tmax (0.28), respectively.

Figure 11. The influence of climate and terrain factors on the solar-induced chlorophyll fluorescence (SIF) of bamboo forests in China and the influence of terrain factors on climate factors (a). The inner ring values in (a) are the direct effect of the climatic and terrain factors on the SIF of bamboo forests in China, and the outer ring values are the direct effect of the terrain factors on the climatic factors. The indirect effects of different terrain factors on SIF changes in the bamboo forests (b). Indirect effects are obtained by calculating the product of the influence factors in the path (for example, the indirect effect of slope gradient (slope) on SIF through annual precipitation (pre) is 0.055 (0.22*0.25). “*” hereafter represents P < 0.05.

Figure 11b shows the indirect effects of different terrain factors on SIF changes in bamboo forests due to climate change. That figure shows that Lat had the greatest indirect effect on the SIF variation in the bamboo forests through climatic factors, followed by DEM and slope, and aspect had the least effect. The changes in temperature and precipitation dominated by Lat and the solar radiation dominated by DEM were the key factors indirectly affecting the SIF changes in bamboo forests.

In addition, to more intuitively reflect the response of climate to terrain, we analyzed the climate under different terrain conditions, as shown in Figure 12. Figures 12 and 8 show that sunny slopes can receive more precipitation, solar radiation, and higher temperatures. When the slope was less than 9°, Pre increased slowly, and Rad, Tmax, and Tmin first increased and then slowly decreased; however, when the slope was greater than 9°, Pre, Rad, Tmax and Tmin all decreased. When the DEM was less than 1000 m, the Pre, Rad, Tmax, and Tmin all gradually increased with increasing DEM; when the DEM was higher than 1000 m, the Pre, Rad, Tmax, and Tmin all gradually decreased. Pre, Rad, Tmax, and Tmin decreased linearly with increasing Lat.

Figure 12. Response of terrain factors to climate change. There are four columns of group plots in Figure 12. From left to right, the responses of slope orientation (aspect), slope gradient (slope), digital elevation model (DEM), and latitude (Lat) to climate factors are shown. The points in the figures represent the mean value of the climate factor corresponding to the gradient of each terrain factor, and the color mapping represents the percentage of the number of pixels contained in that gradient.

In conclusion, terrain factors, including aspect, slope, DEM, and Lat, could influence the changes in the SIF of bamboo forests through the redistribution of resources such as water, light radiation, and heat.

4 Discussion.

This study is the first to apply SIF to long-term, large-scale monitoring of bamboo forest growth. To obtain SIF values that were as accurate as possible from the bamboo forest pixels, GOSIF pixels were screened in this study. When the cumulative area of the bamboo forest pixels within a GOSIF pixel was greater than 60% of the area of the GOSIF pixel, the GOSIF pixel was used as a bamboo forest SIF pixel. Although this method inevitably discarded some of the discrete bamboo forest pixels, the bamboo forest pixels selected using the method above were representative to some extent. The interannual temporal and spatial variation in the bamboo forest SIF from 2008 to 2019 showed an overall increasing trend. This result indicated that bamboo forests in China are growing well, which might be attributed to China’s current emphasis on environmental conservation. This has resulted in the vigorous protection and rational use of bamboo forest resources and the active promotion of the economic development of the bamboo industry. In addition, the monthly trend of the SIF of bamboo forests was single-peaked, with minimum and maximum values in January and June, respectively. This result indicated that the SIF well reflected the seasonal variation in bamboo forest growth and could be used to extract phenological information from bamboo forests.

In this study, partial correlation analysis and univariate polynomial fitting are used to analyze the response of SIF to climate factors and terrain factors, respectively. The correlation between climate factors is strong (Figure 13a), and it is not rigorous to analyze the relationship with SIF directly using simple correlation analysis and polynomial fitting. And the climate data used in this study are annual. Therefore, we used partial correlation analysis to obtain spatial correlations between climate factors and SIF. Due to the small change in terrain over a short period of time, we only used data from one period of terrain. And the correlation between terrain factors is weak (Figure 13b). Therefore, we extract the pixel values of the SIF and terrain factors to do a univariate polynomial fitting. In this way, the response of SIF to terrain differences can be found directly.

Figure 13. Correlation between climate factors (a) and terrain factors (b). “*” hereafter represents P < 0.05.

From 2008 to 2019, the SIF of bamboo forests in China had obvious spatial heterogeneity, showing low values in the west and north and high values in the east and south. The general trend of the bamboo forest SIF during the past 10 years indicated an upward tendency, with 88.08% of all bamboo forest pixels increasing and no pixels decreasing. This result may be due to climate change and terrain differences in different regions. To better study the impact of climate factors on the SIF of bamboo forests, this study used partial correlation analysis for four factors, Pre, Rad, Tmax, and Tmin, to further investigate the spatial response of the SIF of bamboo forests to climate factors. The average partial correlation between Pre and SIF was negative, and the negatively correlated bamboo forests were distributed mainly in the eastern, southern, and coastal regions of China. This may be due to the excessive summer precipitation in the eastern, southern, and coastal areas of China, which is mostly accompanied by extreme weather, such as persistent high temperatures and typhoons. Excessive rainfall can cause soil erosion and affect vegetation growth (Feng, Yang, and Huang 2020; Zhang and Jin 2021). Rad and Tmin were positively correlated with the average partial correlation of the SIF, while Tmax was negatively correlated. This may be because an increase in solar radiation is conducive to the photosynthesis of bamboo forests and can cause an increase in surface temperature and effectively promote tree growth. Vegetation has a certain amount of endurance in extreme temperature environments (Li et al. 2011). When bamboo species are subjected to high-temperature stress, their leaf pigment contents decrease continuously with increasing temperature, which seriously affects their growth and development (Li et al. 2015). When the environmental temperature is lower than the minimum temperature required for normal growth and development, bamboo species will reduce their respiration rate and metabolic intensity. In addition, Tmax had the greatest direct effect among the driving factors on the SIF of bamboo forests, followed by Tmin, Pre, and Rad. Most of the pixels in the partial correlation between climate factors and SIF were not significant. This may be due to the interference of other factors besides climate.

Vegetation growth is a complex process that is influenced by terrain differences and climate change (Lü et al. 2015; Ma et al. 2022). In this study, a great difference was found between the SIF of bamboo forests with different aspect, DEM, Lat, and slope values. The SIF of bamboo forests on sunny slopes was significantly higher than that on shady slopes. The sunny slope has sufficient sunshine, which provides better light conditions for photosynthesis. When the slope was less than 9°, the SIF of bamboo forest increased with increasing slope. When the slope was greater than 9°, the SIF decreased with increasing slope. This may be because the higher slope is good for water and nutrient retention. An excessive increase in slope may lead to an increase in soil erosion intensity, resulting in decreases in nutrients and water (Berger et al. 2010), which may limit the growth of bamboo forests. In the DEM range of 0–1000 m, SIF increased with increasing DEM. The main reason for this relationship is that there is an increase in carbon content and a decrease in human activities as altitude increases (Li, Shi, and Wu 2020), and thus, the bamboo forest can grow better. Low temperature is the most critical factor in limiting the growth of plants at high altitudes (Sun and Du 2017). For DEM above 1000 m, SIF decreased with increasing DEM. This may be because as the altitude increases, the precipitation and temperature gradually fall below the level required for the normal growth of bamboo forests (Huang et al. 2018; Jiang et al. 2021; Li et al. 2023). Therefore, bamboo forest growth is inhibited. With the increase in Lat, the SIF of bamboo forests in China showed a linear decreasing trend. This may be because with increasing latitude, sunshine and precipitation decrease (Park, Jeong, and Peñuelas 2020), and the growth of bamboo forests is inhibited.

The PLS-PM found that the change in bamboo forest SIF was closely related to climate change and terrain differences. Overall, the effects of climate change on bamboo forest SIF were greater than those of terrain differences, which was similar to the results reported by Li et al. (2022). Climate change is the main positive factor affecting the change in bamboo SIF. This may be because Chinese bamboo forests are distributed mainly in the subtropical regions of eastern and southern China. The region has a monsoon climate. The warm and humid climate is suitable for bamboo forest growth (Cai and Jin 2010). Terrain difference had a significant negative impact on the SIF of bamboo forests, but the impact was lower, which may be due to the flat terrain in the subtropical regions of eastern China, where the terrain has little effect on the growth of bamboo forests. However, as an important factor affecting the spatial and temporal heterogeneity of the environment, terrain affects changes in precipitation, radiation, temperature, and other factors to a certain extent (Tao et al. 2018). In addition, we found that the changes in temperature and precipitation dominated by Lat and the solar radiation dominated by DEM could indirectly inhibit the growth of bamboo forests. This may be due to the cooler temperatures and less precipitation at higher latitudes. Bamboo forests at high altitudes are exposed to excessive solar radiation. Excessive solar radiation can reduce the stomatal conductance and photosynthetic rate of leaves and inhibit the growth of bamboo forests (Sun, Chen, and Su 2021). In addition, as shown in Figures 12 and 8, sunny slopes, low latitudes, lower altitudes, and steeper environments were favorable to the growth of bamboo forests. This result is consistent with earlier research results (Cheng et al. 2023; Liu et al. 2018; Wang et al. 2021; Xiong et al. 2021).

There are obvious regional differences in natural geographic conditions in China, and there are several drivers that interact with one another. In this study, no consideration was given to how human activities, socioeconomic factors, and soil characteristics would affect the SIF in bamboo forests. The mechanisms of the drivers affecting the SIF of bamboo forests are complex and may even have time-lag effects. Therefore, we can combine more complete and detailed datasets and comprehensively explore the response mechanisms of bamboo forest growth to various driving factors at higher spatiotemporal resolutions. This will be the focus of our future research.

5. Conclusion

In this study, we extracted SIF information from bamboo forests in China using China’s bamboo forest classification data and GOSIF data. We used SIF as an indicator of the bamboo forest growth status to analyze the long-term spatiotemporal dynamics of bamboo forest growth in China and explored its response mechanism to climate change and terrain differences. In this study, the average bamboo forest SIF values in China from 2008–2019 were spatially characterized by low values in the west and north and high values in the east and south. We also found that the overall bamboo forest SIF in China increased each year, with an annual growth value of 0.0035 Wm⁻²μm⁻¹sr⁻¹. Approximately 88.08% of the total pixels had increasing values, and nearly no pixels had decreasing values. These findings indicate that the growth of bamboo forests in China is generally considered to be in good condition.

The growth of bamboo forests had a different spatial distribution pattern as a result of climate change. From 2008 to 2019, there were notable variations in the spatial distribution patterns of SIF with the Pre, Rad, Tmax, and Tmin partial correlations in bamboo forests in China. Among them, Rad had a significant promoting effect on 13.5% of the pixels, mainly in the central region. Tmin had a significant promoting effect on 48.05% of the pixels, mainly in the central and eastern regions. Pre and Tmax inhibited the growth of most of the bamboo forests. The pixels with significant inhibitory effect were 0.97% and 6.69%, respectively.

The growth of bamboo forests was significantly impacted differently by various terrains. Increases in aspect, slope, and DEM promoted the growth of bamboo forests on sunny slopes less than 9° and at DEM values less than 1000 m, while an increase in Lat always inhibited the growth of bamboo forests. In addition, bamboo forests growing in a southwest direction with a Lat of 23.75°, a slope of approximately 9°, and a DEM value of approximately 1400 m had the fastest rate of growth quality improvement.

Climate change and terrain differences could jointly explain 45% of the bamboo forest growth variation in China, and the impact of the former on bamboo growth was greater than that of the latter. However, terrain differences could have a strong indirect effect on bamboo forest growth by changing climatic conditions. The indirect effect of terrain on bamboo forest growth was greater than its direct effect, and the total effect of terrain on bamboo forest growth was similar to the impact of climate change. Compared to the terrain factors, each climatic factor had a stronger direct impact on bamboo forest growth. Tmax had the largest direct effect of all drivers on bamboo forest growth, and aspect had the smallest effect. The indirect effects of different terrain factors on the SIF changes in the bamboo forests through climate change differed. The DEM and Lat indirectly inhibited the growth of the bamboo forests mainly by changing Tmin and Pre, respectively, while the slope indirectly promoted the growth of the bamboo forests mainly by changing the Tmax.

Acknowledgments

The authors are grateful to the Editor and anonymous reviewers whose comments have contributed to improving the quality of this manuscript.

Disclosure statement

The authors disclosed no possible conflicts of interest(s).

Data availability statement

The datasets generated or analyzed during this study are available from the corresponding author on reasonable request.

Additional information

Funding

This work was supported in part by the Leading Goose Project of Science Technology Department of Zhejiang Province under Grant [2023C02035], in part by the National Natural Science Foundation under Grant [32171785, 32201553, and 31901310], in part by the Scientific Research Project of Baishanzu National Park under Grant [2022JBGS02], and in part by the Key Research and Development Program of Zhejiang Province under Grant [2021C02005].