Impact of precipitation levels on vegetation in ecologically fragile karst areas in the Guangxi (China) karst region

Overview of the study area

The Guangxi Zhuang Autonomous Region is located in the southern part of China, between 20°54′N and 26°23′N and 104°28′E and 112°04′E. This region is bordered by the Yunnan-Guizhou Plateau in the west, Guangdong Province in the east, Beibu Gulf in the south, and Hunan and Guizhou Provinces in the north. Its strategic location is significantly important. Guangxi has a complex and diverse terrain, mainly consisting of mountains and hills. Karst landforms are widely developed, accounting for 37.8% of the total land of the region. The climate is subtropical monsoon, characterized by warm and humid conditions, with abundant rainfall and distinct seasons. Annual average temperature ranges from 16 °C to 23 °C, and the annual precipitation varies from 1,200 to 2,000 millimeters. The vegetation types in Guangxi are rich and varied. Shrublands are the most widely distributed, accounting for 63.40% of the total vegetation area of the region, followed by broadleaf forests, accounting for 17.20%. Bamboo forests are the least, only accounting for 0.88%. In addition, Guangxi has a relatively high forest coverage rate, which reached 62.5% in 2023, ranking the top in the country. Distribution of the forests is shown in Fig. 1.

Data

Daily temperature and precipitation date of 69 meteorological stations from 1961-2020 provided by the Guangxi Meteorological Information Center are used to calculate rainstorm disaster indicators (Table 1).

Satellite remote sensing data came from the Moderate Resolution Imaging Spectro-radiometer (MODIS) product MOD13Q1 (MODIS/Terra Vegetation Indices 16-Day L3 Global 250 m SIN Grid), provided by the National Aeronautics and Space Administration (NASA). This dataset features a spatial resolution of 250 m and a temporal resolution of 16 d, providing high-quality synthesized time-series data. The data used corresponds to version V006 and covers the period from 2000 to 2021. The MOD13Q1 remote sensing dataset of Guangxi Karst region was preprocessed through a series of steps, including band extraction, mosaicking, projection transformation, region extraction, and data format conversion. This process generated a high-quality NDVI dataset, which is synthesized on a monthly basis to create a monthly NDVI dataset. Annual average NDVI values are calculated from this dataset.

Geographic information data includes digital elevation model (DEM) data, administrative boundaries of cities and counties, latitude and longitude coordinates of meteorological stations, vector boundaries of karst areas, and d distribution data of forest species in Guangxi karst areas.

Methods

Using terrestrial ecosystem carbon budget (TEC) models, the net primary productivity (NPP) can be calculated (Xu, Hu & Wang, 2012) as follows: (1)${\displaystyle {\mathrm{NPP}}_{ij}={\mathrm{GPP}}_{ij}-R_{ij}}$ (2)${\displaystyle {\mathrm{GPP}}_{ij}={\varepsilon }_{ij}\times \text{FPAR}\times {\mathrm{PAR}}_{ij}}$ (3)${\displaystyle {\mathrm{NPP}}_i=\sum _{j=1}^n{\mathrm{NPP}}_{ij}}$

where NPP_ij, GPP_ij and R_ij (gC m⁻²) are respectively the net primary productivity, total primary productivity and respiratory consumption of vegetation in the j^th month of the i^th year; ɛ_ij(gC MJ⁻¹) is the actual utilization rate of light energy in the j^th month of the i^th year, reflecting the influence of temperature, water and other factors on photosynthesis; FPAR is the proportion of photosynthetic active radiation absorbed by vegetation; PAR_ij (MJ m⁻²) is the incident photosynthetic active radiation in the j^th month of the i=^th year; NPP_i (gC m⁻²) is the net primary productivity of vegetation in the i^th year, and n is the total number of months in a year, n = 12.

Based on the image element linear decomposition model, vegetation cover is estimated using the NDVI and the pixel dichotomy method. That is, the NDVI value of each pixel can be expressed as a combination of contributions from two components: vegetation cover and non-vegetation cover. This relationship is quantified through the transformation of the vegetation cover fraction. FVC can be calculated as follows (Liu, Liu & Liu, 2010): (4)${\displaystyle {\mathrm{FVC}}_{ij}=\left({\text{NDVI}}_{ij}-{\text{NDVI}}_{\mathrm{s}}\right)/\left({\text{NDVI}}_{\mathrm{v}}-{\text{NDVI}}_{\mathrm{s}}\right)}$ (5)${\displaystyle {\mathrm{FVC}}_i=\frac{1}{n}\sum _{j=1}^n{\mathrm{FVC}}_{ij}}$

where, FVC_ij(%) is the vegetation cover in the j^th month of the i^th year; NDVI_ij is the NDVI in the j^th month of the i^th year; NDVI_s and NDVI_v are the NDVI of full soil cover and meta-vegetation full coverage, respectively, NDVI_s = 0.05 and NDVI_v = 0.95 based on the characteristics of vegetation in China; FVC_i is the vegetation cover in the i^th year. When NDVI < 0.05, the vegetation cover is negative, and there is no vegetation in the area.

The Pearson’s correlation analysis is used to examine the relationship between remotely sensed indicators (such as FVC and NPP) for different vegetation types under different rainfall conditions (Xu, Hu & Wang, 2012). The calculation formula is as follows: (6)${\displaystyle \mathrm{R}=\frac{\sum _{i=1}^n\left(x_i-\overline{x}\right)\left(y_i-\overline{y}\right)}{\sqrt{\sum _{i=1}^n{\left(x_i-\overline{x}\right)}^2\sum _{i=1}^n{\left(y_i-\overline{y}\right)}^2}}}$

where, R is the correlation coefficient of variables x and y; x_i is the vegetation remote sensing parameters in the i^th year; $\overline{x}$ is the mean of multi-year vegetation remote sensing parameters; y_i is the rainfall in the i^th year, and $\overline{y}$ is the mean of multi-year rainfall. The value range of correlation coefficient R is [−1, 1]. The larger R is, the stronger the correlation between the variables is. The significance test is conducted by t statistic.

Responses of vegetation remote sensing parameters to different rainfall levels

Fractional vegetation cover (FVC)

The average absolute values of the correlation coefficients between FVC and the different rainfall levels (extremely heavy rainstorm, heavy rainstorm, rainstorm, heavy rain and moderate rain) were 0.24, 0.19, 0.17, 0.17 and 0.19, respectively. These results indicate a weak correlation. The negative correlation areas accounted for 63.0%, 61.3%, 39.8%, 41.1% and 44.50%, respectively. There was a significant negative correlation in 23.3%, 1.58%, 0.61%, 1.16% and 1.00% of the areas (p < 0.05). FVC was mostly negatively correlated with extremely heavy rainstorms and heavy rainstorm, positively and negatively correlated with rainstorm, and positively correlated with heavy rain and moderate rain. The spatial differences in the correlation between FVC and different rainfall levels are significant (Fig. 2). The spatial distribution characteristics of FVC negatively correlated with rainfall indicates that: the high-value areas of the negative correlation between FVC and extremely heavy rainstorm were mainly located in the southwest (a). The negative correlation areas of heavy rainstorm were concentrated in the central region (b). The negative correlation areas of rainstorm and heavy rain were similar, mainly located in the southwest, north, and some parts of the northeast. The distribution was scattered, and the range of heavy rain was slightly larger than that of rainstorm (c and d). Moderate rain was mainly distributed in some parts of the north, with a relatively dispersed pattern (e). Comparing the correlation of FVC with different rainfall levels, FVC has the strongest correlation with extremely heavy rain, and the negative impact of extremely heavy rain on the vegetation is the most significant. The negative impact areas of extremely heavy rain and heavy rain are extensive, which further emphasizes the potential threat of extreme rainfall events to the vegetation ecosystem. From torrential rain to heavy rain, and then to extremely heavy rain, the correlation between FVC and them has changed significantly. Even the positive and negative correlation has changed in some areas. This change in correlation may be related to the factors such as increased rainfall intensity, leading to changes in soil moisture saturation, surface runoff, and physiological response of vegetation. As the rainfall level increases, soil moisture may reach saturation and surface runoff increase, putting greater pressure on vegetation and causing changes in correlation. In addition, the physiological adaptation mechanism and stress resistance of vegetation may also show different response characteristics under different rainfall intensities, thereby affecting the correlation between FVC and rainfall.

This study reveals the complex correlation between the vegetation cover fraction and different rainfall levels, as well as their spatial distribution characteristics in this area, providing an important scientific basis for a comprehensive understanding of the impact of rainfall on the vegetation ecosystem. The results cannot only help us understand the interaction mechanism between vegetation and rainfall, but also provide a useful reference for formulating targeted vegetation protection and restoration measures, as well as responding to the impact of climate change on ecosystems.

Net primary productivity (NPP)

In this area, the absolute values of correlation coefficients between vegetation NPP and different rainfall levels (extremely heavy rainstorm, heavy rainstorm, rainstorm, heavy rain and moderate rain) were 0.23, 0.19, 0.22, 0.21 and 0.27, respectively. The values indicate a weak correlation. The proportion of areas with a negative correlation was 40.7%, 21.1%, 7.45%, 7.28% and 1.69%, respectively. Negative correlation was significant (p < 0.05) in 0.44%, 0.29%, 0.01%, 0% and 0% of the areas. NPP was mainly positively correlated with different rainfall levels. However, the negative correlation areas of extremely heavy rainstorms were relatively wide, with the proportion smaller than 50%. It is worth noting that there were significant spatial differences in the correlation between NPP and different rainfall levels (Fig. 3). The high-value areas of negative correlation between NPP and extremely heavy rainstorm were located in the central and north-central parts of this area (a). Heavy rainstorm had a relatively concentrated distribution in some parts of the south-central and northeastern regions (b). Rainstorm was found in a few areas of the south-central part (c). Heavy rain was distributed in the east-central and northeastern parts of the area (d). The areas with a negative correlation for moderate rain were scattered and relatively few, appearing sporadically in the north, southwest and east (e). After conducting an in-depth analysis of the correlation between Net Primary Productivity (NPP) and different rainfall levels, there was a significant phenomenon: the negative correlation between NPP and extremely heavy rain was is the most pronounced. This finding indicates that extremely heavy rain, as an extreme rainfall event, has a relatively strong inhibitory effect on the net primary productivity of vegetation. Extremely heavy rain is typically characterized by high-intensity precipitation. The large amount of rainfall in a short period can lead to soil waterlogging, decreased soil oxygen content, and deteriorated light conditions. All of them have an adverse effect on the photosynthesis and growth processes of vegetation, thereby reducing its NPP. The spatial distribution characteristics of the negative impact of different rainfall levels on vegetation NPP reveal that the central region is a concentrated area where vegetation NPP is negatively affected by different rainfall levels. This area may have unique combinations of topography, soil texture, and vegetation types, making its vegetation NPP more susceptible to negative impacts under different rainfall levels.

Through the analysis of the correlation between NPP and different rainfall levels, as well as the study of the central region where vegetation NPP is negatively affected by rainfall, we not only reveal the significant impact of extreme rainfall events on vegetation NPP, but also provide important references and insights for the protection and sustainable development of the vegetation ecosystem in the central region.

Responses of various forest species to different levels of rainfall

There are significant differences in the response of FVC to different rainfall levels in different forest vegetation types (Fig. 4). As the rainfall levels increased, the negative correlation between FVC and rainfall also increased. This was particularly evident when rainfall levels changed from heavy rainstorms to extremely heavy rainstorms, where the correlation between FVC and rainfall increased significantly. However, the impact of different rainfall levels on NDVI varies depending on forest vegetation. For extremely heavy rainstorm, broad-leaved forest (−0.42) and bamboo forest (−0.40) have the most significant impact, while eucalyptus (−0.26) and Chinese fir (−0.27) have relatively small impact. For heavy rainstorm, except for bamboo forest (−0.13), the impact on most forest vegetation types was not significant. In the case of rainstorm, Chinese fir (−0.20) showed a notable response, while eucalyptus (−0.12) was less affected. For heavy rain and moderate rain, the responses of forest types were consistent: economic forest (−0.19) and broad-leaved forest (−0.18) had strong impact, while eucalyptus (−0.11, and −0.12) showed minimal impact.

There are significant differences in the responses of NPP to rainfall levels in different forest vegetation types. Overall, as the rainfall levels increased, the negative correlation between NPP and rainfall increased. This trend was particularly pronounced when rainfall levels changed from heavy rainstorms to extremely heavy rainstorms, where the correlation between NPP and rainfall increased significantly. However, the effects of different rainfall levels on NPP varied depending on vegetation types. For extremely heavy rainstorms, the impact on eucalyptus forests (−0.21) were notable, while the effects on economic forests and broad-leaved forests (−0.13) were relatively small. Heavy rainstorms had a more pronounced impact on pine forests (−0.14) and a less impact on eucalyptus forests (−0.09). In the case of rainstorms, broad-leaved forest (−0.11) showed a stronger response, while eucalyptus forests (−0.07) were less affected. The sensitivity of different forest vegetation was consistent for heavy rain and moderate rain.

The comparison of different remote sensing parameters of vegetation shows that there was a weak correlation between varying rainfalls levels and three vegetation remote sensing parameters. The negative correlation was the highest with NDVI, followed by FVC, and then NPP. Indifferent rainfalls levels, the negative effects of extremely heavy rainstorms were the most pronounced, followed by heavy rainstorms. In contrast, the effects of rainstorms, heavy rain, and moderate rain were weak and showed little difference. In different forest vegetation types, the negative effects of varying rainfall levels on NDVI and FVC were more pronounced in economic forest and broad-leaved forest, while the eucalyptus forest was small. However, the negative effects on NPP showed minimal differences in the forest vegetation types.

There was a certain difference in the negative impact areas of different rainfall levels on FVC of different forest vegetation types (Fig. 5). As the rainfalls level increased from rainstorm to heavy rainstorm, the proportion of negatively impact areas for different forest vegetation types increases. However, in other rainfall transition periods such as moderate rain to heavy rain, heavy rain to rainstorm and heavy rainstorm to extremely heavy rainstorm, this trend is not significant. Due to different rainfalls levels, the average negative impact area in all forest vegetation types ranged from 45%–61%. In terms of severity, the negative impact ranked as: heavy rainstorm > extremely heavy rainstorm > moderate rain > heavy rain > rainstorm. During extremely heavy rainstorm, the negative impact of eucalyptus (69%) and broad-leaved forests (66%) was greater than Chinese fir forest (48%). For heavy rainstorms, broad-leaved forest (66%) experienced a higher impact than eucalyptus (57%) and bamboo forests (58%). Under rainstorm conditions, Chinese fir forest (61%) showed a greater negative impact than bamboo forests (26%). In the case of heavy rain, economic forests (71%) were more affected than eucalyptus forests (18%). Finally, under moderate rain, both economic (58%) and Chinese fir forests (56%) experienced a greater impact than eucalyptus forests (39%).

The negative impact of different rainfall levels on NPP varied in forest vegetation types. In general, as the rainfall levels increased from moderate rain to heavy rainstorm, the proportion of negative impacted area of most forest vegetation types increased significantly. However, when the rainfall changes from heavy rainfall to rainstorm, the proportion of areas negatively affected by economy, Chinese fir and pine forests will decrease. On average, the negative impact areas of all forest vegetation types ranged from 2%–39%, and ranked as: extremely heavy rainstorm > heavy rainstorm > rainstorm > heavy rain > moderate rain. For extremely heavy rainstorms, the negative impact area of eucalyptus (62%) was greater than that of economic forest (20%). Similarly, for heavy rainstorms, the impact area of bamboo forest (45%) was greater than that of broad-leaved forest (17%) and economic forest (14%). For rainstorm, the negative impact area of bamboo forests (26%) was greater than that of pine forests (3%), economic forest (2%) and Chinese fir forest (1%). For heavy and moderate rain, the negative impact areas in forest vegetation types were relatively small, ranging from 3% to 9% for heavy rain and from 1% to 3% for moderate rain.

Comparing different remote sensing vegetation parameters, it was found that there were significant differences in the proportion of negative impact areas of remote sensing for different rainfall levels and three vegetation covers. The proportion of negative impact areas of FVC was 53%, while the average of NPP was 16%. For FVC, heavy rainfall had the largest proportion of negative impacted areas, followed by moderate rainfall. The proportions of heavy and moderate rainfall were the smallest and relatively similar. In contrast, for NPP, rainfall was proportional to the size of impact area. Regarding forests vegetation types, the proportion of negative impact areas of FVC was higher in economic and broad-leaved forests, but lower in eucalyptus and bamboo forests. Conversely, for NPP, the proportion of eucalyptus and bamboo forests was high, while that of economic and broad-leaved forests is low.

Discussion

Areas with sparse vegetation have poor water and soil retention capacity, making them more susceptible to soil and water loss during heavy rainfall events, which may further hinder vegetation growth (Chen et al., 2015; Block & Richter, 2000). In this study, it was found that vegetation in the northeast, northwest and southeast areas had a significant negative correlation with moderate rain and heavy rain. This suggests that in the ecologically sensitive and fragile karst areas, moderate and heavy rain may also have a notable impact on vegetation, in addition to the effects of rainfall levels such as rainstorms. Rainfall in karst areas of Guangxi is mainly concentrated in spring and summer. The spatial distribution of high-value areas for annual extremely heavy rainfall, heavy rainfall, and storm rainfall shows great similarity (a, b, and c). The high-value areas for annual moderate rainfall and heavy rainfall also have significant similarity (d and e). The northeastern region (Guilin City) and the central-western region (Hechi City) of the study area were identified as the main rainstorm centers and areas with concentrated rainfall (Fig. 6) (Huang, Lin & Gao, 2012). The spatio-temporal distribution of areas where vegetation is highly sensitive to extremely heavy rainstorm, heavy rainstorm and rainstorms in this area is significantly from that of regions with frequent rainstorm (the center and southwest). This indicates that in addition to spatio-temporal differences in precipitation (Sun et al., 2021), other factors such as variations in bedrock exposure rate in karst areas (Chen et al., 2018) and the complex composition of vegetation types (Pan, Sun & Wang, 2021), are also key contributors to the pronounced spatio-temporal heterogeneity of vegetation responses to rainfalls in these regions.