Bedrock Type Mediates the Response of Vegetation Activity to Seasonal Precipitation in the Karst Forest

Abstract

Global warming is expected to enhance the severity and frequency of drought in subtropical areas; thus, understanding how vegetation growth responds to precipitation is crucial to comprehending the impact of these changes on ecosystem services, such as carbon storage. However, vegetation activity in subtropical karst regions in Southwest China is hard to explain when we merely consider the influence of climate and soil factors. In this study, we extended traditional research by combining bedrock data we aim to investigate the role of bedrock and its interaction with precipitation on plant growth in the Guizhou Province of China. We analyzed the differences in the precipitation–vegetation growth relationship in noncarbonate and carbonate rock regions, assessing the sensitivity of vegetation from two lithological types to drought. The results reveal that although there are no significant differences in climate and soil parameters between carbonate and noncarbonate regions, the normalized difference vegetation index (NDVI) in carbonate regions is more strongly related to precipitation (carbonate region: R² = 0.67; noncarbonate region: R² = 0.37), while the spring greenness–precipitation relationship show is more stable in the carbonate region. Our results show that the vegetation activity in the carbonate region is more vulnerable during the drought period, highlighting that the vegetation dynamic was not only regulated by climatic factors, and bedrock-caused water stress should be taken into account.

1. Introduction

The southwest of China is anticipated to undergo a series of changes due to global warming in this century, including temperature rising (3.0–4.0 °C) and altered pre-capitation patterns characterized by increased rainfall intervals, a shift from summer to spring distribution, and a decrease in the annual amount of rain [1]. Such changes would increase the intensity and frequency of droughts, resulting in a large reduction in vegetation growth [2]. Particularly, southwestern China is widely recognized as the most extensive Karst region globally, which is highly susceptible to water deficit [3,4,5]; therefore, understanding the vegetation growth response to the precipitation is crucial important to predict the climate–carbon storage feedback. However, many studies indicate that the precipitation–vegetation activity relationship commonly exhibits weak characteristics in the karst region: despite receiving sufficient precipitation, plants still suffer water stress, which suggests that the ecohydrological process is more complex in this region.

The spatial and temporal variation in water availability is strongly dependent on the function and structure of the earth’s critical zone, which means it is not only influenced by land surface climate and soil conditions [6,7,8], but also by deeper components of the ecosystem [9,10]. In the karst region, the significance of bedrock in regulating water availability cannot be disregarded. Karst developed from carbonate rocks, which are highly soluble in water. Carbonate rocks develop large numbers of crevices on their surface due to rainfall-induced water erosion, which significantly increases the bedrock permeability [11]. Moreover, because only small amounts of residues are left after dissolution, the regolith formation rate is shallow, and the thin regolith further limits its water-holding capacity. Consequently, drought events are more likely to occur in carbonate rocks regions (CRR), and it has been observed that for rainfall severe enough to surpass the soil’s field capacity, the water availability it supplies is only adequate to meet the transpiration requirements of plants for a duration of 7 to 14 days [12] (Figure 1).

In CRR, the temporal variation in water availability can be very strong, and drought events frequently occur in early spring and autumn [13,14], depending on precipitation patterns such as amount, intermittency, and magnitude. For example, Nerantzaki and Nikolaidis [15] show that spring drought events are more frequent in the Mediterranean regions because of the decrease in precipitation, whereas in the karst area of southern Italy, long time intervals are the main reason for drought events. Therefore, the frequent transient water scarcity resulting from soil water fluctuations is regarded as one of the most significant factors restraining vegetation growth [16,17]. Vegetation in CRR might be more vulnerable in the period with little or no precipitation. In contrast, vegetation in NCRR can use water contained in deep soil layers during drought episodes, thus showing greater resilience to drought episodes [18]. During a period with sufficient rainfall, the difference in vegetation productivity between CRR and NCRR might be reduced. In other words, vegetation growth is slower in CRR during periods with lower rainfall than in NCRR, but such differences may not exist as the rainfall increases.

Figure 1. Conceptual models of critical regions in carbonate (left) and noncarbonate (right) rock regions in southwest China (cited by Green [19]). Compared to noncarbonate rock, carbonate rock can more easily dissolve and develop several fractures and pavements, and surface water can quickly move through the rock underground. Moreover, carbonate rock usually develops thin regolith, which further influences the water-holding capacity of regolith. Thus, we assume that vegetation in the carbonate rock region is more sensitive to drought than in the noncarbonate region.

Figure 1. Conceptual models of critical regions in carbonate (left) and noncarbonate (right) rock regions in southwest China (cited by Green [19]). Compared to noncarbonate rock, carbonate rock can more easily dissolve and develop several fractures and pavements, and surface water can quickly move through the rock underground. Moreover, carbonate rock usually develops thin regolith, which further influences the water-holding capacity of regolith. Thus, we assume that vegetation in the carbonate rock region is more sensitive to drought than in the noncarbonate region.

As a main ecosystem of southwestern China, forests provide crucial ecosystem services, including watershed protection, erosion prevention, and carbon storage, and play a critical role in sustainable development. Moreover, forest ecosystems provide a habitat for diverse plants and animals. However, the response of forests to precipitation changes is still not clear. In this study, we assume the bedrock has the potential to obscure or distort the relationship between vegetation activity and environmental factors, and we investigate the indirect effects of lithology-associated features that affect vegetation growth in the Guizhou Province (one of the provinces in southwest China) based on a climate, soil, and bedrock database. We compare the vegetation dynamic in CRR with that in NCRR. Second, we test the vegetation growth response to precipitation in both regions. Third, we analyze the role of the bedrock type by considering how it changes the relationship between precipitation and vegetation activity. Specifically, we seek answers to the following two questions: (1) Does the relationship between precipitation and vegetation activity differ between CRR and NCRR? (2) Is vegetation growth more sensitive to droughts in CRR than in NCRR?

2. Materials and Methods

2.1. Study Site and Characteristics

Guizhou Province, situated in southwest China (Figure 2, 24°30′–29°13′ N, 103°01′–109°30′ E), encompasses a region of 176,167 km². The study region has a subtropical humid monsoon climate with annual average temperatures of 11–20 °C and annual precipitation of 730–2300 mm. Significant seasonal changes in precipitation occur, and 75% of the annual precipitation occurs in summer and fall (June–October) based on meteorological station data statistics for 2001–2010. Yellow soil, limestone soil, yellow-brown soil, red soil, purple soil and skeleton soil are the main soil types in Guizhou Province, covering 46.4%, 17.5%, 6.2%, 7.2%, 5.6% and 6.0% of the total area, respectively. Less than 38% of the province’s total area is made up of clastic rock exposure (namely NCRR), which is primarily found in Guizhou’s southeast and southwest regions with a small amount in the north. Regions underlain by carbonate rocks in Guizhou Province account for 62% of its total area. A total of 32.6% of the province is covered by assemblages of homogenous carbonate rocks, with the most widespread carbonate rocks being homogenous limestone, which accounts for 17.4% of the total area and is primarily found in western, southern, and southwestern Guizhou, with fewer smatterings in the east and north; and homogeneous dolomite, which makes up 13.1% of the entire province, is primarily found in the south–north belts in the province’s eastern portion, and is the second most prevalent cover. Dolomite interbedded with clastic rock covers 4.1%; mixed limestone and dolomite cover 2.1%; and lastly, dolomite/clastic rock alternations only comprise 2% of the entire province [20]. To ensure that all CRR developed into Karst landforms, we remove the area in which carbonate rock interbedded with classic rock from our analysis. Due to their thin regolith, karst ecosystems are highly vulnerable and prone to rocky desertification characterized by reduced vegetation cover and severe soil erosion [21]. Vegetation can conserve soil and prevent soil erosion and plays a vital role in maintaining ecosystem stability. Field observations have found that regions with relatively high rock CaCO3 content in Guizhou have developed relatively large crevices and undergone more significant rocky desertification [22].

The forests in Guizhou are dominated by evergreen broad-leaved forests [23], which are widespread across the region. These forests are characterized by species belonging to the families Fagaceae, Lauraceae, Theaceae, and Magnoliaceae. In addition, coniferous forests are also present, particularly stands of pine, cypress, and fir. These forests are found in cooler, higher-altitude regions. Guizhou is also home to several types of grassland vegetation. These grasslands are found in the province’s hilly and mountainous regions and are typically characterized by grassy tussocks and herbaceous plants. Some of these grasslands have been converted to agricultural land, but large tracts remain undisturbed [24].

2.2. Data Sources

The climate data used in this study originated from daily data collected from 78 meteorological stations in 2001–2010. The selection of the period was based on data availability. Moreover, we are confident that the average climate data being derived from a period of ten years is good enough to represent the difference in climate conditions between CRR and NCRR in Guizhou Province. The dataset includes data on precipitation, temperature, and evaporation. The data were downloaded from the National Meteorological Information Center of China (http://data.cma.cn/user/toLogin.html (accessed on 13 June 2022)). The normalized difference vegetation index (NDVI) data are commonly used to monitor vegetation growth. NDVI was derived from the Moderate Resolution Imaging Spectroradiometer (MODIS) Terra biweekly composite image data at a 500 × 500 m resolution (http://bjdl.gscloud.cn/ (accessed on 11 May 2023)) in our study. This product has been corrected for cloud contamination, directional reflectance, sun angle and shadow effects, and aerosol and water vapor effects by selecting the observation with the slightest view angle and the highest NDVI [25]. We selected 657 NDVI images from 2001 to 2010 to match the climate dataset. Forest distribution data were obtained from the Geographical Information Monitoring Cloud Platform (http://www.dsac.cn/DataProduct/Detail/200804 (accessed on 12 March 2023)) at a 500 × 500 m resolution based on Landsat TM/ETM/OLI remote sensing images.

The soil property data referred to here are from the China Soil Map Based Harmonized World Soil Database (v1.1). These data were obtained from the 1:1,000,000 Soil Map of China and were drawn using the results of the China Second National Soil Survey with 8979 representative soil profiles [26]. The lithology distribution map is cited from Wang et al. [27], who used an officially published 1:500,000 geological map as a data source. Noncarbonate and carbonate rocks occupy 38% and 32.6% of the total area in Guizhou Province, respectively. The remaining regions with a mixture of carbonate and noncarbonate rocks were not included in the analyses.

2.3. Data Treatment

First, the NDVI time series was extracted in both the CRR and NCRR. Then, the image with a cloud cover was removed to maintain the high quality of NDVI. The mean value of NDVI was calculated for further analysis, and a cubic spline interpolation in the time series replaced the missing values.

Two synthetic indices were calculated from the time series of NDVI values to reflect the seasonal vegetation activity. One is late spring to summer greenness [28], which represents the most active period of vegetation activity and is calculated as the sum of the NDVI from April to August. The higher values correspond to more active vegetation:

$$SG=\sum _{Apirl}^{August}\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}$$

(1)

Annual relative greenness [29] represents the effect of drought constraints, which is calculated as follows:

$$ARG=\frac{\mathrm{N}\mathrm{D}\mathrm{V}\mathrm{I}{dry}_{min}-NDVI{spring}_{min}}{{NDVI}_{max}-NDVI{spring}_{min}}$$

(2)

where NDVIdry_min represents the minimum NDVI during the dry period of September–October, NDVIspring_min represents the minimum NDVI before the onset of spring greenness (March–April), and NDVI_max represents the maximum NDVI during the whole year. Lower annual relative greenness values show the impact of drought on vegetation growth to be more severe. Both indices were resampled at the cell size of 1 km for their resolution to match environmental data. Over the 2001–2010 period, the biweekly sum precipitation (BSP), biweekly mean temperature (BMT), and biweekly sum evapotranspiration (BSE) were calculated for both NCRR and CRR.

2.4. Statistical Analysis

To explore the relationships between the NDVI and environmental variables, we performed polynomial regressions, where the NDVI was regressed as a cubic, quadratic, and linear function of climatic factors. The Akaike information criterion (AIC) determined the best polynomial regressions.

To identify the most critical factor in predicting the NDVI, we related environmental factors to the NDVI using generalized linear models (GLMs). In the models, environmental factors acted as explanatory variables, while the NDVI was the response variable. We fitted single-predictor GLMs for response variables as quadratic and linear functions of each explanatory variable. We compared these models via the AIC correction for all significant quadratic and linear terms. The collinearity of explanatory variables can influence the GLM analyses. To solve this problem, we evaluated correlations among all significant explanatory variables. Wherever two variables were strongly correlated (Pearson’s |r| > 0.70) [30], we excluded the variable exhibiting the weakest (i.e., greater AIC) independent relationship with the biweekly NDVI. Before carrying out the GLMs, we checked the normality and homoscedasticity of the variables. We use the log, sine, and square root transformations to obtain a normal distribution.

To evaluate how the bedrock type influences the relationship between vegetation activity and precipitation, we first analyzed the relationships between NDVI and precipitation in NCRR and CRR, respectively. Moreover, to test whether the vegetation activity is more sensitive to drought periods, we compared the difference in NDVI between NCRR and CRR in three precipitation classes (monthly precipitation < 100 mm; 100 mm < monthly precipitation < 200 mm; monthly precipitation > 200 mm) using a t-test. We linked SG and AGR with annual precipitation using the linear model, and then compared the slope of the linear model between the NCRR and CRR. Since the data did not meet the assumptions of normality, we used the Wilcoxon signed-rank test for these comparisons.

3. Results

3.1. Climatic and Soil Patterns in the Two Lithological Regions

The temperature and precipitation in Guizhou exhibit clear seasonal patterns: The precipitation data for the period 2001–2010 show that 61% of the precipitation was concentrated from May to August of each year, with only 17% of the precipitation occurring from January to April, and 21% occurring from September to December. In addition, the highest precipitation occurred in June, whereas the lowest occurred in December. The highest monthly mean temperature occurred in December, while the lowest monthly mean temperature occurred in July. There is no significant difference in mean annual temperature (MAT) or annual precipitation (AP) between CRR and NCRR (

Table 1. Comparison of climatic and soil factors between carbonate rock (CRR) and noncarbonate rock regions (NCRR). A t-test was employed to examine the differences.

	CRR		NCRR
	Mean	SD	Mean	SD
MAT (°C)	26.2	5.3	27.1	4.1
EVP (mm)	1056.1	67.2	1038.5	88.2
AP (mm)	996.0	56.1	1121.1	67.21
Soil sand (%)	13.2	5.1	13.8	3.1
Soil clay (%)	79.2	4.3	78.2	1.7
Soil silt (%)	15.26	2.3	15.2	1.2
Soil C (%)	5.2	11.2	5.1	11.1

MAT = mean annual temperature (°C); EVP = mean annual evapotranspiration; AP = annual precipitation (mm); soil sand = soil sand content (%); soil clay = soil clay content (%); soil silt (%).

). The soil properties, including soil carbon content, sand content, soil clay content, and soil slit content, exhibited no significant difference between the two lithological regions.

3.2. Seasonal Forest Growth Patterns in Two Lithological Regions

The temporal patterns of the NDVI between the two lithological regions were similar (Figure 3a, Pearson’s r = 0.88 ***): both regions showed a lower NDVI in spring (January–March) and winter (October–December) and exhibited peak values in August. However, the NDVI values in the NCRR were significantly higher than those in the CRR each month, with a few exceptions occurring in June and July 2006, and September 2008. However, the differences between them changed over time. The most significant differences in NDVI between the two lithological regions were pronounced in spring and winter, whereas the differences were lower during summer (January–August). Spring greenness and annual relative greenness values differed according to the lithological regions (Figure 3b): the spring greenness in the NCRR was significantly higher than that in the CRR. In contrast, the annual relative greenness had the opposite trend.

3.3. Relationships between the NDVI and Precipitation in the Two Lithological Regions

The relationships between NDVI and climatic factors were estimated based on biweekly data. The results showed that the climatic influence on vegetation varied between the two lithological regions (

Table 2. The generalized linear model (GLM) tests how climatic factors are related to the biweekly NDVI. Biweekly climatic factors related to energy and available water acted as explanatory variables, and NDVI acted as the response variable. * p < 0.05, ** p < 0.01.

	Biweekly Climatic Factors	NCRR	CRR
Energy	Actual evapotranspiration	−119.51 *	−161.11
	Sum solar radiation (J/cm2)	−122.34	−167.5
	Minimum temperature (°C)	−90.12	−166.31
	Maximum temperature (°C)	−102.33	−165.81
Available water	Sum precipitation (mm)	−117.62	−167.37 **
	Length dry periods (days)	−101.21	−174.11
	Maximum rainfall (mm)	−121.21	−152.14 *
	Sum evapotranspiration (mm)	−113.21	−161.11

, Figure 4). In CRR, the BSP emerged as the most important single predictor of the NDVI, with biweekly maximum rainfall following. In NCRR, however, none of the environmental factors showed a significant relationship with the NDVI, except actual evapotranspiration.

Further analysis showed that the NDVI exhibited a significant power relationship with precipitation in both regions. However, the fitness in CRR is much higher than in NCRR (Figure 4). The relationship between SG and precipitation also showed a significant difference between CRR and NCRR (Figures S1 and S2). The annual analysis showed that SG in CRR exhibited a positive relationship with precipitation from 2001 to 2010. In contrast, the relationship in NCRR was more varied. SG was negative related to precipitation in 2003, 2005, 2007 and 2010, while showing a positive relationship in the rest of years. The comparisons of the NDVI showed that in months with precipitation of less than 200 mm, the NDVI for NCRR was significantly higher than that of CRR. However, significant differences were not observed in months with precipitation greater than 200 mm (Figure 5). The slope of the linear model is greater in CRR. However, the relationship between ARG and precipitation did not show a significant difference between NCRR and CRR.

4. Discussion

Vegetation development strongly depends on the function and structure of the earth’s critical zone, which means it is not only influenced by land surface climate and topsoil conditions but also by deeper components of the ecosystem, such as bedrock (Figure 2). However, until now, the role of the belowground system in vegetation growth has been relatively poorly understood. We found that the interaction of vegetation with precipitation is mediated by bedrock type. The intermittent drought would be more frequent in the karst zone due to its low water capacity, which plays an essential role in moisture subtropic regimes. In forests characterized by carbonate rock, vegetation growth is more sensitive during the dry spell. In contrast, during periods with adequate precipitation, the difference in forest growth between lithological regions disappears. Notably, Qiao [31] found the NDVI of karst regions related to precipitation to be stronger than in no-karst regions during the drought period, which is consistent with our results. Our findings would encourage further research to clarify the importance of bedrock on vegetation.

Our findings are consistent with past research, proving the role of bedrock cannot be ignored. The bedrock is the foundation of soil formation and its properties can significantly impact the soil’s ability to store water. Zhong et al. [32] found that vegetation growth and soil water content strongly related to lithologies in the karst region of southwest China; by using meta-analysis, Zhu et al. [33] showed that bedrock has a significant relationship with regolith water capacity globally, which is crucial to vegetation growth. Moreover, in the karst area, the bedrock is composed of more permeable materials, such as limestone or sandstone; so, the water can easily pass through the bedrock [34]. In summary, the karst bedrock plays a crucial role in determining the water storage capacity of soil and subsequently affects the growth of vegetation by providing an essential resource for plant survival and growth.

It is imperative that maximum rainfall and sum precipitation are performed as an essential predictor of variation in NDVI in the CRR, which would suggest that the precipitation has a more crucial role in regulating the growth of karst forest. One plausible reason for this phenomenon is the high permeability and thin soil layer. CRR commonly exhibits a low water storage capacity [31,32]. Another possible reason is that the soil developed from carbonate rock is rich in clay, which has a low water-holding capacity because it lacks porosity. Notably, Nachtergaele et al. [34] found that limestone and dolomite usually cover clay soil, which is essential supporting evidence for this inference. It is also likely that because of the adaption strategy of plant species, excessive precipitation would create temporarily stable water sources, such as water in the temporarily saturated zone, and groundwater from lower depths; in addition, some species such as Bursera simaruba, can rapidly absorb and store the amount of water necessary to sustain a high water potential throughout periods of drought, and therefore, their growth can be maintained [2].

The AGE value is minimal for both karst and non-karst zones, suggesting that drought should be an essential limiting factor to the growth of forests in Guizhou Province [35,36,37]. However, the separate univariate analyses conducted for karst and non-karst forest in this study provide evidence that the drought in Guizhou may have been caused by different reasons (Table 2). Actual evapotranspiration was the best predictor for forest growth in the non-karst zone, indicating that the drought event may result from high temperatures. In contrast, the karst forest would be limited by the length of the period with rainfall since the water factor shows the best explanatory power for predicting NDVI in the karst zone. However, we cannot rule out the influence of temperature on karst forest since high temperatures can intensify drought stress during the lack of rainfall and increase the fire frequency in the karst zone [38,39]. Further research would benefit from exploring the interaction of precipitation and temperature in karst forests.

Although the forest in CRR suffers more drought stress, the vegetation activity only exhibits slightly lower values than those in NCRR (Figure 3). One possible reason for that is that the trees in the karst environment develop several special adaptabilities against drought stress [40]; for example, the trees in the karst zone often have thick and sponge leaves to reduce transpiration [41]; some plants can reduce the damage of a drought event by improving the activity of their antioxidant enzymes [42,43]. Such adaptive mechanisms can effectively reduce water loss and improve water use efficiency. However, it is noteworthy that the NDVI of karst forests is significantly lower than that of non-karst forests during spring and autumn when the drought is most severe. This result shows that karst forests would be highly vulnerable when drought occurs and reaches sufficient intensity.

5. Conclusions

The bedrock type can decouple vegetation responses from the climate, demonstrating that substrate prosperity can contribute to the vegetation dynamics in a complex way, suggesting the importance of expanding our focus beyond surface climate and soil, and exploring the connections of each critical component. Our findings can be used to simulate the potential consequences of future climate changes for the different bedrock types. For vegetation located in CRR, a wetter climate can benefit vegetation greenness. In past decades, land management (such as the afforestation project) in Guizhou Province has shown a significant positive impact on vegetation growth; however, the increase in the frequency and intensity of drought events needs to be considered. As regions with carbonate bedrock cover 15% of the Earth’s surface, this threshold should be considered when studying vegetation growth responses to climate change in a karst landscape. To further improve our understanding of bedrock impacts, it is necessary to carry out field observation and control experiments to clarify the water-holding capacity of different bedrock zones. Meanwhile, we also need more details on species-specific drought adaptation mechanisms in karst forests, which can be crucial to enhance our accuracy regarding the prediction of how karst forests respond to climate change.