Impact of bedrock geochemistry on vegetation productivity depends on climate dryness in the Guizhou karst of China

Abstract

The Guizhou karst area is one of the largest continuous areas of karst in the humid climate zone and is representative of karst landforms in China. Large portions of the karst system are characterized by extremely shallow soils underlain by weathered bedrock and water deficits are common. Although the distribution of ecosystem productivity is largely related to variations in the temperature and precipitation, the influence of the substrate in karst areas requires further exploration. We explored the relative importance of the bedrock geochemistry (characterized by the concentrations of Ca, Mg and Si) and climatic factors (temperature and precipitation) to explain the spatial variability in gross primary productivity (GPP) with various degrees of water deficit during the time period 2001–2015. Our results show that the impact of bedrock geochemistry is an important parameter in changing the original relationship between climate and the GPP. The bedrock geochemistry functioned as a “regulator” of the relation between climate and the GPP, which strengthened with decreasing climate favourability. The variations in GPP and surface water storage were significantly different when different elements (Ca, Mg or Si) were dominant. The Mg-rich regions showed the greatest annual variations in the GPP, whereas the Si-rich regions had the strongest surface water storage potential to support vegetation growth. The results of our study are important for systematically evaluating the effects of climate on vegetation productivity and provide a benchmark for global vegetation modelling predictions.

Keywords

Karst Climate Bedrock geochemistry Random forest structural equation model (SEM)GPP

1 Introduction

Ecological hydrology, which aims to understand the complex interactions between vegetation and climate, was developed under the framework of the soil–plant–atmosphere continuum (SPAC), but the effect of the substrate in SPAC processes has not been explored systematically (Katul et al., 2012; Fatichi et al., 2016). The surfaces of the rocks in the complex underground structure of karst areas are cracked, which challenges the conventional patterns of the storage of precipitation in soils for use in vegetation growth.

Karst areas cover 7−12% of the Earth’s continental area and a large percentage of karst areas are characterized by extremely shallow soils underlain by weathered bedrock (Gutierrez et al., 2014). The weathered rock zone is usually thicker than the soil and becomes a large storage reservoir for the supply of available water to plants (Bonacci et al., 2009). Fractures in the weathered rock zone are commonly filled with a mixture of organic material and coarse grains, the hydraulic properties of which resemble those of soil and correspondingly support high root densities (Bornyasz et al., 2005). Some studies have also shown that the uptake of water by plants is not restricted by shallow soils because water is also taken up from the rock layers below (Schwinning, 2010). Water-storing layers of regolith, epikarst and cemented soil horizons have the potential to hold as much water as soil; however, water capture from below the soil horizons has been mostly neglected. Weathering has been shown to transform poorly conductive bedrock into a dynamic water storage reservoir (Dammeyer et al., 2016). The term “rock moisture” was proposed to describe the exchangeable water stored in the unsaturated zone in weathered bedrock. In the dry season, rock moisture storage is gradually depleted by trees through transpiration, leading to a lower value at the end of the dry season (Rempe and Dietrich, 2018). These results provide new insights and areas for further study of the effect of the mechanisms of storing moisture in rocks on surface vegetation, especially in karst areas.

Previous studies have shown that the weathering status of bedrock is closely related to the bedrock geochemistry (Ma 2002). Dissolution is the main factor in the formation and development of the karst environment (Jiang et al., 2014). The epikarst is well developed on carbonate bedrock and can be described by the following chemical equilibrium equations:

{CaCO}_{3} {+ H}_{2} {O + CO}_{2} {= Ca}^{2+} {+ 2HCO}_{3}^{2-}

{MgCO}_{3} {+ H}_{2} {O + CO}_{2} {= Mg}^{2+} {+ 2HCO}_{3}^{2-}

The process of dissolution dynamics involves the transfer and exchange of elements, which form a unique above-ground and underground system. However, only a limited number of case studies have investigated whether the bedrock geochemistry might indirectly affect the local growth of vegetation.

The widespread area of karst in Guizhou Province makes it an ideal location for a study of the critical zone (Li et al., 2003). The Guizhou karst area, one of the largest areas of continuous karst in the humid climate zone, covers ∼109,084 km² and is representative of karst landforms in China (Peng and Wang, 2012). Previous studies have shown that the contribution of climate to vegetation in the karst area of SW China is <30% (Yan et al., 2018), followed by 21 and 13.2% from soil and topographic factors, respectively, indicating that these parameters may not be dominant in determining the spatial patterns of vegetation in this region (Hu et al., 2017). The availability of water is crucial to the growth of vegetation in karst areas (Wang et al., 2013). Water deficits occur from time to time in this region, even with sufficient rainfall and heat. This kind of water deficit represents a drought in the context of a humid climate and differs from that in dry climate zones (Hartmann et al., 2014). The characteristic substrates with shallow soils and the uncertain rock water storage capacity require further exploration to explain the heterogeneity of the distribution of vegetation (Schwinning, 2010). To our knowledge, the importance and effects of bedrock geochemistry on vegetation growth have rarely been discussed and are still poorly understood in the Guizhou karst area.

We used the random forest method to explore the relative importance of the bedrock geochemistry and climate factors to explain the spatial variability of the gross primary productivity (GPP) in the time period 2001–2015, with varying degrees of water deficit. These results contribute to a more comprehensive understanding of how the substrate conditions affect vegetation growth in karst regions.

2 Materials and methods

2.1 Study region

The study region covers all of Guizhou Province, which is >73% karst (Figure 1a). The study region is characterized by a subtropical humid monsoonal climate with a mean annual temperature of 8−22°C and mean annual precipitation of 850−1550 mm. The temperature, precipitation and vegetation productivity all show a generally increasing trend from NW to SE. The Guizhou karst region is characterized by extremely thin soils or soil-like organic matter between the root systems of plants and the rocks, or limited rooting space caused by dense fissures (Figure 1e). The geological formations of the karst are closely related to the bedrock geochemistry. Overall, 64% of Guizhou Province contains a variety of different carbonate rock types and the remaining 36% consists of sandstone and metamorphic rocks, which are typically rich in Si and have low Ca contents (Zhang 2000).

Figure 1.

(a) Location and distribution of karst in Guizhou Province. Spatial patterns of the (b) mean annual temperature, (c) annual precipitation and (d) gross primary productivity. Spatial patterns of the concentrations of (e) Ca, (f) Mg and (g) Si. Photographs of karst vegetation with roots in rock fractures in Guizhou: (h) trees on carbonate and (i) grasses on dolomite.

2.2 Datasets

2.2.1 Climate data

The observed daily data from >70 meteorological stations distributed across the province were used to generate a 1 km spatial interpolation dataset, including the mean annual precipitation and the mean annual temperature. Spatial interpolation was performed using ANUSPLIN interpolation software, which is a widely used tool for analysing and interpolating multivariate data using a smooth spline function (McVicar et al., 2007). It provides a method of approximating a surface using a function that can perform reasonable statistical analysis and data diagnosis and analyse the spatial distribution of data to realize the function of spatial interpolation.

2.2.2 Standardized precipitation evapotranspiration index

Standardized precipitation evapotranspiration index (SPEI) data with a spatial resolution of 0.5° were calculated based on the monthly precipitation and potential evapotranspiration estimated by the Penman–Monteith equation of the Climatic Research Unit TS 3.23 dataset (http://sac.csic.es/spei/database.html). The SPEI is a variable standardized relative to the long-term climatic balance and the smaller the value, the drier the climate (Beguería et al., 2014).

2.2.3 Temperature vegetation drought index

The temperature vegetation drought index (TVDI) is an effective index generated from optical remote sensing imagery to monitor the regional status of surface waters. It was developed from the Vegetation Supply Water Index, which is based on the division of the Normalized Difference Vegetation Index (NDVI) by the land surface temperature (Carlson et al., 1994). It represents the surface water storage deficit, which is different from the atmospheric water deficit (SPEI) (Zhao et al., 2017). We also used the top 10% of each year’s surface water reserves to represent the surface water storage potential. The remote sensing data used in this study are the MODIS product data MOD13C2 for the NDVI and MOD11C3 for the land surface temperature. The processes of correcting, clipping and calculation were implemented in the Environment for Visualizing Images (ENVI5.2), which is a familiar and well-known remote sensing image processing platform (https://www.esrichina.com.cn/EnviZongheye.html). This index was set between 0 and 1, with larger values indicating that less surface water is available for the growth of vegetation.

2.2.4 Geochemistry data

There are three main types of sedimentary bedrock in Guizhou Province: limestone, dolomite and sandstone (or blastopsammite) (Ma, 2002). The indicator elements corresponding to the three bedrocks are Ca, Mg and Si, respectively. The differences in the subsurface rocks can be largely reflected by differences in the content of these three elements. Ca has the highest reaction activity with water containing CO₂, followed by Mg and Si, thus the differences in bedrock composition lead to different degrees of weathering. The geochemical distributions of Ca, Mg and Si were obtained from the Geochemical Atlas of Guizhou Province provided by the China Geological Survey. This atlas was prepared at a scale of 1:200,000 based on regional geochemical scanning using 1,886,164 samples collected at a spatial resolution of 1 km², including stream sediment, soil and rock samples (Feng, 2009).

2.2.5 Gross primary productivity

A long-term series of the terrestrial GPP was derived from MOD17A2, which has been widely used in many studies (Sun et al., 2019). Version 6 is the latest generation of the MOD17A2 GPP product with a native spatial resolution of 500 m. It can be synthesized to different timescales (e.g. weekly, monthly or yearly). To match the spatial resolution of the climate and geochemical distribution data, the spatial resolution of the annual GPP was resampled to 1 km using the nearest neighbour method in ARCGIS 10.2.2. We used MOD17A2 GPP data from 2001 to 2015, which are available from https://e4ftl01.cr.usgs.gov/MOLT/MOD17A2H.006/. The GPP percentage anomaly was used for further analysis and is calculated as follows:

\begin{matrix} G P P p e r c e n t a g e a n o m a l y \\ = \frac{{GPP}_{i} - \bar{GPP}}{\bar{GPP}} \times 100 % (1) \end{matrix}

\begin{matrix} G P P C V = \frac{SD}{\bar{GPP}} (2) \end{matrix}

where GPP is the annual gross primary productivity in year i and $\bar{G P P}$ is a multi-year average of the GPP between 2001 and 2015; SD is the standard deviation.

2.3 Methods

We used a random forest machine learning approach to generate a predictive model capable of detecting the relative importance of climate factors and bedrock geochemistry in the spatial variability of the GPP (Breiman, 2001). We then used univariate analysis of variance to detect whether there was a significant difference in the performance of the bedrock geochemistry under different drought scenarios. We developed a structural equation model (SEM) to distinguish the regulating role of the bedrock geochemistry in climate forcing.

2.3.1 Random forest

The random forest is an machine learning algorithm that performs well even without parameter tuning and descriptor selection (Svetnik et al., 2003). The modelling paradigm of machine learning differs from that of statistical approaches. Statistical approaches to model fitting start by assuming an appropriate data model. By contrast, machine learning avoids starting with a defined data model and assumes that the data generation process is complex and unknown. Machine learning uses an algorithm to learn the responses by observing the inputs and responses to find the dominant patterns. The random forest is a mature algorithm that can deal with a large number of different types of descriptor simultaneously, handling redundant/irrelevant descriptors and integrating interactions and multiple mechanisms of action, which can provide the importance of descriptors while having a high prediction accuracy (Schwalm et al., 2017).

Five descriptors – including the annual mean temperature, annual precipitation, and Ca, Mg and Si concentrations – were selected for use in the random forest model to quantify the relative importance of the descriptors. The analysis was performed using the packages DMwR, ipred and randomForest, respectively, in R 3.5.1. The flowchart is shown in Figure 2.

Figure 2.

Flowchart used to quantify the impacts of factors on the gross primary productivity in Guizhou by the random forest method. D is a dataset composed of a large number of random samples (i = 2001, 2002, 2003…2015.)

2.3.2 Linear regression analysis

The relationships between ecosystem production and environmental factors were expressed as correlation coefficients. Single linear regression analysis was conducted between the GPP anomalies and the relative importance of environmental factors on an annual scale.

2.3.3 Univariate analysis of variance

ANOVA is a common and useful tool for statistical tests of factors in experiments (Hurlbert, 1997). We used the median value to divide the GPP and the drought index equally into high- and low-value groups to determine whether there was a significant difference in the performance of bedrock chemistry under different drought scenarios.

2.3.4 Structure equation model

We developed a SEM to further clarify the contribution of bedrock geochemistry, while simultaneously accounting for the role of the climate in affecting the GPP impacted by the bedrock geochemistry. SEM is a multiple regression analysis used to display the direct and latent relationships among variables. Paths between variables are from independent to dependent variables, with a directional arrow for every regression model. Each path has a standardized coefficient, including direct and indirect effects (Rosseel, 2012).

In our study system, the effect of the bedrock geochemistry was hypothesized to regulate the original relationship between the climate and GPP, which performed differently with varying degrees of water deficit. The SPEI for atmospheric water deficits and the TVDI for shortages in surface water storage were adopted to construct the SEM. Figure 2 shows the plausible interaction pathways derived directly from our theory. We then sequentially eliminated non-significant pathways to obtain a parsimonious set of models (Figure 3). However, removing these pathways did not improve the model and reduced the goodness of fit. The obtained SEMs met the criteria requiring that the p values of χ ² and the goodness of fit test p > 0.05, the comparative fit index > 0.9 and the root mean square error of approximation < 0.05. The SEM was conducted using Amos Version 21.0 (Amos Development Corporation, Chicago, IL, USA).

Figure 3.

Illustration of all plausible interaction pathways derived directly from our theory that the bedrock geochemistry is hypothesized to regulate the original relationship between the climate and the GPP, which performs differently with varying degrees of water deficit. RI: relative importance.

3 Results

3.1 Relative importance of the climate and bedrock geochemistry to the GPP

The multi-year average of the contribution of climate to productivity was about 3% higher than that of the bedrock geochemistry. However, the median results were the opposite, with the importance of the bedrock geochemistry higher than that of the climate (Figure 4a). The large variation in the relative importance suggested that a difference might exist under particular conditions in which the bedrock geochemistry may regulate the strength of the climate–vegetation relationship. However, neither the temperature nor precipitation alone showed a significant relationship with the GPP anomaly, whereas, as a whole, they were significantly correlated with the GPP anomaly (r = 0.51, P < 0.05). The relative importance of the climate was enhanced with increased GPP values, showing that in years with high productivity (GPP anomaly >0), the climate made a high contribution to the GPP, whereas in years with a low productivity (GPP anomaly <0), the relative importance of the climate was small (Figure 4b).

Figure 4.

Relative importance of the climate and bedrock geochemistry to productivity (a). Box-plots show the mean values (small box), the median (the horizontal line in the box), the 25th and 75th percentiles (top and bottom of the box), and the maximum and minimum observed values (edges of top and bottom whiskers). The relationship between the anomalous GPP and the relative importance of the climate (b). The blue dots represent years of high productivity with $GPP Percentage Anomaly$ >0, and the orange dots represent years of low productivity with $GPP Percentage Anomaly$ <0. The red line shows the modelled effect, and the grey-shaded area represents the 95% confidence interval.

3.2 Effect of bedrock geochemistry on the role of climate in affecting the GPP

The relative importance of the bedrock geochemistry showed similar values under two opposite climate conditions: a wet climate with a high SPEI and a dry climate with a low SPEI. The importance of the bedrock geochemistry was relatively high in years with low TVDI values when the land surface was moist, when it was almost 13% higher than that in high TVDI years. Similar to the TVDI, the importance of the bedrock geochemistry during high GPP years was about 9% lower than that during low GPP years (Figure 5a).

Figure 5.

(a) Effect of bedrock geochemistry under different drought indexes. ANOVA of the relative importance of the bedrock geochemistry under different SPEI, TVDI and gross primary productivity conditions. The 50th percentile is used to classify low and high groups for the SPEI, TVDI and gross primary productivity. The box-plots show the mean values, the black lines show the error bars and * represents the significance of the difference. (b) A structural equation model is used to confirm the effects of the bedrock geochemistry under different climatic conditions. The rectangles represent the actual variables and the ellipses represent the potential variables. The solid blue arrows represent significant paths and the grey lines indicate non-significant pathways. The numbers near the lines indicate the standard path coefficients and * represents P <0.05. GPP: gross primary productivity.

We fitted paths for all the environmental variables that directly or indirectly affected the productivity and the final model passed the statistical test (P > 0.05). The contribution of bedrock geochemistry in the middle worked as a “regulator” of the relation of climate to the GPP (Figure 5b). Temperature and precipitation had both direct and indirect effects on the GPP, with the direct effects of both being positive. The indirect effects of the climate on the GPP were regulated by the bedrock geochemistry. A significant negative effect of the bedrock geochemistry on the GPP weakened the relationship in which climate favourability showed a consistent positive correlation with the GPP. The contribution of the bedrock geochemistry varied according to the climate and it strengthened with decreasing climate favourability. The fitting results of the SEM indicated that the impact of the bedrock geochemistry was exerted through changing the original relationship between the climate and the GPP by substrate processes such as water storage.

3.3 GPP variation and water storage potential under different dominant elements

After showing that the bedrock geochemistry played a significant part in the available surface water and the GPP, we further demonstrated how the two (surface available water and GPP) differed in areas where Ca, Mg and Si were enriched. The coefficient of variation for the GPP (GPP_CV) was the largest in the Mg-rich region (Figure 6a). The productivity of vegetation in Si-rich regions was the least varied and that of Ca-rich regions was in between (Figure 6a). Si-rich regions had the strongest surface water storage potential to support vegetation growth, followed by the Ca-rich regions (Figure 6b). The Mg-rich regions had the lowest water storage potential.

Figure 6.

Differences in the coefficient of variation in the gross primary productivity and the surface water storage potential in the three element-rich regions (a b). The 5th percentile was used as a criterion for screening an element-rich region. GPP: gross primary productivity.

4 Discussion

It has been widely accepted that the climate is the most important factor affecting vegetation growth and that climatic extremes, such as drought, can cause a decrease in vegetation growth (Allen et al., 2010; Eamus et al., 2013). Our results indicated that bedrock geochemistry may be an easily overlooked regulator of the climate–vegetation relationship. The results of variance analysis showed that the relative importance of bedrock geochemistry did not perform significantly differently under different atmospheric water deficits represented by the SPEI. However, the importance of the bedrock geochemistry markedly increased with increased surface water storage and reduced productivity. These results suggest a strong regulation effect of the bedrock geochemistry on the availability of water for the growth of above-ground vegetation in Guizhou Province.

The bedrock geochemistry influenced the effect of climate on the GPP, which could be explained by rock weathering (Jiang et al., 2020). There was no doubt that the weathered rock products showed water storage capacities – for example, some highly weathered limestone products had water storage capacities of 0·35–0·50 m³ m⁻³ and supported the uptake of water by plants (Querejeta et al., 2006). Our results show that the variations in the productivity of vegetation in Mg-rich regions were the greatest, followed by the variations in Ca-rich regions. The differences in Ca and Mg contents were related to the degree of weathering, the number and size of crevices, and the water retention capacity (Wang et al., 2004; Hong et al., 2018). Ca-rich regions were more likely to dissolve and eventually form large, widespread fractures on the weathered surface (Wang et al., 2001), leading to a high investment in deep root growth through the crevices (Schwinning, 2010). By contrast, the chemical weathering of the Mg-rich regions was slower than that of the Ca-rich regions. The weathered surface of Mg-rich regions eventually formed dense, but narrow, fractures that were suitable for shallow-rooted plants, for which productivity is more prone to annual fluctuations than deep-rooted plants in Ca-rich regions (Liu et al., 2019). Si-rich regions had the strongest surface water storage potential to support vegetation growth, followed by Ca-rich regions, whereas the Mg-rich regions had the poorest water storage potential. In the Guizhou karst area, Si-rich regions had a greater possibility of forming surface soils with a high clay content than the Ca- or Mg-rich regions, which tended to form epikarst (Buss et al., 2017). In general, as the clay content increased, the size of the soil pores decreased, capillary action strengthened, and the soil water retention capacity was enhanced (Tian et al., 2017). Compared with Ca- or Mg-rich regions, the Si-rich regions provided a high water storage potential for vegetation growth by creating a weathered layer with smaller pores and holding water at higher suction pressures (Abuel-Naga and Bouazza, 2010).

Previous studies on the impact of bedrock geochemistry on vegetation mainly focused on two aspects: the nutrient elements and heavy metal elements (Zhang et al., 2015). The bedrock geochemistry was either a neglected source of nutrients or contained heavy metal elements that inhibited vegetation growth (Morford et al., 2011; Hahm et al., 2014). By contrast, our research focused on the interannual differences in the contribution of bedrock geochemistry to vegetation growth. Although the distribution of ecosystem productivity was better explained by the temperature and precipitation based on the multi-year average, this relationship changed under varied climate conditions. The large variation in the relative importance suggested that the bedrock geochemistry may regulate the strength of the climate–vegetation relationship.

The soil water balance module in most dynamic vegetation models usually assumes that the soil is homogeneous and has a certain thickness (Jung et al., 2007). The lithology is hardly considered, which might lead to bias in the estimation of vegetation growth. The elemental composition of the bedrock has also been rarely considered in ecohydrological models of karst (Hartmann et al., 2014). Previous studies have shown that if geochemical processes are considered in ecohydrological models of karst, then the models are more complex (Kirchner, 2010). The accurate simulation of the inflow path of precipitation, the residence time and the outflow path has remained challenging (Hartmann et al., 2012). At present, it has become common for researchers to predict karst ecological and hydrological processes based on artificial neural networks, least-squares support vector machines, machine learning or other algorithms (Li et al., 2018). These models almost completely avoid the problem of parameter acquisition and often have high accuracy. Our results obtained using machine learning showed that the bedrock property-related substrate had a more important impact on the above-ground vegetation than previously expected, especially under drought stress. Our results further showed the dominant element that caused the difference in the importance of the bedrock geochemistry under different drought scenarios and provided a new way to further predict the future impact of climate on vegetation in karst regions.

5 Conclusions

After a comprehensive, broad-scale assessment of the relative importance of the climate and bedrock geochemistry, we found that the effects of climate on the GPP were regulated by the bedrock geochemistry. The contribution of the bedrock geochemistry varied according to the climate, strengthening with decreasing climate favourability. Si-rich regions had the strongest surface water storage potential to support vegetation growth, followed by Ca-rich regions. The Mg-rich regions had the lowest water storage potential and the greatest annual variation in the GPP. The vegetation productivity in Si-rich regions was the least varied and that in Ca-rich regions was in between. Our results, which suggest the importance of bedrock geochemistry to vegetation growth, could contribute to systematically understanding the effects of different environmental factors on local vegetation growth in karst regions.

Footnotes

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This study was funded by the National Natural Science Foundation of China (No. 41571130044).

ORCID iDs

Xinrong Zhu

Hongyan Liu

Boyi Liang

References

Abuel-Naga

Bouazza

(2010) A novel laboratory technique to determine the water retention curve of geosynthetic clay liners. Geosynthetics International 17: 313–322.

Allen

Macalady

Chenchouni

, et al. (2010) A global overview of drought and heat-induced tree mortality reveals emerging climate change risks for forests. Forest Ecology and Management 259: 660–684.

Begueria

Vicente-Serrano

Reig

, et al. (2014) Standardized precipitation evapotranspiration index (SPEI) revisited: parameter fitting, evapotranspiration models, tools, datasets and drought monitoring. International Journal of Climatology 34: 3001–3023.

Bonacci

Pipan

Culver

DC.

(2009) A framework for karst ecohydrology. Environmental Geology 56: 891–900.

Bornyasz

Graham

Allen

MF.

(2005) Ectomycorrhizae in a soil-weathered granitic bedrock regolith: Linking matrix resources to plants. Geoderma 126: 141–160.

Breiman

LJML.

(2001) Random forests. Mach Learn, 45: 5–32.

Buss

Lara

Moore

, et al. (2017) Lithological influences on contemporary and long-term regolith weathering at the Luquillo Critical Zone Observatory. Geochimica et Cosmochimica Acta 196: 224–251.

Carlson

Gillies

Perry

(1994) A method to make use of thermal infrared temperature and NDVI measurements to infer surface soil water content and fractional vegetation cover. Remote Sensing Reviews 9: 161–173.

Dammeyer

Schwinning

Schwartz

, et al. (2016) Effects of juniper removal and rainfall variation on tree transpiration in a semi-arid karst: Evidence of complex water storage dynamics. Hydrological Processes 30: 4568–4581

10.

Eamus

Boulain

Cleverly

, et al. (2013) Global change-type drought-induced tree mortality: Vapor pressure deficit is more important than temperature per se in causing decline in tree health. Ecology and Evolution 3: 2711–2729.

11.

Fatichi

Pappas

Ivanov

. (2016) Modeling plant–water interactions: An ecohydrological overview from the cell to the global scale. WIRES-Water 3: 327–368.

12.

Feng

(2009) Geochemical Atlas in Guizhou Province China. Geological Press, Beijing, China. [in Chinese].

13.

Gutierrez

Parise

De Waele

, et al. (2014) A review on natural and human-induced geohazards and impacts in karst. Earth-Science Reviews 138: 61–88.

14.

Hahm

Riebe

Lukens

. (2014) Bedrock composition regulates mountain ecosystems and landscape evolution. Proceedings of the National Academy of Sciences of the United States of America 111: 3338–3343.

15.

Hartmann

Lange

Weiler

, et al. (2012) A new approach to model the spatial and temporal variability of recharge to karst aquifers. Hydrology and Earth System Sciences 16: 2219–2231.

16.

Hartmann

Goldscheider

Wagener

, et al. (2014) Karst water resources in a changing world: Review of hydrological modeling approaches. Reviews of Geophysics 52: 218–242.

17.

Hong

Xie

Y-Q

Qin

X-M

, et al. (2018) Geochemical characteristics and ecological interaction of earth elements in the soil-plant system in northwest Guizhou Province. Journal of Ecology and Rural Environment 34: 890–896.

18.

Zeng

, et al. (2017) Plant community characteristics and their relationships with soil properties in a karst region of southwest China. Contemporary Problems of Ecology 10: 707–716.

19.

Hurlbert

(1997) Experiments in ecology. Endeavour 21: 172–173.

20.

Jiang

Lian

Qin

(2014) Rocky desertification in Southwest China: Impacts, causes, and restoration. Earth-Science Reviews 132: 1–12.

21.

Jiang

Liu

Wang

, et al. (2020) Bedrock geochemistry influences vegetation growth by regulating the regolith water holding capacity. Nat Commun 11: 2392.

22.

Jung

Vetter

Herold

, et al. (2007) Uncertainties of modeling gross primary productivity over Europe: A systematic study on the effects of using different drivers and terrestrial biosphere models. Global Biogeochemical Cycles 21 : 553–566.

23.

Katul

Oren

Manzoni

, et al. (2012) Evapotranspiration: A process driving mass transport and energy exchange in the soil-plant-atmosphere-climate system Reviews of Geophysics 50.

24.

Kirchner

. (2010) A double paradox in catchment hydrology and geochemistry. Hydrological Processes 17: 871–874.

25.

Wang

Zhou

, et al. (2003) The correlation between rock desertification and lithology in karst area of Guizhou. Acta Geographica Sinica 58: 314–320. [in Chinese]

26.

, et al. (2018) Dam construction impacts on multiscale characterization of sediment discharge in two typical karst watersheds of southwest China. Journal of Hydrology 558: 42–54.

27.

Liu

Jiang

Dai

, et al. (2019) Rock crevices determine woody and herbaceous plant cover in the karst critical zone. Science China Earth Science 11: 1756–1763

28.

2002. China Geological Atlas [in Chinese]. Beijing: Geological Publishing House.

29.

McVicar

Van Niel

Hutchinson

Liu

(2007) Spatially distributing monthly reference evapotranspiration and pan evaporation considering topographic influences. Journal of Hydrology 338: 196–220.

30.

Morford

Houlton

Dahlgren

RA.

(2011) Increased forest ecosystem carbon and nitrogen storage from nitrogen rich bedrock. Nature 477: 78–88.

31.

Peng

Wang

. (2012) Effects of land use, land cover and rainfall regimes on the surface runoff and soil loss on karst slopes in southwest China. Fuel & Energy Abstracts 90: 53–62.

32.

Querejeta

Estrada-Medina

Allen

, et al. (2006) Utilization of bedrock water by Brosimum alicastrum trees growing on shallow soil atop limestone in a dry tropical climate. Plant and Soil 287: 187–197.

33.

Rempe

Dietrich

. (2018) Direct observations of rock moisture, a hidden component of the hydrologic cycle. Proceedings of the National Academy of Sciences of the United States of America 115: 201800141.

34.

Rosseel

(2012) lavaan: An R package for structural equation modeling. Journal of Statistical Software 48: 1–36.

35.

Schwalm

Anderegg

WRL

Michalak

, et al. (2017) Global patterns of drought recovery. Nature 548: 202–205.

36.

Schwinning

(2010) The ecohydrology of roots in rocks. Ecohydrology 3: 238–245.

37.

Sun

Zhou

, et al. (2019) Precipitation mediates the temporal dynamics of net primary productivity and precipitation use efficiency in China’s northern and southern forests. Annals of Forest Science 76: 12.

38.

Svetnik

Liaw

Tong

, et al. (2003) Random forest: A classification and regression tool for compound classification and QSAR modeling. Journal of Chemical Information and Computer Sciences 43: 1947–1958.

39.

Tian

Wang

, et al. (2017) Characterization of nano to micron-scale pore structure in tight sand gas reservoir from Turpan-Hami Basin. Journal of Nanoscience and Nanotechnology 17: 6096–6108.

40.

Wang

Zhou

Xiao

BX.

(2013) Progress of study on karst soil moisture characteristics of southwest China. Soil & Water Conservation in China.

41.

Wang

Sun

, et al. (2004) How types of carbonate rock assemblages constrain the distribution of karst rocky desertified land in Guizhou Province, PR China: Phenomena and mechanisms. Land Degradation & Development 15: 123–131.

42.

Wang

Luo

(2001) Geostatistical and geochemical analysis of surface water leakage into groundwater on a regional scale: A case study in the Liulin karst system, northwestern China. Journal of Hydrology 246: 223–234.

43.

Yan

, et al. (2018) Spatial and temporal patterns of changing vegetation and the influence of environmental factors in the karst region of Southwest China. Acta Ecologica Sinica 38: 8774–8786.

44.

Zhao

Cong

, et al. (2017) Spatial-temporal variation of drought in China from 1982 to 2010 based on a modified temperature vegetation drought index (mTVDI). Scientific Reports, 7: 17473.

45.

Zhang

Guo

(2015) The source, spatial distribution and risk assessment of heavy metals in soil from the Pearl River Delta based on the National Multi-Purpose Regional Geochemical Survey. PLoS One 10: 12.

46.

Zhang

(2000) Construction and use of digital geological map of 1:500000 in Guizhou Province. Guizhou Geology 17: 14-17.