Gray Forecast of Ecosystem Services Value and Its Driving Forces in Karst Areas of China: A Case Study in Guizhou Province, China

A sound ecosystem is the prerequisite for the sustainable development of human society, and the karst ecosystem is a key component of the global ecosystem, which is essential to human welfare and livelihood. However, there remains a gap in the literature on the changing trend and driving factors of ecosystem services value (ESV) in karst areas. In this study, Guizhou Province, a representative region of karst mountainous areas, was taken as a case to bridge the gap. ESV in the karst areas was predicted, based on the land use change data in 2009–2018, and the driving mechanisms were explored through the gray correlation analysis method. Results show that a total loss of CNY 21.47 billion ESV from 2009 to 2018 is due to the conversion of a total of 22.566% of the land in Guizhou, with forest land as the main cause of ESV change. By 2025 and 2030, the areas of garden land, water area, and construction land in Guizhou Province will continue to increase, whereas the areas of cultivated land, forest land, and garden land will decline. The total ESV shows a downward trend and will decrease to CNY 218.71 billion by 2030. Gray correlation analysis results illuminate that the total population and tertiary industry proportion are the uppermost, among all the driving factors that affect ESV change. The findings in this study have important implications for optimizing and adjusting the land use structure ecological protection and will enrich the literature on ESV in ecologically fragile areas.


Introduction
Ecosystem services (ESs) are increasingly the focus of international attention in the 21st century [1], which represents the benefits that life on earth obtains, directly or indirectly, from nature to maintain its ecosystem functions [2,3]. Nevertheless, human activities have exerted a dramatic impact on the ecosystem, resulting in the degradation of approximately 60% of the Earth's ESs [4,5]. The karst landform, featured in poor ecological stability and sensitive to ecological disturbance [6], are widely distributed globally [7]. Changes of the ecosystem functions, especially the extensive land degradation and vegetation deterioration in karst areas, have caused severe soil erosion and drought [8]. Studies have shown that there is a significantly positive correlation between incidence of poverty and ESs in karst areas [9]. As one of the most fragile ecosystems in the world [10], karst areas in China account for approximately 15% of the total land areas, making China one of the countries with the largest karst landforms in the world [11]. The ecological foundation of karst areas in China is so fragile that it can easily be damaged. Tremendously degraded plant Currently, research concerning the driving factors of ESV changes are plentiful [37][38][39][40][41], while direct drivers, such as land use change, and the application of new technologies attracts more attention [42,43]. For example, Liu et al. [44] superimposed the impact of future climate scenarios and land use change on ESs to explore the overall effect of the changes on ESs. During the period of rapid economic development, population growth, economic development, and urban expansion have exerted unprecedented pressures and negative effects on the regional ESs [39], and the impact of national policies on the ESs should not be neglected, such as the Grain for Green Program [17,45,46]. However, studies on whether socioeconomic characteristics are related to ESs prefer conceptual model and scenario analysis [26]. Villamagna et al. [47] insisted that socioeconomic factors are the driver of the ESs in agricultural landscape. Minin et al. [48] explored the trade-offs of ESs, land use, economic development, and biodiversity to identify priority areas through four policy scenarios. Song et al. [41] illuminated that socioeconomic development effect exerted the greatest impact on wetland ESV in Northeast China. A preliminary qualitative conclusion can be drawn from the above research, that socioeconomic factors do impose effect on ESs; yet, they cannot distinguish or sort the specific types of socioeconomic factors, which, as a result, cannot facilitate targeted policy application.
Guizhou is one of the provinces with the most typical karst landform development in China, exhibits extensive karst landform features and obvious regional differentiation [10]. In the past decade, Guizhou has experienced rapid economic development and urbanization, which have profoundly affected the karst ecosystem. Conducting research on kasrt areas is a new challenging attempt, which can enrich the types of studies on ESV in ecologically fragile areas. Taking Guizhou Province as a study case, this study analyzes the spatio-temporal change of land use and ESV during 2009-2018 and its driving forces in Guizhou Province. Based on the gray GM(1,1) model, the study made gray prediction on the future land use structure and the development trend of ESV. Additionally, the driving factors behind the change of ESV were explored by adopting gray correlation analysis method. The application of gray system to forecast the development trend and reveal the driving forces behind is an innovation. The findings of this research can aid in targeted policy application towards achieving beneficial environmental outcomes. The research objectives are as follows:

Study Area
Guizhou Province (103 • 36'~109 • 36' E, 24 • 37'~29 • 14' N) is located in the east slope of Yunnan-Guizhou Plateau, with the terrain high in the west and low in the east (Figure 1). Guizhou Province covers an area of 17,609,858 hm 2 , accounting for 1.83% of the total land area of China. The landform types are complex and diverse, mainly plateau and mountain, exhibiting typical karst geomorphological characteristics. The forest coverage rate is reaching 60%, and the land reclamation rate is as high as 20.03%, which is approximately twice the national average level. The high-intensity reclamation of territorial space has caused large areas of rocky desertification and soil erosion. Under the background of rapid urbanization, industrialization, and significantly unbalanced regional development, the land use in Guizhou Province underwent profound transformation from 2009 to 2018. The area of construction land increased by 181,959 hm 2 , whereas the area of cultivated land decreased by 36,340 hm 2 , the forest land decreased by 84,117 hm 2 , and the grass land decreased by 59,077 hm 2 . Land use change disturbs the ecosystem severely, and Guizhou Province has extensive karst landform features, generating a combination of serious land degradation issues [49]. Therefore, conducting research on the impact of land use change on ESV and unearthing the driving factors behind the ESV change in Guizhou is of great significance, which can serve as scientific guidance for the ecological protection in karst areas, as well as to promote the coordinated development of ecological protection and social economy. decreased by 59,077 hm 2 . Land use change disturbs the ecosystem severely, and Guiz Province has extensive karst landform features, generating a combination of serious degradation issues [49]. Therefore, conducting research on the impact of land use ch on ESV and unearthing the driving factors behind the ESV change in Guizhou is of g significance, which can serve as scientific guidance for the ecological protection in areas, as well as to promote the coordinated development of ecological protection social economy.

Data Sources
This study used ten-year data from 2009 to 2018. In details, the land use data predominantly from the land use change survey data of Guizhou Province, and the oeconomic data are from China Statistical Yearbook and Guizhou Statistical Yearb DEM is a 90 m ASTER GDEM data derived from the Geospatial Data C (http://www.gscloud.cn/, accessed on 15 November 2021), which can accurately reflec topographic features of the study area. According to China's land use classification tem, the original land use classes of the data consist of seven first-class land use ty including cultivated land, garden land, forest land, grass land, construction land, w area, and unused land.

Evaluation Model of Ecosystem Services Value
Costanza et al. [50] clarified the principles and methods of ESV estimation in a s tific sense. In China, Xie et al. [2,51], based on Costanza's research and China's chara istics, grouped China's ecosystem and re-determined the ESV coefficient, which bec the basis for Chinese scholars to assess ESV [10,52,53]. Based on the research resul Costanza et al. [50] and Xie et al. [2,51], the literature on the study of ESV in diffe regions were analyzed, the ecosystem types and ecological value coefficients (VC), c sponding to the land use categories in the Guizhou Province, were determined by bining the geographical and climate situation of the study area ( Table 1). The calcula of ESV is as follows:

Data Sources
This study used ten-year data from 2009 to 2018. In details, the land use data are predominantly from the land use change survey data of Guizhou Province, and the socioeconomic data are from China Statistical Yearbook and Guizhou Statistical Yearbook. DEM is a 90 m ASTER GDEM data derived from the Geospatial Data Cloud (http://www.gscloud.cn/, accessed on 15 November 2021), which can accurately reflect the topographic features of the study area. According to China's land use classification system, the original land use classes of the data consist of seven first-class land use types, including cultivated land, garden land, forest land, grass land, construction land, water area, and unused land.

Evaluation Model of Ecosystem Services Value
Costanza et al. [50] clarified the principles and methods of ESV estimation in a scientific sense. In China, Xie et al. [2,51], based on Costanza's research and China's characteristics, grouped China's ecosystem and re-determined the ESV coefficient, which became the basis for Chinese scholars to assess ESV [10,52,53]. Based on the research results of Costanza et al. [50] and Xie et al. [2,51], the literature on the study of ESV in different regions were analyzed, the ecosystem types and ecological value coefficients (VC), corresponding to the land use categories in the Guizhou Province, were determined by combining the geographical and climate situation of the study area ( Table 1). The calculation of ESV is as follows: where ESV refers to the total value of ESs in the study area (CNY); A i refers to the distribution areas (hm 2 ) of the i type of land use in the study area; VC i refers to the ecological VC per unit area of type i land use category (CNY /hm 2 yr); i refers to the type of land use in the study area.

Sensitivity Index
Coefficient of sensitivity (CS) or coefficient of elasticity is usually used to determine the sensitivity and robustness of the price (coefficients) [54]. To verify the representative of ecosystem types to each landscape type, as well as the accuracy of VC, CS was introduced to determine the degree of dependence of the total ESV on the change of VC over time, namely changes in ESV attribute to a 1% change in VC, which can be expressed in Equation (2) [55]: where i and j represent the initial and adjusted value, respectively; k represents the land use category. The CS is applied to verify whether the selected VC is suitable for this study area. According to Equation (2), CS is always less than 1 [20]. The higher the ratio, the more critical the accuracy of the VC, and the greater the sensitivity of the ESV to it. In this study, the VC of each landscape type was adjusted up and down by 50%, respectively, to illustrate the sensitivity of ESV to VC.

Gray Forecast Model
Gray prediction model [10,31] is based on the understanding of the uncertain characteristics of the system evolution, using sequence operators to generate and process the original data, dig out the law of system evolution, establish the gray model, and make scientific quantitative predictions for the future state of the system. In this study, the gray GM(1,1) model is adopted to forecast the land use structure and the development trend of ESV in Guizhou Province in 2025 and 2030. If t (i) is defined as the original data sequence and x (1) t (i) is a sequence of number generated by an accumulation, then the standard equation of the first-order, linear constant coefficient differential equation of GM(1,1) model is as follows: 6 of 20 The standard solution, corresponding to the gray GM(1,1) model, is as follows: where a and u are unknown parameters to be determined; t is time.
To improve the accuracy and reliability of gray GM(1,1) model prediction, a posterior difference test method was adopted to test the model accuracy. The posterior error ratio C and the small error frequency P are defined as follows: where S 1 is the standard deviation of the original data; S 2 is the mean of standard deviation of the predicted data; ε k is the error of forecast data; and ε is the prediction error. The value of C indicates the degree of dispersion of the difference between the predicted value and the actual value (the lower, the better).

Gray Correlation Analysis
Gray correlation analysis is a method to measure the correlation degree between factors according to the similarity or difference degree of the development trend among factors. Its principle is based on the macro or micro geometric approximation between behavior factors [56]. When the experiment is not explicit, or the experimental method cannot be carried out accurately, gray analysis can help to make up for the deficiencies in statistical regression [57] and can be used to measure the approximate correlation between the series [58]. The more similar the curves, the greater the correlation between relative data series [56]. In other words, if the changing trends of two factors (or indicators) are basically consistent or synchronized, it can be regarded as the correlation between the two is relatively strong, and vice versa. In this study, seven factors were selected from three aspects of population, economy, and policy to analyze the driving factors of ESV in Guizhou Province, which were the total population (TP), urbanization level (UL), gross domestic product (GDP), secondary industries proportion (SIP), tertiary industries proportion (TIP), total investment in fixed assets (TIFA), and afforestation project (AP) ( Table 2). After the original data are normalized by the initial value or average value, the gray correlation coefficient can be obtained by the following equation: where ξ ij (k) is called the gray correlation coefficient at time t = k for x 0 and x i , min | are the minimum difference and maximum difference from the normalized reference series to all the other normalized series, respectively, and ρ is called the distinguished coefficient which can be adjusted to better distinguish between the normalized reference series and normalized comparative series, ρ is taken between 0 and 1, usually 0.5. Then, Equation (7) is introduced to calculate the gray correlation grades, which is actually the average of the gray correlation coefficients.
where r ij is the gray correlation grade. n refers to the number of factors. The closer to 1 the existing gray correlation grade is, the greater the impact of the comparison sequence on the reference series, that is, the greater the correlation impact on the ESV.

Land Use Change in Guizhou Province
As shown in Table 3, whether in 2009 or 2018, forest land accounted for more than 50% of the total land areas, constituting the main body of the land use structure. Due to the large areas of forest land, the ecological function in Guizhou is of great significance and it is sensitive in the construction of ecological environment. From 2009 to 2018, the total area of land use change in Guizhou Province was 374,900 hm 2 . Among them construction land and forest land are the main types of land use change, with a total change area of 266,100 hm 2 (construction land increased by 182,000 hm 2 , forest land decreased by 84,100 hm 2 ), accounting for 70.98% of the total land use change. Among the land use categories, the area of forest land, grass land, cultivated land, and unused land all decreased to varying degrees (Table 3), with a reduction of 84,100 hm 2 , 59,100 hm 2 , 36,300 hm 2 , and 8400 hm 2 , respectively. The reduction rates were 0.93%, 3.6%, 0.8%, and 1.7%, respectively. Construction land, garden land and water area increased by 182,000 hm 2 , 4700 hm 2 , and 1300 hm 2 , respectively, with an increase rate of 12.0%, 3.0%, and 0.5%. The gray forecast analysis shows that by 2025 and 2030, cultivated land, forest land, grass land, and unused land exhibit a downward trend, whereas garden land, construction land, and water areas will be reversely varied ( Figure 2).   The main driving factors for the decrease of cultivated land and the increase of garden land are the implementation of the Grain for Green Program [17,45,46] and the adjustment of the agricultural industry structure, which has turned part of the cultivated land into orchards, tea gardens, and other land use types. The decrease in forest land, grass land and unused land, and the increase in construction land is predominantly due to the development of social economy [10]. Additionally, Guizhou Province has implemented a large number of resettlements, infrastructure, poverty alleviation industries, and people's livelihood security projects, and the demand for construction land is far beyond expectations. The reason for the increase of water area is that the three-year battle of water transport construction launched by Guizhou Province promoted the construction of water conservancy and hydropower projects.

Change of Total Ecosystem Services Value in Guizhou Province
The total ESV in Guizhou Province from 2009 to 2018 is calculated by Equation (1) ( Table 4). Results show that the total ESV in Guizhou Province is CNY 223.64 billion in 2009 and CNY 221.50 billion in 2018, a decrease of CNY 21.47 billion in 10 years, with a change rate of 9.60%. The VC of forest land is relatively high, leading to the continuous decrease of total ESV in the Guizhou Province, due to the reduction of forest land in the past decade. In ten years, the area of forest land decreased by 84,117 hm 2 , resulting in a reduction of CNY 16.26 billion in ESV, accounting for 75.73% of the total ESV decrease, indicating that the change of forest land is the main reason for the change of ESV. In terms of the total ESV of each land use type, the ESV of cultivated land, forest land, grass land, and unused land exhibit a downward trend, whereas garden land and water area display a floating change ( Figure 3). The main reason for the decline in ESV in Guizhou is predominantly, due to the development of social economy. Though the implementation of the Grain for Green Program did play an active role [10,59,60], leading to the rapid growth of garden land, but it only worked in the early stage. The three-year battle of water transport construction in Guizhou exerted great impact on the change of water area, but then the effect weakened. The decrease of ESV caused by construction occupation, due to economic development is difficult to compensate. Besides, the total ESV of forest land and cultivated land reach up to 89%. Therefore, Guizhou Province should attach importance to the protection of forest land and cultivated land ecosystem in the future land use, continue to strengthen the ecological environment construction through forest land protection and comprehensive land management.

Sensitivity Analysis
The CS can determine the dependence of ESV over time on ecological VC, so as to verify whether the selected ecological VC is suitable for this study area. By adjusting the VC by ±50%, the ESV and CS of each category in Guizhou Province in 2009 and 2018 were calculated ( Table 5). The horizontal comparison shows that the ecosystem CS is in the order of forest land > cultivated land > grass land > water area > garden land > unused land, and the highest value is 0.779. Namely, when the VC of forest land increases by 1%, the corresponding ESV increases by 0.779%, exerting the greatest impact on the ESV of the study area. The lowest value is 0.001, that is, when the VC of the garden or unused land increases by 1%, the corresponding ESV increases by 0-0.001%, which has little impact on

Sensitivity Analysis
The CS can determine the dependence of ESV over time on ecological VC, so as to verify whether the selected ecological VC is suitable for this study area. By adjusting the VC by ±50%, the ESV and CS of each category in Guizhou Province in 2009 and 2018 were calculated ( Table 5). The horizontal comparison shows that the ecosystem CS is in the order of forest land > cultivated land > grass land > water area > garden land > unused land, and the highest value is 0.779. Namely, when the VC of forest land increases by 1%, the corresponding ESV increases by 0.779%, exerting the greatest impact on the ESV of the study area. The lowest value is 0.001, that is, when the VC of the garden or unused land increases by 1%, the corresponding ESV increases by 0-0.001%, which has little impact on the ESV of the study area. From the perspective of different periods, there is no significant change in the sensitivity index of all land types from 2009 to 2018. The greatest change is forest land, but it only increases by 0.007%. In addition, the 10-year change rate of the total ESV in the study area varies from 0.897% to 1.563% after the adjustment of ecological VC, and the difference is small. All these indicate that ESV in the study area is inelastic to VC, namely the VC adopted in the study area does not affect the authenticity of ESV changes over time, which is in line with the actual situation in the study area, and the research results are credible. Since the spatial distribution pattern of the sensitivity index of ESV to VC of different land use types were similar in different years, we only listed the maps in 2018 ( Figure 5). The spatial distribution showed that the sensitivity index of cultivated land in the west was higher than that in the east (Figure 5a), whereas the sensitivity index of forest land shows the opposite (Figure 5c). Besides, the sensitivity index of garden land, grass land, water area, and unused land show higher distribution characteristics in the middle and southwest Guizhou Province (Figure 5b,d-f).

Gray Forecast of Ecosystem Services Value
Based on the land use data of Guizhou Province from 2009 to 2018, Matlab_R2021a was applied to construct a gray GM(1,1) model suitable for all land use types in the study area to predict the land use structure in 2025 and 2030. The forecast assumes that the total area of the study area remains unchanged, and the predicted value of unused land is the total area of Guizhou Province minus the predicted value of other land use types and the ESV of different land use types is calculated (Table 6). Results show that by 2025 and 2030, the ESV of garden land and water area will increase slightly, whereas others will decrease to varying degrees, with forest land the greatest ( Figure 6). The total ESV will decrease to CNY 219.83 billion and CNY 218.71 billion in 2025 and 2030, respectively. According to the model prediction accuracy rating standard, when C < 0.35 and p ≥ 0.95, the model accuracy is level 1 (good). When C ≥ 0.65 and p ≤ 0.75, the model accuracy is level 4 (unqualified). According to Table 6, the posterior error ratio of this model, Ci ≤ 0.28 < 0.35, and Pi = 1 > 0.95, shows that the prediction accuracy of the model is good and the prediction value is highly reliable.

Gray Forecast of Ecosystem Services Value
Based on the land use data of Guizhou Province from 2009 to 2018, Matlab_R2021a was applied to construct a gray GM(1,1) model suitable for all land use types in the study area to predict the land use structure in 2025 and 2030. The forecast assumes that the total area of the study area remains unchanged, and the predicted value of unused land is the total area of Guizhou Province minus the predicted value of other land use types and the ESV of different land use types is calculated (Table 6). Results show that by 2025 and 2030, the ESV of garden land and water area will increase slightly, whereas others will decrease to varying degrees, with forest land the greatest (Figure 6). The total ESV will decrease to CNY 219.83 billion and CNY 218.71 billion in 2025 and 2030, respectively. According to the model prediction accuracy rating standard, when C < 0.35 and p ≥ 0.95, the model

Driving Forces of Change in Ecosystem Services Value in Guizhou Province
Gray correlation grades of influential factors on ESV change are listed in Table 7. It can be seen that all the gray correlation grades are greater than 0.5 and maintain high value, indicating that the seven parameters are the driving factors affecting the ESV change in Guizhou Province. From 2009 to 2018, the relative importance of influencing factors to ESV change is as followed: TP > TIP > SIP > AP > UL > GDP > TIFA. Among them, TP exerts the greatest impact on ESV, with a gray correlation grade of 0.9965, which is close to 1. The smallest gray correlation grade belongs to TIFA, which equals to 0.6620. Except from 2009 to 2012 TIP accounts for the highest gray correlation degree, TP ranks the highest of gray correlation degree in other years, which has been maintained above 0.97. Besides, the gray correlation grade of TIP has also remained above 0.93, revealing that TP and TIP were the two main driving factors of ESV change in Guizhou Province.

Comparing with Previous Studies
Judging from the overall change trend of ESV in Guizhou Province, the ecological protection trend in karst areas is not optimistic. Studies have shown that the process of urbanization in the past is often at the cost of the ecological environment, resulting in a significant loss of ESV [26]. In the early stage (2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008), the implementation of Grain For Green Program did play an active role in ESs in karst areas [10,59,60], yet, according to our study, in the past decade, though the Grain For Green Program continued, the land demand resulted from urbanization and population boom far exceeded expectations, leading to the decline of ESV in karst areas. From 2009-2018, construction land in Guizhou Province has increased by 181,959 hm 2 , with a change rate of 11.96%. As a result, ESV in Guizhou Province has decreased by CNY 2.15 billion, with an average annual decrease of 3.6%. Undoubtedly, urbanization has exerted a negative impact on the ecosystem supply capacity, which is in line with the research of Chen et al. [61].
In this study, ESV was evaluated in combination with the equivalent value table proposed by Xie et al. [2,51] and Costanza et al. [50]. Results showed that the contribution of different land use types to the total ESV was significantly different, among which the forest land contributes the largest, followed by cultivated land and grass land, which is consistent with the research of Leh et al. [62] and W. Chen et al. [63]. Some studies believe that the ESV of construction land is negative [64], whereas some studies consider that construction land can provide certain ESs (e.g., entertainment, tourism, culture, and carbon storage) [22,65]. This research is in conformity to the research of Wu et al. [26], which suggested that construction land contributes nothing to ESV but the loss of it; hence, we assign the coefficient of construction land to be 0.
Previous studies mostly used coarse data sets that lacked correction, leading to differences in ESV assessment in the same place [66,67]. Meanwhile, the inconsistency in the land use classification can also lead to differences in ESV assessment. For example, when Wang (2018) [68] measured the ESV of Guizhou from 2016 to 2017, the land use types were divided into cultivated land, forest land, water area, grass land, construction land, and unused land; the calculated ESV for 2016 and 2017 were CNY 207.90 and CNY 207.70 billion, respectively, both less than the ESV calculated in this study, which are CNY 221.87 and CNY 221.66 billion, respectively. Zhao et al. [69] classified land use types into five categories: cropland, forest land, grass land, water bodies, and unused land. According to their calculation results, the total amount of ESV in 2013 and 2017 were both higher than that in this study, CNY 261.80 and CNY 261.20 billion, respectively. In the evolution of China's land use classification system, garden land is an important secondary category. However, when ESV is measured, garden land is rarely considered, but mostly counted into forest land [70], which is also a factor attracts the difference in ESV assessment. This study believes that garden land differs from forest land, the VC of which should be less than that of forest land. Therefore, it is listed separately for evaluation.

Evolution Mechanism of Ecosystem Services Value in Guizhou Province
Our research results are in accordance with previous studies that land use change is one of the main drivers of ESV loss [71], and the most significant change in ESs attributed to land use change is predominantly urbanization driven [7]. However, the research results also show that neither socioeconomic factors nor physical factors are the sole or primary potential causes of ESs change [27], and the main driving forces are also variable in different periods [72]. In addition to land use structure, economic, social, demographic, and policy factors are also driving factors affect ESV. The results of this study show that from 2009 to 2018, TP imposes the greatest impact on ESV, followed by TIP in economic factors. Leading driving forces vary in different periods. The most important factor affecting the regional variation of ESV is TIP, between 2009 and 2012, while TP between 2012 and 2018. The urbanization process is accompanied by the population surge, and the rise of the tertiary industry has become an important symbol of the acceleration of urbanization. The Guizhou Province is located in the center of the east Asian karst region, with the most complex, complete types, and largest distribution area in the world [70]. Thus, the analysis of the driving factors affecting the ESV change in Guizhou will provide decision-making reference for the study of ESV change and ecological governance in karst areas and will enrich the literature on ESV in ecologically fragile areas.

Policy Implications
In the early stage of economic development, ecological considerations have commonly been interpreted as a factor restricting economic development and can, thus, be ignored easily [12]. Existing evidence indicates that economic development yields obvious ecosystem degradation through land use change [27]. Rapid urbanization in Guizhou has driven the expansion of human-dependent lands (e.g., housing and transportation infrastructure), while ecological lands (e.g., forest land, grass land, and garden land) were heavily encroached. The ratio of GDP to ESV in China is, approximately, 1:1 [73], whereas the ratio of GDP to ESV in Guizhou from 2009 to 2018 increased from 1.75:1 to 6.93:1, indicating that the ESV in Guizhou is far below the average level in China. Policy can play a vital role in maintaining regional ESV [74,75]. A considerable number of studies have focused on ESs, and how to incorporate ESs into land use and ecological conservation decision-making has long been discussed [76]. However, the non-uniqueness and effectiveness of assessment methods along with the difficulty of incorporating assessment results into actual management have led to the policy implications of ESs [77]. Additionally, the trade-offs and synergies in ESs [78], are also the key to land use decision-making [42]. For example, land development to guarantee food security will inevitably lead to the weakening of water and soil conservation and the hydrological regulation function.
As a typical representative of China's karst landform, Guizhou is positioned as a province with balanced grain production, marketing, energy, and mineral base. Meanwhile, it is also as an important ecological security barrier in the upper reaches of the Yangtze river and the Pearl River. Thus, its ecological restoration should be placed in a prominent position; under the premise of protection, the economic development must meet the carrying capacity of resources and environment. Within this context, Guizhou Province must coordinate ecological protection and economic development. In accordance with the principle of "ecological priority and green development" [61], the adjustment of economic structure should be accelerated, industrial transformation and upgrading must be promoted, and rationalization of the spatial layout of industrial functions must be guided, so that to promote green and high-quality development. Meanwhile, it would also be an innovation to incorporate new concepts, such as green GDP, into the performance evaluation system for government officials [3,79].

Limitations and Future Directions
Certain limitations could be observed in this study. The VC adopted in this study is based on the research of Xie et al. [2,51] and Costanza et al. [50]. Actually, the VC of the same ESs in different regions have certain deviations. However, this study does not consider applying different indicators to estimate the same ESV in different regions. Besides, obtaining data in the karst areas is not easy, which can be a direction of future research. Although the assessment of ESs has been extensively studied, considerable progress still needs to be made to improve the evaluation of ESs [80]. In predicting future ESV changes, the temporal heterogeneity of VC is also easy to be ignored. Therefore, how to accurately assess and quantify ESs in a complex and constantly changing ecosystem will also be a question worth exploring. Furthermore, the research of Aschonitis et al. [54] showed that the ecosystem sensitivity index can only be applied to rank the importance of land use based on its contribution to the total ESV and should be abandoned for assessing the robustness and sensitivity of ESs. Future studies can be discussed in this regard. In terms of land use simulation and prediction, the gray prediction model has the advantage of being able to make predictions, in the case of a small number of data samples; however, the impact of factors, such as policies, in the long-term series will make for the prediction results to be inconsistent with the actual development situation, which needs to be noted when applied for forecasting in the future. In the next step, our study will further improve the method model and analyze the value differences of ecosystem types in different times within the same region, and the spatial distribution value differences of ecosystem types will be further explored.

Conclusions
The gray GM(1,1) model is introduced in this study to reveal the spatial-temporal changes of land use and ESV. The contribution of different land use types to ESV, in virtue of the sensitivity index, is measured. The gray correlation analysis method is adopted to explore the driving mechanism affect ESV changes in Guizhou Province. The areas of forest land, grass land, cultivated land, and unused land in Guizhou Province decreased to varying degrees, whereas the construction land, garden land, and water areas exhibited an upward trend from 2009 to 2018. Among them, construction land and forest land are the main types of land use change. The ESV in Guizhou Province decreased by CNY 21.47 billion, with forest land the main cause of ESV change. The sensitivity indexes of ESV to VC were all less than l, indicating that the evaluation results of ESV in Guizhou Province were credible. From the perspective of different periods, there was no significant change in the sensitivity index of all land use types from 2009 to 2018. The forecast results showed that by 2025 and 2030, the areas of garden land, water area, and construction land in Guizhou Province will continue to increase, whereas the areas of cultivated, forest and garden land will be on a decline. Nevertheless, the total ESV will decrease to CNY 218.71 billion by 2030. Gray correlation analysis results showed that population, economy, and policy factors can all affect ESV change, among which TP and TIP are the two most important driving factors. Consequently, top priority should be given to the ecological factors in future planning and socio-economic development. We expect that the research result and proposed policy suggestions of this study will constitute an important aid to the correct treatment of the spatial characteristics of land uses change and ESV in karst areas.