Study on the driven mechanism of hydrologic drought based on the lithology-combined structure of the Karst drainage basin in South China


 Hydrologic drought, considered as a typical natural phenomenon in the background of global climate changes, is the continuation and development of meteorological and agricultural droughts, and is the ultimate and most thoroughly drought. The research area controlled by the 55 hydrological sections in South China is selected in this paper, and the intensity and frequency of hydrologic droughts are analyzed by the Standardized Runoff Index (SRI), and the driven mechanism of watershed lithologies to hydrologic droughts is discussed. The results show that (i) the hydrological drought of Karst drainage basins is shown the gradual aggravation from the west to east parts in South China, with the significant north–south stripe distributions at the SRI_3 and SRI_6; (ii) the occurring probability of hydrological droughts is the Limestone-type Karst Basin (II and III, 0.17) < Dolomite-type Karst Basin (I and IV, 0.22) < Non-Karst Basin (V, 0.25) in terms of combination types of basin lithologies, and (iii) the Karst Basin (I and III, 0.18) < Semi-Karst Basin (II and IV, 0.2) < Non-Karst Basin (V, 0.25) in terms of basin lithologies. Therefore, this proves that the most water-stored spaces are found in Karst Basins under the differential dissolution or erosion effects of soluble water, followed by in the Semi-Karst Basin, the least water-stored spaces in the Non-Karst Basin.


INTRODUCTION
Hydrologic drought, considered as a typical natural phenomenon in the background of global climate changes, is the continuation and development of meteorological and agricultural droughts, and is the ultimate and most thoroughly drought (Kang & Guo 1991). The drought cannot be simply attributed to the climate anomalies and precipitation reduction under the special and fragile Karst ecological environments, but to the water storage capacity of underlying surface media shown as the lithology, landform, soil, vegetation, etc. The watershed lithology is one of important medium factors, and their functions are mainly shown that (i) for the different types of watershed lithologies, there are different kinds of watershed landforms developed in Karst basins, which will lead to the different runoff characteristics on the surface and underground, and affect the watershed storage capacity (Liang 1995(Liang , 1997Liang & Wang 1998); (ii) the different soil structures and vegetation types are developed due to different types of watershed lithologies with different anti-weathering abilities and weathering crust thicknesses formed in Karst basins, and they influence the infiltration and runoff of rainfall on the surface and underground, which results in the different watershed storage capacity; and (iii) for the different types of watershed lithologies, with the different mechanical properties and the different development intensities and scales of bedrock fissures, are developed the different water-stored spaces or drainage channels in Karst basins, which affects the watershed storage capacity. For example, Liang (1995) discussed the flood peak effects caused by the coupling differences between the different spatio-temporal scales of Karst drainage basins and the different spatial configurations of landform types. It found that there is a positive return period of hydrologic droughts in southeastern semi-arid region of Iran. It was found that the vegetation cover play important roles in the hydrologic drought length (Modarres & Sarhadi 2010). Drought duration, severity, and frequency are analyzed for steep topography in the central Vietnam and other parts of Southeast Asia at different timescales (Firoz et al. 2018). The more severe droughts are observed in groundwater dependent areas, where have longer durations rather than higher intensities (Rudd et al. 2017). The relationship between the El Niño-Southern Oscillation (ENSO) and hydrologic variability in the United States is discussed using Empirical Orthogonal Function (EOF)/Principal Component Analysis (PCA), and it is found the positive relationships in the Pacific Northwest, whereas negative relationships in southern California and the northern Great Plains (Ryu et al. 2010). Finally, to explore the feasibility of the SRI in Southwest China, Liu et al. (2016) simulated a long-term daily hydrological and meteorological data series by the Xinanjiang model and calculated the SRI. It proves that the onset, severity, and duration of extremely droughts are detected well by the SRI. For the studies on hydrological droughts in China, Feng & Wang (1997) and Feng & Jia (1997) mainly discussed the influence factors of runoff volume in the dry season, and the identification method of hydrological droughts based on the theory of runs, and analyzed the hydrologic drought severities by the time fractal dimensions. Feng (1993Feng ( , 1994Feng ( , 1995 studied the probability density and distribution functions of extreme hydrologic drought durations. The joint distribution of hydrologic drought characteristics is constructed employing the Copula Joint Distribution Function (Yan et al. 2007;Ma & Song 2010;Xu et al. 2010;Zhou et al. 2011). However, most of the researches are mainly to make the identification, characteristic analysis, and prediction for hydrologic droughts by different drought indices, respectively. For instance, Zhai et al. (2015) established a new hydrologic drought assessment index such as Standard Water Resources Index (SWRI) and developed a complete basic framework of hydrologic drought identification, assessment, and characteristic analysis combined with the distributed hydrologic model, the copula function, and the statistical test method. The Standardized Streamflow Drought Index (SSDI) is established based on the optimal distributions such as the logistic, normal, two-parameter lognormal, or Weibull probability distribution , and it is validated the applicability and rationality in the Fenhe River Basin. Combined with the percentages of runoff anomaly and precipitation anomaly, Wu et al. (2016) constructed the Regional Hydrological Drought Index (RHDI) and obtained the corresponding frequency of drought grade, and further determined the threshold value of the different drought levels for regional hydrological droughts. The Copula Model of two-variable joint distribution of hydrologic drought characteristics is built based on the test method of Cramer-von Mises Statistics associated with Rosenblatt transfer (Tu et al. 2016), and it is analyzed the hydrologic drought characteristics and water shortage responses in the Dongjiang River Basin under a changing environment. Ren et al. (2016) quantitatively separated the contribution of climate change and human activities on runoff reduction based on the Variable Infiltration Capacity Model (VICM) and analyzed the spatio-temporal evolution characteristics of hydrological droughts by the SRI. The Standard Precipitation Evapotranspiration Index (SPEI) and the SDI are used to analyze the evaluation characteristics of both meteorological and hydrologic droughts and to discuss the response of hydrologic droughts to meteorological droughts (Li et al. 2016). The SRI, SDI, and associated indicators with the trend, time lag cross-correlation are applied to analyze the spatial-temporal characteristics of meteorological and hydrologic droughts across the Yellow River Basin (YRB) during the periods 1961-2010 (He et al. 2015a). According to the correlation characteristics of seasonal runoff, the copula prediction model of hydrologic droughts based on the copula function, runoff distribution function, and SRI is constructed by Zhang et al. (2016), and the empirical analysis is carried out by the hydrologic station of the Aksu River West Bride.
However, the present studies on the hydrologic droughts in Karst basins, except for some relevant research contents of this team He et al. , 2014He et al. , 2015bHe et al. , 2018aHe et al. , 2018b, have not seen more detailed studies reporting. Thus, the objectives of this study are (i) to discuss the types of Karst drainage basins based on the spatial coupling structures of different lithologies in Section 4.1; (ii) to comparatively analyze the hydrological drought characteristics in terms of different time scales and spatial coupling of lithologies in Section 4.2; and (iii) to systematically study the driven mechanism of the single lithology and the different lithologies coupled to hydrological droughts in Section 4.3. Therefore, it is beneficial to promote the development of Karst hydrographic geomorphology.

STUDY AREAS
The South China (SC), taking Guizhou, Yunnan, and Guangxi as the center, is a typical distribution region of the Cone Karst, Sword Karst, and Tower Karst. This study was to select the Karst drainage basins controlled by 55 hydrological stations in the SC as the research area. It is enclosed by the eastern longitudes of 101°55 0 55″-110°55 0 45″ and northern latitudes of 22°J ournal of Water and Climate Change Vol 00 No 0, 3 Uncorrected Proof 42 0 57″-29°13 0 11″ with an area of 352,526 km 2 and an average elevation of 1,065.62 m. The research area is mainly included the most area of Guizhou Province (37.97%), southeastern area of Yunnan Province (25.36%), and northwestern and northern areas of Guangxi Province (36.67%) (Figure 1). It is located in the climatic zones of north subtropical humid climate and south subtropical semi-humid climate with the mean annual rainfall of 1,000-1,300 mm, and the average annual temperature of 16-23°C. The study area is divided into two parts of the Yangtze and Pearl River Basins by the Wumeng-Miaoling Mountains. That means the Jinshajiang River System, the Upper Yangtze River Mainstream System, the Wujiang River System, and the Dongting Lake System of the Yangtze River Basin are located in the north of research area, whereas the Duliu River System (in Guizhou), the Yuanjiang River System (in Yunnan), and the Xijiang River System (in Guangxi) of the Pearl River Basin in the south of research area. ) and calculated the monthly average runoff volume during the periods 1970-2013. The hydrological stations selected in this paper must have 44 years of continuous hydrological observation, and continuous defect data must be less than 3 months. All the data will be made quality analysis (reliability, homogeneity, and representativeness), and for the months of defect data are interpolated by the cubic spline function. The interpolations must keep the same change trend as the total values during the periods 1970-2013. In order to eliminate the influence of different basin areas, hydrologic data were standardized in this study.

Lithologic data
Firstly, we all know that the geological lithological types and structures are basically unchanged during the periods 1970-2013, because the formation and evolution of geologic lithologies are a slow geological process. This paper made some preprocessing for TM images corresponding to the mean runoff of the minimum month in 2006 (Time: January to December 2006) and extracted the study area of remote sensing images controlled by 55 hydrological stations (He et al. 2012). Secondly, the research area was extracted from the Comprehensive Geological Map of 1:500,000 in Guizhou, Yunnan, and Guangxi Provinces carried out by the geometric correction and projection correction, which was fused with the TM images. Thirdly, the remote sensing information of watershed lithologies was automatically extracted using object-oriented classification technology based on the Comprehensive Geological Map of 1:500,000 in Guizhou, Yunnan, and Guangxi Provinces (Jiao & Liang 2002). Finally, to make the statistics for the lithologic types of areas and to calculate the area percentages (Table 1).

Identification of hydrological drought
Hydrologic drought is a phenomenon when the river flow is lower than its normal value. In other words, the river flow cannot satisfy the water supply demand in a certain period (Van Loon & Van Lanen 2012;Van Loon & Laaha 2015). According to the definition of hydrological droughts and Karst basin characteristics (Tsakiris et al. 2007;Liu et al. 2012;López-Moreno et al. 2013;Jehanzaib et al. 2020;Zhou et al. 2020), the SRI was quoted to describe the hydrological drought characteristics in this paper. The SRI can be expressed as follows: , for 0:5 , H i,k (x) 1:0 in which q is the probability of zero precipitation and F i,k (x) is the cumulative probability of the gamma distribution.
in which g i,k and b i,k are the shape and scale parameters, respectively, and G(g i,k ) is the gamma function.
where Q i,j refers to the accumulated runoff volume of the jth month in the ith hydrologic year, and X i,j denotes the accumulated runoff volume of the kth reference period in the ith hydrologic year. k ¼ 1, 2, 3, 4 means October to December, October to March, October to June, and October to September, respectively (Nalbantis & Tsakiris 2009;Tigkas et al. 2012;Tabari et al. 2013). A positive SRI means humid, whereas a negative SRI means drought. Moreover, when SRI is negative, the larger the absolute value of the SRIs means the serious hydrological drought. According to SRI, hydrological droughts can be divided into five levels, namely the Non-Drought for the 1.0 SRI 3.0, the Mild Drought for the À0.99 SRI 0.99, the Moderate Drought for the À1.49 SRI À1.0, the Severe Drought for the À1.99 SRI À1.5, and the Extreme Drought for the À3.0 SRI À2.0 (Table 2) (Nalbantis & Tsakiris 2009;Wen et al. 2011;Jehanzaib et al. 2020).

Analysis of hydrologic drought mechanism
According to the definition of the Bayesian formula (Liao et al. 2007), to assume the independent random events (n) B 1 , B 2 , . . . , B n are satisfied with the B i > B j ¼ F (i = j) (Φ represents impossible event), and P < n i¼1 ¼ 1, P(B i ) . 0 (i ¼ 1, 2, . . . , n). For any event A (P(A) . 0), there is where P(B k ) refers to the hypothesis probability before the test, and P(B k jA) refers to the probability after the test. This paper supposed that X ¼ (x ij ) mÂn is the information matrix of lithological types in the research areas, and Y ¼ (y i ) mÂ1 is the drought level matrix of hydrologic droughts. In which x ij is the area percentage of the jth type of lithologies, the ith level of hydrologic droughts, and y i is the area percentage of the ith level of hydrologic droughts.
To presume that B i represents the hydrologic drought event y i affected by the jth type of lithology x ij . Therefore, the uncertainty of hydrological droughts can be described by the conditional probability P(y i jx ij ). The conditional probability P(y i jx ij ) is expressed as In a maximum likelihood classification, the maximum value of {P(y i jx ij ), i ¼ 1, 2, . . . , 5} will be selected, if there is P(B i jx ij ) . P(B r jx ij ), for all i = r. Namely, the hydrological drought is belonged to the category B i . It is called the probability vector for all probability values P(B i jx ij ) of the impact of each lithologic type index on hydrologic drought levels, which will be used to describe the uncertainty of lithological type indicators.

Analysis of the watershed lithology-combined structure
As shown in Table 1, the watershed lithologies, without a single type, are mainly mixed with two or more than types of lithologies in Karst drainage basins. Therefore, to make the classification for the watershed sample areas by the system clustering method based on the watershed lithology-combined characteristics, and to draw the cluster pedigree by the SPSS software ( Figure 2).
It can be seen that there are certain similarities between different basins in the cluster pedigree chart. The watershed lithological types could be divided into five types of combining structures according to the marking distance equal to 20 (Wang et al. 2002).
Class I (c26, c42, c34, c5, c19, c46, c16, c37, c41, c33, c13, c21, c9, c29): The largest (65%) is the area percentage of the dolomites, mud dolomite with mudstone (T), followed by the medium thickness limestones with breccia dolomite and siliceous limestone (P) with the area percentage 19%, and the smallest (1%) is the area percentage of the limestones and dolomites in the upper part of the basins, dolomite in the lower part, and quartz sandstone in the bottom (ε). Therefore, the basin with the lithology-combined structures is called a Dolomite-type Karst Basin.
Class II (c6, c20, c3, c24, c2, c12, c45, c17, c15): The largest (66%) is the area percentage (66%) of medium thickness limestones with breccia dolomite and siliceous limestone (P), followed by the dolomite, mud dolomite with mudstone (T) with the area percentage 21%, and the smallest (3%) is the area percentage of limestones with mud limestone and shale (J), and carbon shale with sandstone (D). Therefore, the basin with the lithology-combined structures is called a Limestone-type Semi-Karst Basin.
Class IV (c1, c4, c23, c25, c31, c47, c10): The largest (26%) is the area percentage of limestones and dolomites in the upper part of the basin, dolomite in the lower part, and quartz sandstone in the bottom (ε), followed by the thick limestones with dolomite (O) with the area percentage 16%, and the smallest (2%) is the area percentage of dolomites with silicon strip dolomite in the upper part of the basin, and mud dolomite in the bottom (Z). Therefore, the basin with the lithology-combined structures is called a Dolomite-type Semi-Karst Basin.
Class V (c14, c36, c40, c49, c52, c50, c51, c7, c39, c22): The largest (27.85%) is the area percentage of carbon shales with sandstone (D), followed by the sericite slates with siltstone in the upper part of the basin, and siltstones with sericite slate in the bottom (Pt) with the area percentage 15%, and the smallest (2%) is the area percentage of thick limestones with dolomite (O). Therefore, the basin with the lithology-combined structures is called a Non-Karst Basin.

Analysis of hydrological drought characteristics at different time scales
Hydrological drought characteristics in South China at the different time scales (SRI_3, SRI_6, SRI_9, and SRI_12) (SC) were calculated based on the data of monthly mean runoff during the periods 1970-2013 by the SRI in this study (Figure 3).
In terms of the temporal evolution of hydrologic droughts, (i) the distribution areas of hydrological droughts in the Karst drainage basins of South China are larger, and drought intensities are shown a decreasing trend. Among them, the hydrological droughts at the SRI_3, SRI_6, and SRI_9 are relatively serious with the area percentages of 98.99, 99.63, and 98.46%, respectively, and relatively lighter at the SRI_12 with the area percentage of 77.43% (Figure 3(d)). It means that the hydrological drought areas in Karst basins are gradually decreasing with the increase of time scales. (ii) The distribution areas of moderate hydrological droughts at the SRI_3 and SRI_6 are the largest with the area percentages of 43.23 and 52.26% shown in Figure 3(a) and 3(b), respectively, followed by the mild hydrologic droughts with the area percentages of 27.74 and 20.12%, and the smallest the Non-Drought. It indicates that the runoff regulation functions of Karst drainage basins are weaker or nonsignificant with the time scales decreasing. While the distribution areas of mild hydrological droughts at the SRI_9 and SRI_12 are the largest (62.19 and 66.21%), followed by the moderate hydrological drought (27.32%) at the SRI_9, the Non-Drought (22.57%) at the SRI_12, and the smallest the Non-Drought at the SRI_9, the severe hydrologic droughts at the SRI_12 (Figure 3(c) and 3(d)). (iii) For the severe or above droughts, the largest are the drought areas at the SRI_3 and SRI_6 with the area percentages of 25.04 and 27.28%, followed by the SRI_9 (8.97%), and the smallest (6.57%) at the SRI_12 in Figure 3(d). This means that the drought areas of the severe or above hydrologic droughts are Uncorrected Proof also shown a decreasing trend with the increase of time scales. It further demonstrates that the runoff regulation effects or water storage functions of Karst drainage basins are more significant with the time scales increasing.
From the spatial distribution, (i) the hydrological drought intensity of the Karst drainage basins in South China is a gradually aggravating trend from the west to east parts and is showing a significant north-south stripe distribution at the SRI_3 and SRI_6 (Figure 3(a) and 3(b)). (ii) The mild, moderate, and severe or above hydrologic droughts are mainly distributed in the eastern, central, and western of research areas at the SRI_3 and SRI_6 with the area percentages of 28.75, 46.23 and 25.04%; and 20.51, 52.23 and 27.28%, respectively, and in the most part of the central, southwest, and southeast of research areas at the SRI_12 (Figure 3(c)). (iii) Whether at the SRI_3 and SRI_6, or at the SRI_9 and SRI_12, the severe or above hydrologic droughts are found in the Zhijin Basin of Guizhou Province, the Yiguba Basin of Yunnan Province, the Tianzhu and Du'an Basin of Guangxi Province, while the mild or below hydrologic droughts in the Panjiangqiao Basin of Guizhou Province and the Fengyu Basin of Guangxi Province.

Analysis of hydrological drought characteristics based on different lithology-combined structures
As shown in Figure 4, (i) the moderate or below hydrological droughts (À1.5 , SRI) are observed in Classes I to V at the four types of time scales (SRI_3, SRI_6, SRI_9, and SRI_12). The severe or above hydrologic droughts (SRI , À1.5) are gradually decreased with the increase of time scales, especially the severe hydrologic droughts (À2.0 , SRI , À1.5) is shown a parallel distribution and rapid decline. (ii) For the SRI_3 (Figure 4(a)), it is presented a peak value for the mild and extreme hydrologic droughts in Classes I, IV, and V, and a low valley phenomenon for the extreme hydrologic droughts in Class II and the mild hydrologic droughts in Class III, as well as a trend of first increase and then decrease for the moderate hydrologic droughts (À1.5 , SRI , À1.0) in Classes I to V. (iii) In terms of the SRI_6 (Figure 4(b)), the peak value of mild hydrologic droughts is shown a north shifting phenomenon and that of extreme hydrologic droughts is shown a south shifting phenomenon. The distribution areas of moderate hydrologic drought (À1.5 , SRI , À1.0) in Classes I to V are appeared a gradually decreasing trend and are obviously shown a peak value phenomenon in Class III. For the Classes I to V, the densities of moderate or below hydrological droughts are smaller with the density gradient changing slowly, and these of severe or above hydrologic droughts are larger with the density gradient changing rapidly. (iv) For the 9-month scale (SRI_9) (Figure 4(c)), the significant increasing trend and the obvious peak-valley alternating phenomena in Classes I to V are found for the distribution areas of mild hydrological droughts, the rapidly decreasing trend for these of severe or above hydrologic droughts, and the parallel distribution for these of moderate hydrological droughts. The mild hydrologic droughts are shown the two peaks and one valley in Class I, the single-peak distribution in Class V, and the peak flattening phenomenon in Class III, respectively. The severe or above hydrologic droughts are presented the peak value in Classes I, III, and IV, and the low valley in Classes II and V. The larger densities and density gradient changes rapidly are observed for the severe or above hydrological droughts in Classes I to V. (v) As shown in Figure 4(d) at the SRI_12, the distribution areas of mild hydrologic droughts in Classes I to V are accounted for a larger proportion, while these of severe or above hydrologic droughts in Class V has disappeared. There are still two peaks and one valley for the distribution regions of mild hydrologic droughts in Class I, the single-peak distribution for these of severe or above hydrologic droughts, and the low valley phenomenon in Class II. The gradient distributions are obviously appeared for the hydrological droughts in Classes II to V, and the gradient changes of moderate or below hydrologic droughts are relatively slow, and these of severe or above hydrologic droughts are relatively steeping.

Analysis on the driven mechanism of hydrologic drought in Karst Basins
4.3.1. Driven mechanism of single lithological type to hydrologic drought Figure 5 shows that (i) the 11 kinds of watershed lithologies have produced some influences on hydrologic droughts, and these impacts are mainly concentrated in the mild, moderate, and severe hydrologic droughts at the SRI_3 and SRI_6 (Figure 5(a) and 5(b)), and in the moderate or below hydrologic droughts at the SRI_9 and SRI_12 (Figure 5(c) and 5(d)).
(ii) For the 3-month scale (SRI_3) (Figure 5(a)), it mainly occurs the mild and moderate hydrologic droughts in the Pt to K distribution regions. Especially the occurrence probabilities are the largest (0.5-0.6) in the C, P, and T distribution regions, followed by in the Z, ε, O, J, and K distribution regions with the occurrence probabilities 0.35-0.45, while the smallest (0.2-0.3) in the Pt, S, and D distribution regions. (iii) For the 3-month scale (SRI_3) (Figure 5(a)), there are mainly the mild and moderate hydrologic droughts in the Pt to K distribution areas. Especially the occurring probability is the largest (0.5-0.6) in the C, P, and T distribution areas, followed by in the Z, ε, O, J, and K distribution areas with the occurring probabilities 0.35-0.45, while the smallest (0.2-0.3) in the Pt, S, and D distribution areas. The largest (0.75) is the occurring probabilities of severe and extreme hydrological droughts in the Pt, followed by in the S, D, and J distribution regions (0.15-0.35). (iv) It mainly occurs the moderate hydrologic droughts at the SRI_6 (Figure 5(b)). Especially the largest is the occurring probabilities in the C, P, T, and K distribution regions, followed by in the Z, ε, O, S, D, and J distribution regions with the occurring probability 0.35-0.4, the smallest (0.2) in the Pt distribution region. The largest probability (0.75) of severe or above hydrologic droughts is observed in the Pt distribution region, the smallest (0.1-0.2) in the rest of the lithologic distribution regions. The occurrence probability of the severe or above hydrologic droughts is the largest (0.75) in the Pt distribution area, while the smallest (0.1-0.2) in the rest of the lithologies. (v) For the 9-month and 12-month scales (SRI_9, SRI_12) shown in Figure 5(c) and 5(d), the mild hydrologic droughts are mainly presented in the Pt to K distribution regions. Among them, the occurrence probability of mild hydrologic droughts is the largest (0.55-0.8) in the C to K and Z distribution regions at the SRI_12, followed by in the Z to O distributions (0.3-0.6), while the smallest (0.2-0.3) in the S, D, and Pt distribution regions. For the C to K distribution regions, the mild to moderate hydrologic droughts are shown a slower change and smaller gradient at the SRI_9, a faster change and larger gradient at the SRI_12, and the severe and extreme hydrologic droughts have basically not happened. There are the slowest change and smallest gradient for the mild Journal of Water and Climate Change Vol 00 No 0, 14 Uncorrected Proof to moderate hydrologic droughts in the S and D distribution regions, and the severe and extreme hydrologic droughts have a certain occurrence probability (0.15). A slower change and smaller gradient are found for the mild to moderate hydrologic droughts in the Pt to O distribution regions, however the severe hydrological droughts have disappeared, and extreme hydrological droughts have a larger occurring probability (0.55).

Driven mechanism of lithology-combined structures to hydrologic drought
It is known from the lithology-combined structures ( Figure 6) that (i) the five kinds of structures have all produced the influences on the hydrological droughts. Among them, the impacts are mainly concentrated in the mild and moderate hydrologic droughts at the SRI_3 and SRI_6, the mild or below hydrologic droughts at the SRI_9 and SRI_12. This indicates that the response of Karst basins to atmospheric precipitation is shown a lag, and the lagged effects of different lithology-combined structures have great differences. It shows more and more significant with the increase of time scales. (ii) For the 3-month scale (SRI_3) shown in Figure 6(a), the mild and moderate hydrologic droughts are observed in Classes I to V, especially Uncorrected Proof with the largest probability (0.6) in Class II, followed by in Classes I and IV with the occurrence probability 0.5. It hardly occurs the mild hydrologic droughts in Class III, but occurs the moderate hydrologic droughts with the occurring probability 0.4. The severe and extreme hydrologic droughts hardly happen in Classes I, II, and IV, and show the occurrence probabilities of 0.2-0.3 in Classes III and V. The mild to extreme hydrologic droughts are shown the larger gradient and faster change in Classes I and II, and the smaller gradient and slower change in Classes IV and V, as well as a certain fluctuation characteristics in Class III. (iii) For 6-month scale (SRI_6), it mainly occurs the mild hydrologic droughts in Classes I, II, and IV with the occurring probabilities of 0.5-0.6 ( Figure 6(b)), and the moderate hydrologic drought in Classes III and V with the occurring probabilities of 0.35-0.4. Similar to the SRI_3, the severe and extreme hydrologic droughts do not basically happen at the SRI_6 in Classes I, II, and IV, and show a certain occurring probabilities of 0.05-0.2 in Classes III and IV. The occurrence probability of mild hydrologic droughts is gradually decreasing at the SRI_6 in Class V, and that of moderate hydrologic droughts is just the opposite. (iv) Compared with the SRI_3 and SRI_6, it mainly occurs the mild hydrologic drought at the SRI_9 and SRI_12 (Figure 6(c) and 6(d)). Especially the highest (0.45-0.9) is the occurring probability of mild hydrologic droughts in Class V, followed by in Classes I, II, and IV (0.6-0.75), and the smallest (0.4-0.45) in Class III. The mild to extreme hydrologic droughts in Classes I to V are presented the slower change and smaller gradient at the SRI_9, and the faster change and larger gradient at the SRI_12. It hardly occurs the severe or above hydrologic droughts at the SRI_9 and the moderate or above hydrologic droughts at the SRI_12. From the Classes I to IV, the mild hydrological droughts are shown the larger distribution areas with a higher probability at the SRI_9, and the smaller distribution areas with a lower probability at the SRI_12. Especially in Class V, the mild and moderate hydrologic droughts at the SRI_9 are mainly appeared the mild hydrologic droughts at the SRI_12.

DISCUSSION
Attributed to the soluble aqueous medium under the differential erosion or solution effects of soluble water, many sizes of spaces are formed in Karst drainage basins, which provide some places for atmospheric precipitation lag flow on the surface and underground, and makes the basins have a certain storage capacity. This paper justly attempts to explore the watershed storage capacities in terms of lithologic types and spatial coupling structures, and to reveal the mechanism of hydrological droughts. According to this research purpose, this paper firstly selected 55 samples in Karst distribution areas of South China, and collected monthly runoff data from 1970 to 2013, and extracted the basin lithology data. Secondly, the temporal and spatial evolution characteristics of hydrological droughts in Karst drainage basins of South China were analyzed using the SRI, and the mechanism of hydrological droughts were discussed employing the cluster statistical method and Bayesian formula of probability theory in terms of single lithology and spatial coupling of basin lithologies. Therefore, this study has realized the initial research assumption, and these research results have good applicability for Karst drought monitoring and prediction. However, this study mainly considered the Karst research areas, and the research results have certain limitations for the application in Non-Karst areas.
The double aqueous medium and dual water system structures on the surface and underground are developed in Karst drought basins compared with the normal basins, which results in the significant response of watershed runoff to atmospheric precipitation at different time scales (Wang et al. 2002;He et al. 2015aHe et al. , 2015b. This study demonstrates that the hydrological droughts are relatively serious at the 3-, 6-, and 9-month scales, and a relatively lighter at the SRI_12. The occurrence level of hydrological droughts is mainly shown the extreme hydrological drought (6.28%) , severe hydrological drought (6.52%) , non-drought (7.15%) , moderate hydrological drought (32.8%) , mild hydrological drought (47.25%). It proves that Karst drain basins have a stronger water storage function, and their water storage capacities become more and more significant with the increase of time scales. Meanwhile, many types of lithologies, impure textures, and complex structures are observed in Karst drainage basins, where are developed the different sizes of water-stored spaces under the differential erosion or solution effects of soluble water, such as the solution gap, solution hole, and pipeline, as well as the underground cave and underground corridor. As a result, the response of different lithologic combination structures to atmospheric precipitation is significantly different He et al. , 2014López-Moreno et al. 2013). This study indicates that in terms of lithology combination structures, the hydrological drought is the most serious in Class V with the drought area 95.94%, followed by in Classes IV, III, and I with the drought areas of 94.41, 92.25, and 91.95%, respectively, and the relatively lighter (90.95%) in Class II. The severity of hydrological droughts is the Classes II and III (91.79%) , Classes I and IV (92.15%) , Class V (95.94%) in terms of the combination types of watershed lithologies, and the Classes I and III (90%) , Classes II and IV (92.2%) , Class V (95.94%) in terms of Karst lithologies. These conclusions are supported by He et al. (2015b).
The watershed lithologies are the material bases controlling the geomorphic development and soil formation. The different watershed lithologies are shown the different resistance abilities because their lithologic compositions, particle sizes, and composite structures have the larger differences, which promotes or inhibits the formation of water-stored spaces in Karst drainage basins. Therefore, the driving capacity of different watershed lithology types and spatial coupling structures to hydrological droughts have significant differences. This study shows that the driving probability of single lithology to hydrologic droughts is the Pt (0.15) , J, D and S (0.3) , T, O, and ε (0.4) , K (0.45) , P and Z (0.5) , C (0.6) in the distribution areas of mild hydrologic droughts, the Pt (0.15) , ε (0.2) , Z (0.25) , J, D, S, O, P, and C (0.3) , T (0.35) , K (0.4) in the distribution areas of moderate hydrologic droughts, the C (0) , Pt, P, and T (0.05) , Z, O, and K (0.1) , ε, D (0.15) , J and S (0.2) in the distribution areas of severe hydrologic droughts, and the C (0) , O, K, and ε (0.05) , P, T, and Z (0.1) , J and S (0.15) , D (0.2) , Pt (0.6) in the distribution areas of extreme hydrologic droughts. On the whole, the O, C, and T (0.15) , Pt, Z, ε, D, and P (0.2) , S, J, and K (0.25). This may be that the O, C, and T, dominated by the soluble calcite, are easy to be formed the different sizes of water-stored spaces under the differential dissolution or erosion effects of soluble water, which enhances the watershed storage capacity and inhibits the occurrence of hydrologic droughts in a certain extent. On the contrary, the S, J, and K, dominated by the insoluble quartz, feldspar, and kaolinite, are difficult to be formed the water-stored spaces, which weakens the watershed storage capacity and promotes the occurrence of hydrologic droughts. Thus, this study shows that the driving probability of hydrological droughts is the Class II (0.15) , Class III (0.19) , Class IV (0.2) , Class I (0.22) , Class V (0.25) in terms of lithology combination structures, the Classes II and III (0.17) , Classes I and IV (0.22) , Class V (0.25) in terms of the composite types of watershed lithologies, and the Classes I and III (0.18) , Classes II and IV (0.2) , Class V (0.25) in terms of Karst lithologies. It means that under the differential erosion or solution effects of soluble water, the water-stored spaces formed in Karst Basins is the most, followed by in the Semi-Karst Basins, the least in the Non-Karst Basins. This further proves that the watershed storage capacity is the Non-Karst Basin (V) , Semi-Karst Basin (II and IV) , Karst Basin (I and III).
To sum up, the watershed runoff and atmospheric precipitation do not happen at the same time and have a certain lag in time. Its lag time and intensity are greatly affected by the watershed storage capacity. The Karst drainage basin has a certain storage capacity attributed to the Karst lithology with the solubility under the differential dissolution or erosion effects of soluble water. This study proves that there are the most water-stored spaces or the strongest storage capacity in the Limestone Karst Basin (II and III), followed by in the Dolomite Karst Basin (I and IV), the least water-stored spaces or the weakest storage capacity in the Non-Karst Basin (V). Similarly, the most water-stored spaces or the strongest storage capacity are found in the Karst Basin (I and III), followed by in the Semi-Karst Basin (II and IV), the least water-stored spaces or the weakest storage capacity in the Non-Karst Basin (V). Therefore, this study provides a technical guidance for Karst drought monitoring and warning and a theoretical basis for drought relief, and effectively promotes the development of hydrogeology.

CONCLUSION
The watershed lithology is an important component of basin underlying surfaces. The development of the landform and river system and the formation of watershed-stored spaces controlled by the watershed lithologic types and structures will promote or inhibit the occurrence of hydrological droughts. As analyzed above, the driven mechanisms of Karst lithologies and their composite structures to hydrological droughts can be summarized as follows: 1. There is few single lithologic type in Karst drainage basins, and two or more than lithologic types mixed. The basins could be divided into five types according to watershed lithology-combined structures, namely the Dolomite Karst Basin (I), Limestone Semi-Karst Basin (II), Limestone Karst Basin (III), Dolomite Semi-Karst Basin (IV), and Non-Karst Basin (V), respectively. 2. The hydrological droughts of Karst drainage basins in South China are widely distributed, and the drought intensities are gradually decreasing with the time scales increasing. Among them, the hydrological droughts are relatively serious at the SRI_3, SRI_6, and SRI_9, and relatively lighter at the SRI_12. It indicates that the effects of Karst drainage basins on the runoff regulations are more and more significant with the increase of time scales, and hydrologic drought intensities are more and more lighter. Meanwhile, the hydrological droughts from the west to east parts, South China are gradually aggravating with the north-south stripe distribution that is the most significant at the SRI_3 and SRI_6.
3. The driven effects of Karst basins on hydrologic droughts are closely related to the solubility of lithology. Hence, the occurrence probability of hydrological droughts is the O, C, and T (0.15) , Pt, Z, ε, D, and P (0.2) , S, J, and K (0.25). The driven effects of different types of basins on hydrologic droughts have significant differences due to the different lithologic types and spatial coupling structures. Among them, it mainly occurs the mild and moderate hydrologic droughts at the SRI_3 and SRI_6, and the mild or below hydrologic droughts at the SRI_9 and SRI_12. The occurrence probability of hydrological droughts is the Limestone Semi-Karst Basin (