Watershed Characterization and Hydrograph Recession Analysis: A Comparative Look at a Karst vs. Non-Karst Watershed and Implications for Groundwater Resources in Gaolan River Basin, Southern China

Karst watersheds are often treated as non-karst watersheds that can lead to several hazards. Hence, how do karst watersheds differ from non-karst watersheds and what are the effects of karstification on groundwater availability and quality? In this study, we contrast between a karst and non-karst watershed by elucidating their geomorphological peculiarities and potential impact on spatio-temporal availability and quality of groundwater. GIS morphometric mapping and hydrograph recession analysis are applied to map the watershed features and estimate hydrograph recession coefficient to define the groundwater drainage characteristics as well as the influence of karst drainage attributes (KDA). Furthermore, we characterize streamflow components based on the hydrograph recession limbs (segments) and infer their contributing geomorphological factors. Results show that the karst watershed has higher recession coefficients for successive recession limbs. Consequently, it drains larger volumes of groundwater primarily due to the KDAs, which transmit interflow and groundwater flow more rapidly through large cavities to springs as well as stream channels. The KDAs generate what we term karst drainage flow (KDF), defined by the second recession limb which has high recession coefficient as the first limb (overland flow) and strongly contrasts with the non-karst watershed from visual and ANOVA analysis. The effect is that karst aquifer yield over time is significantly lower and highly exposed to pollution compared to the non-karst aquifer. Consequently, sustainable water management practices should be adopted to ensure the availability and safety of groundwater reserves.


Introduction
Groundwater resources in karst regions are under increased pressure from rapid population growth and increased agricultural as well as industrial activities [1].In China alone, over 100 million people rely on karst water resources [2].The multiple porosity matrix in carbonate rocks (limestone, dolostone, gypsum, quartzite, and halite) gives karst aquifers the capacity to house large volumes of water, but also, this extreme heterogeneity (secondary and tertiary porosities) from fractures and underground conduits can result in enhanced flow thereby reducing storage and also increase aquifer are characterized by large depressions, sinkholes, caves, and several springs, some of which are shown in Figure 2. The landcover is predominantly a mixture of Agriculture land and Evergreen Forest.Gaojiaping (non-karst) watershed has about 75% of its entire catchment underlain by an igneous rock while 15% and 10% is metamorphic and carbonate rocks, respectively.The exposed igneous and metamorphic strata are mainly Proterozoic where the igneous comprises of granite and diorite.The metamorphic rocks are comprised mainly of slant gneiss, graphite gneiss, gneiss, and granulite.This composition collectively forms the ancient crystalline basement in the area where the landcover is mainly evergreen forest.are characterized by large depressions, sinkholes, caves, and several springs, some of which are shown in Figure 2. The landcover is predominantly a mixture of Agriculture land and Evergreen Forest.Gaojiaping (non-karst) watershed has about 75% of its entire catchment underlain by an igneous rock while 15% and 10% is metamorphic and carbonate rocks, respectively.The exposed igneous and metamorphic strata are mainly Proterozoic where the igneous comprises of granite and diorite.The metamorphic rocks are comprised mainly of slant gneiss, graphite gneiss, gneiss, and granulite.This composition collectively forms the ancient crystalline basement in the area where the landcover is mainly evergreen forest.

GIS Watershed Characterization
Several studies have applied GIS and Remote Sensing data and methods for different terrains, and they have proved to be efficient for the characterization of drainage parameters [19][20][21].The digital elevation model (DEM) and its derivative dataset (slope, aspect etc.) constitute the building blocks for defining the watershed boundary, flow direction, and stream network, which are generally prepared as input for hydrologic models [22][23][24][25][26][27][28][29].Extrapolating the stream order is conducted based on the hierarchical order method as proposed by Strahler [30] from flow accumulation and flow direction data-sets derived based on the eight-direction pour point algorithm [29,[31][32][33] as a module in ArcGIS surface analysis utilizes DEM raster for determining flow direction of grid cells out of their eight neighbouring cells.Extrapolating the stream order is conducted based on the hierarchical order as proposed by Strahler [30].This is achieved from the flow accumulation and flow direction data-set as derived from the DEM.Since several studies have been carried out describing these methods we will not dwell on it here.For a more detailed description, the reader may refer to Reference [29].Since several studies have been carried out describing these methods, we will not dwell on it here.For a more detailed description, the reader may refer to Dixon and Uddameri [29].Figure 3 shows 30 m × 30 m resolution SRTM DEM data for both study areas.
Water 2019, 11, x FOR PEER REVIEW 4 of 20

GIS Watershed Characterization
Several studies have applied GIS and Remote Sensing data and methods for different terrains, and they have proved to be efficient for the characterization of drainage parameters [19][20][21].The digital elevation model (DEM) and its derivative dataset (slope, aspect etc.) constitute the building blocks for defining the watershed boundary, flow direction, and stream network, which are generally prepared as input for hydrologic models [22][23][24][25][26][27][28][29].Extrapolating the stream order is conducted based on the hierarchical order method as proposed by Strahler [30] from flow accumulation and flow direction data-sets derived based on the eight-direction pour point algorithm [29,[31][32][33] as a module in ArcGIS surface analysis utilizes DEM raster for determining flow direction of grid cells out of their eight neighbouring cells.Extrapolating the stream order is conducted based on the hierarchical order as proposed by Strahler [30].This is achieved from the flow accumulation and flow direction dataset as derived from the DEM.Since several studies have been carried out describing these methods we will not dwell on it here.For a more detailed description, the reader may refer to Reference [29].Since several studies have been carried out describing these methods, we will not dwell on it here.For a more detailed description, the reader may refer to Dixon and Uddameri [29].Figure 3 shows 30 m × 30 m resolution SRTM DEM data for both study areas.

Hydrograph Recession Curve Analysis
As stated earlier, the recession curve refers to the part of the hydrograph that comes after the crest (peak).The recession curve can be characterized by a uniform recession limb or a set of turbulent segments as the curve regresses over time.The segments or inflection points are generally associated with the non-linearity in rock and soil media storage structure which that contributes to streamflow, generally in the form of interflow and baseflow.The segments on the recession curve have been analyzed in the past using several methods, including the exponential method to determine the discharge processes that make up the total streamflow.The exponential function was first expressed by Boussinesq [34] and Maillet [35], expressed as; Since then, it has been widely used for hydrograph recession and base-flow analyses for low flow forecasting, drainage basin studies, pollution sensitivity studies [12,[14][15][16][17][18]36,37].This model can describe a relatively closed, independent and large reservoirs whose main water source is precipitation and major losses are through streamflow discharge [16].Miao and Miao [38] showed how it could be used to describe the drainage of the aquifer reserves over time.The exponential equation can be expanded and further expressed for multiple time-steps as:

Hydrograph Recession Curve Analysis
As stated earlier, the recession curve refers to the part of the hydrograph that comes after the crest (peak).The recession curve can be characterized by a uniform recession limb or a set of turbulent segments as the curve regresses over time.The segments or inflection points are generally associated with the non-linearity in rock and soil media storage structure which that contributes to streamflow, generally in the form of interflow and baseflow.The segments on the recession curve have been analyzed in the past using several methods, including the exponential method to determine the discharge processes that make up the total streamflow.The exponential function was first expressed by Boussinesq [34] and Maillet [35], expressed as; Since then, it has been widely used for hydrograph recession and base-flow analyses for low flow forecasting, drainage basin studies, pollution sensitivity studies [12,[14][15][16][17][18]36,37].This model can describe a relatively closed, independent and large reservoirs whose main water source is precipitation and major losses are through streamflow discharge [16].Miao and Miao [38] showed how it could Water 2019, 11, 743 5 of 20 be used to describe the drainage of the aquifer reserves over time.The exponential equation can be expanded and further expressed for multiple time-steps as: where Q 0 is the initial flow; Q t is the flow at any moment, and α is the recession coefficient (coefficient of discharge) with dimension [T −1 ].It is estimated by plotting the recession limb in a semi-logarithmic plot where Equation (1) appears as a straight-line plot with constant slope independent of the initial Q 0 value [16].Based on the logarithmic scale, the recession curves can generally be classified into several segments, as shown in Figure 4, based on inflection points which are indicative of a transition (micro-regimes) in the drainage structure.Consequently, we can infer major sources and storage systems such as overland and subsurface flows as well as estimate specific volumes of each segment (area under the curve) (Figure 4).For heterogeneous aquifers, on a small scale, especially in areas with highly karstified media, the recession curve can be classified into three (3) or more segments [16,17].Aquifers with a linear storage function have been reported to have no significant inflections and thus can be hardly segmented.Therefore, it can be understood that the non-linear recession curve cannot be solved only by Equation (1).However, by assuming each recession segment as one being "constant" and the second "decaying," then it can be solved using the form expressed in Equations ( 3) and (4).
Water 2019, 11, x FOR PEER REVIEW 5 of 20 where Q0 is the initial flow; Qt is the flow at any moment, and α is the recession coefficient (coefficient of discharge) with dimension [T −1 ].It is estimated by plotting the recession limb in a semilogarithmic plot where equation 1 appears as a straight-line plot with constant slope independent of the initial Q0 value [16].Based on the logarithmic scale, the recession curves can generally be classified into several segments, as shown in Figure 4, based on inflection points which are indicative of a transition (micro-regimes) in the drainage structure.Consequently, we can infer major sources and storage systems such as overland and subsurface flows as well as estimate specific volumes of each segment (area under the curve) (Figure 4).For heterogeneous aquifers, on a small scale, especially in areas with highly karstified media, the recession curve can be classified into three (3) or more segments [16,17].Aquifers with a linear storage function have been reported to have no significant inflections and thus can be hardly segmented.Therefore, it can be understood that the non-linear recession curve cannot be solved only by Equation (1).However, by assuming each recession segment as one being "constant" and the second "decaying," then it can be solved using the form expressed in Equations ( 3) and ( 4).From equation ( 4), considering Q is known, α (recession coefficient) can then be solved for each segment.Additionally, each segment can be quantified and uniquely defined by integrating the flow at initial time step t0, which represents the lower limit and the next time step being the upper limit, while maintaining the recession coefficients for each time step, respectively, expressed as; From Equation (4), considering Q is known, α (recession coefficient) can then be solved for each segment.Additionally, each segment can be quantified and uniquely defined by integrating the flow Water 2019, 11, 743 6 of 20 at initial time step t 0 , which represents the lower limit and the next time step being the upper limit, while maintaining the recession coefficients for each time step, respectively, expressed as; In addition, the total percentage volume of each component (V i , i=1, 2, 3) to total water storage capacity (V 0 ) is calculated as;

Streamflow Data
The streamflow data were collected using an automatic ultrasonic water level gauge (CJ 800) stationed at the outlet of both Miaogou and Gaojiaping watersheds.By visually analysing the hydrograph trend (Figure 5), especially the recession limb, it would be noted that for the karst watershed, the recession is smoother and less turbulent than that of the non-karst watershed.This can be attributed to the very large subsurface karst conduits that easily drain percolated precipitation out of the watershed.The recession limb of the non-karst watershed is more turbulent and thus, indicative of a more segmented drainage due to some active geomorphological impediments.In other words, the karst watershed seems to have a more linear reservoir system.Also, the data shows that the karst watershed generally has less volume of discharge during dry periods (November-March, which marks the dry winter season in the catchment area).During this period, groundwater levels are generally at their lowest making abstraction more difficult, especially in Maiogou.Thus, fifteen hydrographs were selected from 2015 and 2016 for the recession curve analysis for this study.
In addition, the total percentage volume of each component (Vi, i=1, 2, 3) to total water storage capacity (V0) is calculated as;

Streamflow Data
The streamflow data were collected using an automatic ultrasonic water level gauge (CJ 800) stationed at the outlet of both Miaogou and Gaojiaping watersheds.By visually analysing the hydrograph trend (Figure 5), especially the recession limb, it would be noted that for the karst watershed, the recession is smoother and less turbulent than that of the non-karst watershed.This can be attributed to the very large subsurface karst conduits that easily drain percolated precipitation out of the watershed.The recession limb of the non-karst watershed is more turbulent and thus, indicative of a more segmented drainage due to some active geomorphological impediments.In other words, the karst watershed seems to have a more linear reservoir system.Also, the data shows that the karst watershed generally has less volume of discharge during dry periods (November-March, which marks the dry winter season in the catchment area).During this period, groundwater levels are generally at their lowest making abstraction more difficult, especially in Maiogou.Thus, fifteen hydrographs were selected from 2015 and 2016 for the recession curve analysis for this study.

Topographic Analysis
Topography is defined by the general relief of an area, slope as well as the drainage divide which delineates the surface water basin.The hypsographic curve of Miaogou watershed shown in Figure 6 illustrates a complex surface relief where the minimum altitude is 436m while the maximum is 1780 m in an area of only 45 km 2 .The frequency polygons are divided into 10 topographic classes for both watersheds.In the karst catchment, there are considerable amounts of low to flat-lying areas, which are actually karst depressions that extend between karst cones (domes).About 8.5% of the total area of the watershed ranges between 436-770 m while 63% range between 840-1100 m and 31.5% range between 1175-1780 m.This will have a substantial impact on time of concentration for runoff as well as erosion and sediment transport.Furthermore, since infiltration is inversely related to slope, it, therefore, means that the flat slope areas will contribute more to groundwater recharge, especially with the extensive distribution of sinks in and around these low areas.Conversely, the relief in Gaojiaping watershed is more evenly distributed with gentler slopes.No consecutive areas between two contour intervals were found to be higher than 10% of the entire watershed.This is very different to Miaogou watershed where four classes (7-10) of contour intervals alone cover 15.02, 19.4,17.6, and 12.45% of the watershed area, respectively (Figure 6).This highlights the impact of carbonate dissolution in creating a rugged topography compared to the more homogenous crystalline watershed.The phenomenon of highly heterogeneous relief system in Miaogou makes for a more dynamic inter-fluvial process within the watershed thereby promoting more carbonate dissolution, higher drainage density (as is the case), and easy diversion of runoff processes into caves, sinks, and dolines within the catchment.
Water 2019, 11, x FOR PEER REVIEW 7 of 20

Topographic Analysis
Topography is defined by the general relief of an area, slope as well as the drainage divide which delineates the surface water basin.The hypsographic curve of Miaogou watershed shown in Figure 6 illustrates a complex surface relief where the minimum altitude is 436m while the maximum is 1780 m in an area of only 45 km 2 .The frequency polygons are divided into 10 topographic classes for both watersheds.In the karst catchment, there are considerable amounts of low to flat-lying areas, which are actually karst depressions that extend between karst cones (domes).About 8.5% of the total area of the watershed ranges between 436-770 m while 63% range between 840-1100 m and 31.5% range between 1175-1780 m.This will have a substantial impact on time of concentration for runoff as well as erosion and sediment transport.Furthermore, since infiltration is inversely related to slope, it, therefore, means that the flat slope areas will contribute more to groundwater recharge, especially with the extensive distribution of sinks in and around these low areas.Conversely, the relief in Gaojiaping watershed is more evenly distributed with gentler slopes.No consecutive areas between two contour intervals were found to be higher than 10% of the entire watershed.This is very different to Miaogou watershed where four classes (7-10) of contour intervals alone cover 15.02, 19.4,17.6, and 12.45% of the watershed area, respectively (Figure 6).This highlights the impact of carbonate dissolution in creating a rugged topography compared to the more homogenous crystalline watershed.The phenomenon of highly heterogeneous relief system in Miaogou makes for a more dynamic inter-fluvial process within the watershed thereby promoting more carbonate dissolution, higher drainage density (as is the case), and easy diversion of runoff processes into caves, sinks, and dolines within the catchment.

Drainage Network Analysis
Surface hydraulic transport mechanisms are driven by stream channels and their hierarchy which defines drainage density.Miaogou and Gaojiaping stream networks were extracted, and the main channel length as illustrated in Figure 7. Results from hydrographic computations (Table 1) show the karst watershed area to be 45 km 2 and main stream channel to be about 15 km long while the total length of drainage is 30.86 km.Gravelius index was calculated to be 1.93, which is indicative of an elongated drainage system with the potential of having low runoff time of concentration and consequently gentler hydrograph peak.The drainage density is about 0.68 km 2 , which is same as the stream density is 0.68, while the average length of surface runoff is 0.36.

Drainage Network Analysis
Surface hydraulic transport mechanisms are driven by stream channels and their hierarchy which defines drainage density.Miaogou and Gaojiaping stream networks were extracted, and the main channel length as illustrated in Figure 7. Results from hydrographic computations (Table 1) show the karst watershed area to be 45 km 2 and main stream channel to be about 15 km long while the total length of drainage is 30.86 km.Gravelius index was calculated to be 1.93, which is indicative of an elongated drainage system with the potential of having low runoff time of concentration and consequently gentler hydrograph peak.The drainage density is about 0.68 km 2 , which is same as the stream density is 0.68, while the average length of surface runoff is 0.36.On the other hand, Gaojiaping watershed area is computed to be 54 km 2 (Table 1) while its primary stream channel is 11.91 km long and shorter than that of Miaogou.The total length of its drainage is 25.06 km and also lesser than Miaogou, implying that overall, Gaojiaping watershed has shorter stream channels for the conveyance of its total runoff as compared to Miaogou. Figure 8 shows Gaojiaping to have a fan-shaped perimeter with a Gravelius index of 1.65, also lesser than that of Miaogou watershed which implies a higher runoff time of concentration and consequently generate higher flow in shorter period.In addition, the drainage in Gaojiaping is characterized by a dendritic pattern.

Geomorphic Analysis
Generally, geomorphic dynamics is defined by three factors as proposed by Strahler [40], they include gravitational stresses, molecular stresses and chemical processes.The impact of these forces is generally driven by fluvial process, which is governed principally by relief and underlying geology.In Miaogou for example, which is mainly carbonate rocks, interaction with water with presence of CO2 results in dissolution of the bed rock.This, however is not the case for metamorphic rocks in Gaojiaping, which require large amount of heat (solar) to drive its weathering process (rock exfoliation).Thus, karst in humid-tropics which have abundant rainfall are characterized by karst towers or karst cones (domes) that appear like conical haystack rounded hills [41] with elongated ground depressions extending between them which creates low-flat areas as in our study area [42].Visual interpretation of Figure 8(A) reveals some of these features, where about 19 karst cones were delineated in the watershed.The highest cone peak is 1514 m above mean sea level while the lowest is 888m above mean sea level.Additionally, five major karst depressions were also delineated and measured to be around 2.42, 0.48, 1.13 and 0.18 km 2 respectively.In addition, 18 sinkholes were identified during ground survey and where found to be mainly distributed in the upper half of the watershed.Also, it should be noted that Figure 8 has no vertical exaggeration, thus, the high concavity of the karstified watershed further highlights the impact of geomorphic work on the overall karstification phenomenon.The valleys along the depressions illustrate the weathering of folding cavities.Gaojiaping on the hand has low concavity, the slopes are generally homogeneous perhaps due to higher resistance to geomorphic work by its crystalline formations (Figure 8B).On the other hand, Gaojiaping watershed area is computed to be 54 km 2 (Table 1) while its primary stream channel is 11.91 km long and shorter than that of Miaogou.The total length of its drainage is 25.06 km and also lesser than Miaogou, implying that overall, Gaojiaping watershed has shorter stream channels for the conveyance of its total runoff as compared to Miaogou. Figure 8 shows Gaojiaping to have a fan-shaped perimeter with a Gravelius index of 1.65, also lesser than that of Miaogou watershed which implies a higher runoff time of concentration and consequently generate higher flow in shorter period.In addition, the drainage in Gaojiaping is characterized by a dendritic pattern.

Geomorphic Analysis
Generally, geomorphic dynamics is defined by three factors as proposed by Strahler [40], they include gravitational stresses, molecular stresses and chemical processes.The impact of these forces is generally driven by fluvial process, which is governed principally by relief and underlying geology.In Miaogou for example, which is mainly carbonate rocks, interaction with water with presence of CO 2 results in dissolution of the bed rock.This, however is not the case for metamorphic rocks in Gaojiaping, which require large amount of heat (solar) to drive its weathering process (rock exfoliation).Thus, karst in humid-tropics which have abundant rainfall are characterized by karst towers or karst cones (domes) that appear like conical haystack rounded hills [41] with elongated ground depressions extending between them which creates low-flat areas as in our study area [42].Visual interpretation of Figure 8A reveals some of these features, where about 19 karst cones were delineated in the watershed.The highest cone peak is 1514 m above mean sea level while the lowest is 888m above mean sea level.Additionally, five major karst depressions were also delineated and measured to be around 2.42, 0.48, 1.13 and 0.18 km 2 respectively.In addition, 18 sinkholes were identified during ground survey and where found to be mainly distributed in the upper half of the watershed.Also, it should be noted that Figure 8 has no vertical exaggeration, thus, the high concavity of the karstified watershed further highlights the impact of geomorphic work on the overall karstification phenomenon.The valleys along the depressions illustrate the weathering of folding cavities.Gaojiaping on the hand has low concavity, the slopes are generally homogeneous perhaps due to higher resistance to geomorphic work by its crystalline formations (Figure 8B).

Hydrograph Recession Coefficient (α) Estimation
From the streamflow data, fifteen (15) storms enumerated in Tables 2 and 3 were selected for hydrograph recession analysis.The recession curve corresponds to the concave part of the

Hydrograph Recession Coefficient (α) Estimation
From the streamflow data, fifteen (15) storms enumerated in Tables 2 and 3 were selected for hydrograph recession analysis.The recession curve corresponds to the concave part of the hydrograph following a peak i.e., when the crest of the hydrograph is at its highest and is not influenced by recharge processes [43].Using Equation (3), recession coefficients of successive limbs are calculated.In all, four major limbs were segmented for the fifteen storms even though not all storms had four distinct stages.The recession time of the selected hydrographs ranged from a minimum of 70 h (2.9 days) to over 289 h Water 2019, 11, 743 10 of 20 (12 days).Here, it is noteworthy, Bonacci [44] stated that one of the challenges in recession analysis is the time-step selection, where measurement methods in the past often restrict the treatment to a time unit of one day, however, he stresses that this unit can be too long and thus suggested future research to focus on hourly time-step.For this study, we also found this hourly time step as well as the overall duration of about 12 days suitable for the purpose of our study, which is to determine the impact of geomorphological features on the drainage function of karst watersheds against non-karst watersheds.Table 2 shows the results for recession coefficients of Miaogou for four (4) major recession stages (limbs) of the storms.The average value for the first segment is 0.0683 for an average period of 34 h while the second stage has an average value of 0.0121 for an average of 93 h.For the third and fourth stages, the recession coefficient drastically depletes to third and fourth order magnitudes at 0.0039 and 0.0007 respectively.The highest coefficients for both first and second stage segments are 0.1008, 0.0149, and 0.0972, 0.0104 for storms 3 and 4, respectively.For Gaojiaping, the average value for the first segment is 0.0405 for an average period of 31 h while the second stage has an average value of 0.0076 for an average of 96 h.For the third and fourth stages, the recession coefficient drastically depletes to third and fourth order magnitudes at 0.0026 and 0.0004, respectively.The highest coefficients for both first and second stage segments are 0.0630, 0.0100, and 0.0615, 0.0150 for storms 8 and 2, respectively.Figure 9 shows the average recession coefficients of the four stages of the recession curves of the two catchments.
The contrast between the two watersheds is highest during the first and second stages of the recession limb which is characterized by the fast flow drained immediately after the peak flow.While the slow flow represented by the third and fourth stages are delayed.This phenomena is principally in conformity "matrix-restrained flow" and "conduct-influenced flow" (fast and slow flows respectively) regimes elucidated by Kovács et al. [45].Fiorillo [43] asserts that the "matrix-restrained flow" is assumed to discharge instatenously to the outlet and therefore, requires well-developed conduit pervading the entire karst system while the "conduct-influenced flow" is slower and thus, modelled as an equal porous medium (non-karst).The longest period analyzed for Miaogou was greater than 1858 h which had maintained a recession coefficient of about 0.0004, while that of Gaojiaping is slightly shorter at about 1729 h in Gaojiaping where the recession coefficient was steady about at 0.0002 due the denser rock matrix.Gaojiaping is slightly shorter at about 1729 hours in Gaojiaping where the recession coefficient was steady about at 0.0002 due the denser rock matrix.

ANOVA on Recession Coefficient
The analysis of variance (ANOVA) technique was used to determine the significant statistical difference between the recession coefficients of both Miaogou and Gaojiaping watersheds.Results from Tables 2 and 3 for each recession stage of both watersheds were individually analysed as a single group with an alpha level of 0.1.It was observed that comparing each recession segment would statistically be better as opposed to lumping all the recession stages together.This is done with a view to ensuring singular groupings for the ANOVA can identify which stage or stages have the most statistical variance and therefrom infer which watershed attributes have the major impact on streamflow.
The first stage of the recession curve, with a total of 15 records for each watershed showed statistically significant difference in the recession coefficient of the two watersheds as determined by one-way ANOVA at (F (1, 28) = 16.232,P = 0.000389).Hence, the null hypothesis is rejected based on the alpha value set at 0.1, and the P value being significantly lower at 3.89 × 10 −4 .Similarly, the second stage of the recession curve was analysed where the one-way ANOVA was found to have statistically significant difference at (F (1, 28) = 9.884, P = 0.00392).The P value of 3.92 × 10 −3 was found to be significantly lower than the alpha value.Finally, for the third stage, there was the statistically significant difference between the recession coefficient of the two watersheds as determined by one-way ANOVA at (F (1, 28) = 5.859, P = 0.0222).It is noteworthy that from the third stage, the statistical difference begins to reduce in magnitude as the degree of variance between the alpha and P value begins to close up.It can therefore be concluded from visual interpretation of Figure 9 that for the third and fourth stages, the statistical variance is not extremely wide.
Figure 10 also shows the variability discharge ratio ln(Q max /Q mean ) of the karst watershed to be much higher by an order of nearly two magnitudes.In addition, the karst discharge variability shows an exponential growth from 2.26 to 4.14 in about 220 h whereas the non-karst discharge, at about 220 h begins to decay rather rapidly.The variability discharge ratio has often been used to determine the "flashiness" of a spring, which distinguishes diffuse (slow) and conduit (fast) flows [46].
The first stage of the recession curve, with a total of 15 records for each watershed showed statistically significant difference in the recession coefficient of the two watersheds as determined by one-way ANOVA at (F (1, 28) = 16.232,P = 0.000389).Hence, the null hypothesis is rejected based on the alpha value set at 0.1, and the P value being significantly lower at 3.89 × 10 − 4. Similarly, the second stage of the recession curve was analysed where the one-way ANOVA was found to have statistically significant difference at (F (1, 28) = 9.884, P = 0.00392).The P value of 3.92 × 10 − 3 was found to be significantly lower than the alpha value.Finally, for the third stage, there was the statistically significant difference between the recession coefficient of the two watersheds as determined by one-way ANOVA at (F (1, 28) = 5.859, P = 0.0222).It is noteworthy that from the third stage, the statistical difference begins to reduce in magnitude as the degree of variance between the alpha and P value begins to close up.It can therefore be concluded from visual interpretation of Figure 9 that for the third and fourth stages, the statistical variance is not extremely wide.
Figure 10 also shows the variability discharge ratio ln(Qmax/Qmean) of the karst watershed to be much higher by an order of nearly two magnitudes.In addition, the karst discharge variability shows an exponential growth from 2.26 to 4.14 in about 220 hours whereas the non-karst discharge, at about 220 hours begins to decay rather rapidly.The variability discharge ratio has often been used to determine the "flashiness" of a spring, which distinguishes diffuse (slow) and conduit (fast) flows [46].

Streamflow Component Separation
Three typical storms characterized by low, medium, and high volumes of stream discharges were selected for the streamflow components separation.This is to have a balance between dry period and wet periods during the hydrologic year.Using Equations (5-7), as well as the recession

Streamflow Component Separation
Three typical storms characterized by low, medium, and high volumes of stream discharges were selected for the streamflow components separation.This is to have a balance between dry period and wet periods during the hydrologic year.Using Equations ( 5)-( 7), as well as the recession coefficients generated in Tables 2 and 3, respectively, the streamflow recession segments and their respective volumetric flows, computed.The initial and steepest part of the recession should characterize a different hydraulic flow type, as it generally reflects the fall of the hydraulic head in the conduit network.This is because the hydraulic behavior depends on two main hydraulic heads which characterize karst aquifers, and connected to the peculiar structure of the medium [47].The result, shown in Tables 4 and 5 were quantified based on the function that higher recession coefficient implies higher water drainage capacity per unit time which is primarily attributed to topographical and geomorphological characteristics.The geology is particularly critical here since it defines the drainage porosity which in effect is more impactful after the first overland discharge.Also, the function of the aquifer specific yield is known to be given by the ratio of the volume of water drained to the rock volume.Hence, for Miaogou, we defined four ( 4 correlates well with the drainage characteristics of karst aquifers having multiple porosity structures due to karstification [48].Conversely, in Gaojiaping, micro fracture flow, which is generally through regolith deposits and soil sediments contribute the most significant percentage of water ratio.Thus we can conclude that the aquifer matrix of the non-karst basin is more homogenous than that of the karst basin.When the volumes of each stream component are examined and contrasted between the case areas, the disparity becomes apparent.For example, Figure 11 shows the disparity in combined overland and karst drainage flows when compared to the overland, and large fracture flows in between the case study areas.In the first stage of the recession, there is a difference of about 40 m 3 while the second stage has a difference of about 30 m 3 .This implies that the karst watershed tends to discharge larger volumes of water per unit time over the surface as well as subsurface at nearly the same rate.Bonacci [44] elucidated this phenomenon with cases from Switzerland and France [49] and Russia [50] as well as other parts of the world [12], which show that the water retention capacity in karst areas may not be great.This corroborates the findings in this study, even though, the high surficial flow observed in Miaogou, is not primarily attributed to any particular karst feature but more to the basin geometry and morphology, which are ultimately defined by the underlying carbonate system.Hence, there is an indirect correlation but it is merely circumstantial since other geological formations can produce basins geometries that discharge runoff quickly.Similarly, Figure 12 shows the disparity in micro fracture flow for both watersheds where the non-karst catchment has a larger volumetric discharge rate.This property of non-karst watershed is even more pronounced during the wetter phases beginning from the first stage through to the third stage which is drier due to the absence of precipitation.From Figure 11 the drainage volumes for all for flow types for each of our case study areas and stacked against each other.It can be concluded that the karst watershed has a faster and more robust drainage system due to the impacts of topography and especially geomorphology, as exemplified by the karst drainage attributes (KDA) such as caves, sinks, and closed depressions.The non-karst watershed, on the other hand, has the potential to hold more groundwater reserves despite being a crystalline hard rock formation.Similarly, Figure 12 shows the disparity in micro fracture flow for both watersheds where the non-karst catchment has a larger volumetric discharge rate.This property of non-karst watershed is even more pronounced during the wetter phases beginning from the first stage through to the third stage which is drier due to the absence of precipitation.From Figure 11 the drainage volumes for all for flow types for each of our case study areas and stacked against each other.It can be concluded that the karst watershed has a faster and more robust drainage system due to the impacts of topography and especially geomorphology, as exemplified by the karst drainage attributes (KDA) such as caves, sinks, and closed depressions.The non-karst watershed, on the other hand, has the potential to hold more groundwater reserves despite being a crystalline hard rock formation.
stage which is drier due to the absence of precipitation.From Figure 11 the drainage volumes for all for flow types for each of our case study areas and stacked against each other.It can be concluded that the karst watershed has a faster and more robust drainage system due to the impacts of topography and especially geomorphology, as exemplified by the karst drainage attributes (KDA) such as caves, sinks, and closed depressions.The non-karst watershed, on the other hand, has the potential to hold more groundwater reserves despite being a crystalline hard rock formation.

Implications on Groundwater Availability and Quality in Karst Areas
High recession coefficient watershed for successive stages of the recession hydrograph has implications on the availability of groundwater especially during dry periods with no precipitation [17].Fiorillo et al. [17] also reports that when the recession coefficient decreases or increases, the 12. Contrast in Micro Fracture Flow Volume for all four recession stages.

Implications on Groundwater Availability and Quality in Karst Areas
High recession coefficient watershed for successive stages of the recession hydrograph has implications on the availability of groundwater especially during dry periods with no precipitation [17].Fiorillo et al. [17] also reports that when the recession coefficient decreases or increases, the aquifer is emptying more slowly or more quickly than expected, respectively.This is due to the significant influence in hydraulic potential of the aquifers storage capacity resulting from an increase in effective porosity in depth.In the case of Miaogou, higher recession coefficient, along with growing population and industrial activities in the area puts the groundwater system under more pressure.Recent reports from Huangliang Township state that, during the dry winter season, groundwater reserves have been found to be short.
Table 6 shows the average percentile of each streamflow component from the three storms that were selected in this study.In Miaogou, overland flow accounts for 68.12%, karst drainage flow (KDF) accounts for 25.47%, medium fracture flow is 3.93% while micro fracture flow is only 2.48%.However, in Gaojiaping, only 35.44% of its average streamflow is from overland flow compared to Miaogou's 68.12%.Furthermore, the delayed flow, that is, micro fracture flow is 31.74%making it more than twelve times the size of Miaogou karst watershed.This highlights low groundwater yield in Miaogou due to drainage excess by KDAs.These two characteristics agree with morphometric analyses in above where it was established that Miaogou has a higher concavity that dominates about 50% of its cumulated surface while its average stream network slope is higher than that of Gaojiaping watershed.Thus, we see a higher overland flow percentage as well as a higher macro fracture percentage while the converse is seen in the medium and micro fracture flows, respectively.This disparity between the karst and non-karst watershed is shown in Figure 13, illustrating the inverse relationship between the respective streamflow components and also strongly highlights the groundwater resources availability and potential of both catchments.Besides the availability of groundwater reserves, another significant concern is groundwater quality.Malik's [18] revised aquifer pollution sensitivity index show that multiple recession stages with high recession coefficients values imply high heterogeneity and thus higher sensitivity to pollution.Also, Jakada et al., [5] carried out an intrinsic vulnerability assessment of Miaogou catchment area based on the EPIK parametric method and found that most of the catchment is highly vulnerable to pollution.Specifically, Huangliang Township, which has recently been reporting groundwater shortages, falls under the "very high" vulnerability index.Thus, results from this work validate this conclusion as well as establish the fact that groundwater quality will continue to deteriorate until drastic environmental management practices are enforced.

Conclusion
This is the first attempt at a comparative hydro-geomorphological and hydrograph recession analysis between a karst and non-karst watershed.By comparing these two distinct watersheds, under common precipitation conditions, a major contrast was observed based on their hydrograph recession characteristics, which has direct correlation with shallow and deep aquifer drainage properties.Results show that the karst watershed in contrast to non-karst watershed tended to drain larger volume of precipitation over shorter period.This is due to larger amounts of interflow influenced by large fractures, while other karst features such as sinks and closed depressions act as siphoning conduits that transmit runoff rapidly through well-developed underground cavities to springs which drain into the surface water system (streams and river channels).These common features usually found in highly karstified areas, referred to in this study as karst drainage attributes Besides the availability of groundwater reserves, another significant concern is groundwater quality.Malik's [18] revised aquifer pollution sensitivity index show that multiple recession stages with high recession coefficients values imply high heterogeneity and thus higher sensitivity to pollution.Also, Jakada et al., [5] carried out an intrinsic vulnerability assessment of Miaogou catchment area based on the EPIK parametric method and found that most of the catchment is highly vulnerable to pollution.Specifically, Huangliang Township, which has recently been reporting groundwater shortages, falls under the "very high" vulnerability index.Thus, results from this work validate this conclusion as well as establish the fact that groundwater quality will continue to deteriorate until drastic environmental management practices are enforced.

Conclusions
This is the first attempt at a comparative hydro-geomorphological and hydrograph recession analysis between a karst and non-karst watershed.By comparing these two distinct watersheds, under common precipitation conditions, a major contrast was observed based on their hydrograph recession characteristics, which has direct correlation with shallow and deep aquifer drainage properties.Results show that the karst watershed in contrast to non-karst watershed tended to drain larger volume of precipitation over shorter period.This is due to larger amounts of interflow influenced by large fractures, while other karst features such as sinks and closed depressions act as siphoning conduits that transmit runoff rapidly through well-developed underground cavities to springs which drain into the surface water system (streams and river channels).These common features usually found in highly karstified areas, referred to in this study as karst drainage attributes (KDA), generate what we also termed karst drainage flow (KDF).The KDF is defined by the second limb of the recession hydrograph and also shares similar properties as overland flow due to high discharge volume as well high recession coefficient.Kresic and Bonacci [39] also argue that the steep curve one the recession limb represents turbulent drainage of large fractures and conduits, followed by a transitional portion that is less turbulent and reflects the contribution of smaller fractures and rock matrix, ending with the slowly decreasing curve, the so-called master recession curve, where the drainage of rock matrix and small fissures is dominant.However, it must be stated that this study by no means attempts to draw conclusions that hydrograph recession analysis fully describes the internal groundwater structure, as the initial catchment area is karst can never be reliably defined [44].Consequently, the statistically significant variance in recession coefficients between the second stages of the two watersheds simply emphasizes a strong relationship between the recession coefficient and KDAs, as well as catchment morphometric properties, as posited by references [11][12][13].
Additionally, with increasing pressure on groundwater reserves globally, our results seek to draw attention to karst aquifer water retention abilities [12,49,50] relative to groundwater quality, not only locally in the study area but also in other parts of the world.Since higher recession coefficient index for successive recession segments (regimes) imply strong sensitivity to groundwater pollution, the environmental implications cannot be ignored.Escolero et al. [51] state that "the high permeability of karst terrains result in many practical problems such as scarcity and poor predictability of groundwater supplies; scarcity of surface water bodies; instability of the ground (due to the presence of sinkholes); leakage of surface water reservoirs, and an unreliable waste disposal environment".Previous groundwater vulnerability studies established that the Miaogou area has a high intrinsic susceptibility to pollution due to karstification, hence, sustainable water management measures must be adopted in the area.

Figure 2 .
Figure 2. Some karst features located in Miaogou watershed and environs (A) Xiaosanyou Cave; (B) doline with dotted lines enunciating the perimeter; (C) large sinkhole with dotted lines enunciating the perimeter; (D) vracture with dotted lines enunciating the orientation of fracture; (E) epikarst; and (F) bailongquan spring.

Figure 2 .
Figure 2. Some karst features located in Miaogou watershed and environs (A) Xiaosanyou Cave; (B) doline with dotted lines enunciating the perimeter; (C) large sinkhole with dotted lines enunciating the perimeter; (D) vracture with dotted lines enunciating the orientation of fracture; (E) epikarst; and (F) bailongquan spring.

Figure 2 .
Figure 2. Some karst features located in Miaogou watershed and environs (A) Xiaosanyou Cave; (B) doline with dotted lines enunciating the perimeter; (C) large sinkhole with dotted lines enunciating the perimeter; (D) vracture with dotted lines enunciating the orientation of fracture; (E) epikarst; and (F) bailongquan spring.

Figure 4 .
Figure 4. Schematic presentation of the hydrograph recession limb with four recession segments and corresponding volumes of discharged water (adapted from Kresic N. and Bonacci O. [39]).

Figure 4 .
Figure 4. Schematic presentation of the hydrograph recession limb with four recession segments and corresponding volumes of discharged water (adapted from Kresic N. and Bonacci O. [39]).

Figure 8 .
Figure 8. 3D Geomorphometric illustration (A) Miaogou watershed characterized by an elongated shape with the presence of spring, sinkholes, caves, karst depressions and karst cones (B) Gaojiaping watershed characterized by a broader fan-shaped area with no karst features.

Figure 8 .
Figure 8. 3D Geomorphometric illustration (A) Miaogou watershed characterized by an elongated shape with the presence of spring, sinkholes, caves, karst depressions and karst cones (B) Gaojiaping watershed characterized by a broader fan-shaped area with no karst features.

Figure 9 .
Figure 9. Average recession coefficients of the four stages of the recession curves.Figure 9. Average recession coefficients of the four stages of the recession curves.

Figure 9 .
Figure 9. Average recession coefficients of the four stages of the recession curves.Figure 9. Average recession coefficients of the four stages of the recession curves.

Figure 10 .
Figure 10.Variability discharge ratio ln(Qmax/Qmean) showing higher discharge rates with respect to time in karst watershed.

Figure 10 .
Figure 10.Variability discharge ratio ln(Q max /Q mean ) showing higher discharge rates with respect to time in karst watershed.
) classes of flow regime that contribute to streamflow; (1) overland flow, (2) karst drainage flow, (3) medium fracture flow and (4) micro fracture flow.For Gaojiaping the classes are (1) overland flow, (2) macro fracture flow, (3) medium fracture flow, and (4) micro fracture flow.The streamflow separation results show that overland flow and karst drainage flow have the largest percentages for all three storms that were evaluated.This phenomenon Water 2019, 11, 743 13 of 20

Figure 11 .
Figure 11.Contrast in Overland + Karst Drainage Flow Volume vs. Overland + Macro Fracture Flow Volume.

Figure 11 .
Figure 11.Contrast in Overland + Karst Drainage Flow Volume vs. Overland + Macro Fracture Flow Volume.

Figure 12 .
Figure 12.Contrast in Micro Fracture Flow Volume for all four recession stages.

Figure 13 .
Figure 13.The plot of disparity in flow type average percentages contribution to streamflow.

Figure 13 .
Figure 13.The plot of disparity in flow type average percentages contribution to streamflow.

Table 2 .
Table of recession coefficients and recession stages at Miaogou Karst Watershed.

Table 3 .
Table of recession coefficients and recession stages for Gaojiaping Watershed.

Table 3 .
Table of recession coefficients and recession stages for Gaojiaping Watershed.

Table 4 .
Table of streamflow component separation for Miaogou watershed.

Table 5 .
Table of streamflow Component Separation for Gaojiaping watershed.

Table 6 .
Flow type statistics for all three storms and their average percentages.