Sinkhole flooding and aquifer recharge in arid to dry sub-humid regions: a systematic review

Sinkhole flooding is an essential hydrological process to recharge karst aquifer in arid to dry sub-humid regions. On the other hand, the increase of rain extremes is one of the major consequences of the global warming, together with the expansion of drylands. Thus, appropriate runoff regulation in endorheic karst basins in order to reduce the risk of flooding and improve the quantity and quality of the water drained by sinkholes will be more and more crucial. With these premises, a systematic review was performed by using WoS engine to infer the best practices for the karst water management in regions actually or potentially affected by water scarcity. Hydrological models are essential to manage the consequences of climate change on karst water resource, however the review shows that providing the tools necessary for reliable modeling is still challenging. Finally, due to the intrinsic vulnerability of the karst aquifers, pollution reduction and wastewater recycling policy will play key role in the next decades.


Introduction
In arid to dry sub-humid regions, Sinkhole Flooding (SF) is a process essential to recharge karst aquifers, but difficult to manage [1][2][3][4][5]. Climate change and population growth are continuously increasing human pressure on water resources and thus, it is appropriate to investigate how accumulated scientific knowledge about SF can support sustainable management in conditions of water depletion. Not by chance, defining the issue it deals with as the "land degradation in arid, semi-arid and dry sub-humid areas resulting from various factors", the United Nation Convention to Combat Desertification has brought together, for research purposes, the lands actually or potentially affected by water scarcity [6]. Before addressing the question of SF, some definitions and basic features must be given.
It must be also considered that, since the ancient civilizations, hydraulic works have frequently altered the hydrology of the karst catchments, thereby changing both the infiltration pathways and the floodplain features. Especially the sinkholes have historically served for runoff regulation, stormwater discharge and wastewater disposal. A further function, the intentional (managed) groundwater recharge, was added in recent times [3,[22][23][24]. As a whole, karst aquifers currently supply a quarter of the global population with water. They are highly sensitive to changes in recharge patterns because they combine a high transmissivity with a low specific yield, thus their correct use is vital to ensure the water availability for years to come [25,26].
Despite SFs are essential for groundwater balance, they are still difficult to foresee and manage [27,28]. They occurs as consequences of extreme precipitation when: 1) the rate of stormwater flow exceeds the discharge capacity of the sinkhole (SF type 1); 2) the underground karst system is unable to drain the entire stormwater flow (SF type 2); 3) the groundwater level rapidly rise especially because of the diffuse infiltration (SF type 3) [29,30] (Figure 1).

Figure 1.
Simplified scheme of SF types. It is not representative of the whole complexity of the phenomenon and simply considers change in width of the karst shafts. The increase in diffuse infiltration from SF type 1 to SF type 3 is a generic trend.
Together with the expansion of drylands, just the increase of rain extremes is one of the major consequences of the changes in climate pattern and hydrological cycle due to the global warming [7,9,[31][32][33]. With these premises, a systematic review of documented cases of SF occurred in arid to dry sub-humid areas was performed to infer the best practices for the upcoming water management plans. Particular attention has given to observational studies, because of the general lack of flow time-series data in the karst catchments literature, i.e. the basic data for any models validation or verification [14,[34][35][36].

Methods
The systematic review of literature differs from the typical narrative review by the effort to establish a clear, not subjective and replicable procedure for the studies identi-fication. Became initially important in medical and healthcare field, it gained increasing popularity in Earth and Environmental Sciences through the last decade.
In the present work, the systematic review was conducted according to [37][38][39][40]. It aimed to the comprehensive identification and appraisal of the studies relevant for the pursued subject. Web of Science (WoS) was chosen as bibliographic data source because suitable for the purpose of the review [41][42][43]. The identification of the studies was implemented in all WoS databases. A limitation of the adopted procedure is that it only includes items written in English. Moreover, a temporal period was not specified. The screening process has consisted of three stages. In the first one, WoS engine search was used as the primary platform for "title/abstract/author keywords/Keywords Plus". The abstract screening was performed in Stage 2 and the full-text articles assessment for eligibility in Stage 3 ( Figure 2). Careful exclusion procedures were adopted in Stages 2 and 3.

Results
WoS engine allowed the identification of eightyseven records, only one of which was excluded as duplicate ( Figure 2). The selection of Stage 2 has determined the exclusion of fiftyfour articles because apparently irrilevant to the review purpose and belonging to different subject areas (i.e. ecology, hydrobiology, geomorphology, sinkhole occurrence, regional hydrogeology, karst hydrochemistry, baseflow estimation, particles transport, paleoclimate, civil engineering, pollution impact, vulnerability assessment, rainfall chemistry, mathematical model). Despite of the positive evaluation of the abstracts, eleven out of thirty-two articles were then excluded by the full-text assessment of the Stage 3 (Appendix A). These exclusions were caused by the irrelevance of the article on the sinkhole flooding process, or because the article deals with cases of humid regions (Table A1). Eleven articles identified by other sources (examination of the bibliography of the selected articles; personal literature knowledge; use of common search engines) have been added for the review analysis.  [37,40]).
"Aquifer recharge" is the main objective for ten articles (third column of Table 1), while 1 for the others it is "flood management" (twelve articles) or "contamination risk" (six). Four 2 articles address "multiple objectives". Twenty-two studies state the type(s) of examined SF 3 or provide information for such classification. 4 The main topics identified by the literature examination (types of sinkhole flooding, 5 Sub-Section 3.1; flood events series, Sub-Section 3.2; sinkhole discharge capacity, Sub-6 Section 3.3, aquifer recharge process, Sub-Section 3.4; water resource planning, Sub-Section 7 3.5) are summarized below. Half of the twenty-two reviewed studies address the SF type 1, while among the 10 remaining studies prevail the SF type 2 (six articles). Only two cases were confidently 11 ascribed to the SF type 3. Finally, in four studies the attribution is uncertain or SF is the 12 result of two concurrent causes ( Middle East contamination risk und. The ascription to the SF type 1, imply the knowledge of the discharge capacity of the 14 sinkhole or, at least, of its underground shape and dimensions [1,5]. Despite the lack of data 15 on the underground karst system, Herczeg et al. [45] provided pioneer field observation  (Table 1). Estimating the aquifer recharge for the small endorheic As Sulb Basin (located 22 in the hyper arid As Summan Plateau, Saudi Arabia, Middle East) by camera monitoring, 23 Schulz et al. [61] were incidentally able to visually define the occurred SF type 1. Where 24 the sinkhole is a dissolution doline covered by residual soil, the issue to define the SF type 25 does not arise as it is of the first type (see e.g. [19]). 26 The ascription to the second type of SF requires accurate knowledge about the de- 27 velopment of the underground karst system, previous achieved mainly by speleological 28 exploration and mapping [2,46]. Dealing with an extreme flood occurred within a flat karst 1988 [44]. Some authors who have studied cases later occurred in the climatic regions 32 herein considered, were of the same opinion [47][48][49]54,66]. Other authors have considered 33 SF type 2 as a concurrent cause together with type 1 or type 3 [35,50,67] (see Table 1). The 34 case described by Liguori and Manno [48] (Platani Basin, Italian Peninsula) is quite unique 35 in the context of this review, dealing with sinkholes induced by quarrying and then flooded 36 because of the water filling of excavation tunnels. Drilling data and speleological surveys, 37 both suggesting the decreasing of the width of conduits with depth for the southern slope 38 of the Falakro Mountain (Greece, Balkan Peninsula) allow to attribute to type 2 also the SF 39 studied by [47]. The recurrent SF affecting the Gokova Graben (Anatolian Peninsula) must 40 be attribute to the type 2 as a result of the spectral analyses on the precipitation and spring 41 discharge time series carried out by [66]. 42 The type 3 is apparently the least common of SF. In fact, only two of the thirty-one 43 reviewed studies attribute the flooding to the quick rise of the groundwater level [13,58]. 44 Both these studies support such an attribution by means of aquifer monitoring. For a 45 few other studied cases, type 3 is tentatively referred to as concurrent cause in possible 46 conjunction with type 2 ( Table 1). 47

Flood events series 48
Several essential features arose from the full-text assessment on SF series reported 49 in literature. A three years series of floods is analysed by Bailly et al. [54] as regards the 50 swallow hole named Puits de l'aven, which is placed within the karst basin of Coulazou 51 River (France, South Continental Europe) (see also [49]). The thirteen events occurred from study, which made use of various instruments such as rain gauges, discharge gauging 56 stations, and monitoring wells, was decisive for the obtained findings. 57 Guo et al. [15,56] monitored the recurrent floods affecting karst depressions placed 58 within the sub-humid and seasonally drought-prone Fengshan County (Guangxi) and 59 Guangnan County (Yunnan), South China. Floods were found to last up to four months 60 during the rainy season, in stark contrast to the extreme drought conditions of the dry 61 season. Moreover, several process of sinkhole clogging (see e.g. [4,63,72]) were detected. 62 The flood history of some endorheic catchments was scrutinized in the reviewed 63 literature. Parise [46] found more than twenty flood events struck the karst basin of  The discharge capacity is a basic parameter in the relation between SF and aquifer 80 recharge. However, it is rarely quantified or at least estimated in the literature studies 81 [16,63]. Such a result confirms the concern of Field [73] which likely first observed the "little  The discharge capacity is a simple direct function of the water level in the sinkhole 85 when flow in the underground karst system in not under pressure. Exceeded a threshold 86 value of the sinkhole water level, the flow comes under pressure and the discharge capacity 87 becomes dependent on the difference in head hydraulic between the sinkhole and another 88 measuring point (a spring, as an example) [74]. Despite of its good reliability, such an 89 approach requires extensive measures for the computation of a coefficient and thus it is 90 rarely used. However, dealing with sinkhole of simple shapes (cylinder-shaped, cone- 91 shaped, bowl-shaped), simplified mathematical models were developed to estimate the 92 discharge capacity [73]. 93 The above theoretical approaches are not frequently used. [54] performed an analysis 94 on the infiltration rate in the Puits de l'aven sinkhole (see 3.2) during recession, using 95 it to estimate the hydraulic conductivity of the karst system rather than to establish the 96 discharge capacity. 97 Rough evaluations of the total discharge capacity for clusters of sinkholes located  In the next decades, a major challenge will be quantifying the change in aquifers 141 recharge due to the variation in precipitation frequency and amount. As deducted by the 142 literature, for this scope the use of reliable models will play a role of reference [61,65].

165
In arid to dry sub-humid regions, the Hortonian runoff concept [78] should be carefully 166 considered in modelling for water management [79][80][81]. Thus, according to Iacobellis et 167 al. [59], the soil moisture condition before the rainfall event must be quantified together 168 with the sinkhole discharge capacity. The latter is used, in the model proposed by these 169 authors, to calculate the "storage capacity", i.e. the maximum volume of water that can be 170 stored in a sinkhole before the flooding of the surrounding area. Moreover, the rainfall 171 input should be provided by means of the "intensity duration frequency curve" for a given 172 return period.

174
Over the last two decades, a worldwide decline in water storage for sub-continental 175 sized endorheic regions has been detected [82,83]. At catchment scale, several studies 176 highlighted this phenomenon in many karst basins (see e.g. [84][85][86][87]). In arid to dry sub-177 humid regions, aquifer recharge through sinkholes is a discontinuous process that occur as 178 a result of flooding or intense runoff [45,47,61]. If wet soil moisture condition precedes the 179 rainfall event, extensive runoff is triggered and contribute to large SF with considerable 180 water accumulation [28,59]. Where SF occurs in karst areas threatened by water scarcity, 181 pollutant free water accumulated on the surface should be discharged into the aquifer 182 rather than evaporate to the atmosphere. The main influencing factors of such a process 183 are: duration and intensity of the rain, discharge capacity of the swallow holes, average 184 soil moisture, and mean air temperature [19,34,65]. the stress on karst water sources is likely to significantly growth over the next decades.

190
However, given the current uncertainties, a very difficult question to address is how much 191 potential changes in temperature and precipitation will affect water availability in karst 192 regions [6,88,89].

193
The need to correctly regulate the runoff in endorheic karst basins in order to reduce 194 the risk of SF type 1 (i.e. the most recurrent type of SF according to the reviewed studies, 195 see Sub-Section 3.1) and, at the same time, improve the quantity and quality of the water 196 drained by the sinkholes is crucial anyway [16,18,55] Sub-Section 3.4).

231
Where the anthropic impact on a karst area is significant to manage the groundwater 232 recharge is mandatory [35,63]. Currently, the volume of managed aquifer recharge at global 233 scale is estimated in 10 km 3 /year, while its rate of expansion in 100 km 3 /year [24]. This Due to the consequences of climate change and the intrinsic vulnerability of the karst 243 aquifers, pollution reduction and wastewater recycling policy will play key role in the next 244 decades [16,24,35,93]. While in karst areas belonging to humid regions the dilution process 245 can partially reduce the contaminant concentrations [94], for those belonging to arid to dry 246 sub-humid regions it plays no role. Therefore, treated wastewater facilities will have to 247 avoid fast direct aquifer recharge (through sinkholes or injection wells) in favour of slow 248 and delayed diffuse infiltration [95]. Soil covered epikarst should be preferentially used 249 in wastewater recharge because of its natural treatment processes such as contaminant 250 adsorption and spread of the injected effluent.

251
This review concludes remarking that hydrological models are essential to manage 252 the consequences of climate change on karst water resource [88,93].