Spatial distribution of groundwater recharge , based on regionalized soil 1 moisture models in Wadi Natuf karst aquifers , Palestine 2

12 While groundwater recharge is considered fundamental to hydrogeological insights and basin management, only 13 relatively little attention has been paid to its spatial distribution. And in ungauged catchments it has rarely been 14 quantified, especially on the catchment scale. 15 16 For the first time, this study attempts such analysis, in a previously ungauged basin. Our work based on field data 17 of several soil moisture stations, which represent five geological formations of karst rock in Wadi Natuf, a semi18 arid to sub-humid Mediterranean catchment in the occupied Palestinian West Bank. For that purpose, recharge 19 was conceptualized as deep percolation from soil moisture under saturation excess conditions, which had been 20 modelled parsimoniously and separately with different formation-specific recharge rates. 21 For the regionalisation, inductive methods of empirical field-measurements and observations were combined with 22 deductive approaches of extrapolation, following the recommendations for hydrological Prediction in Ungauged 23 Basins (PUB), by the International Association of Hydrological Sciences (IAHS). Our results show an average 24 annual recharge estimation in Wadi Natuf Catchment (103 km), ranging from 24 to 28 Mm/yr, equivalent to 25 recharge coefficients (RC) of 39-46% of average annual precipitation. 26 27 Thus, for the first time, formation-specific RC-values could be derived, assessed and quantified in their spatial 28 distribution, and by creating a schematic conceptual basin classification framework for regionalisation that is also 29 applicable in many comparable sedimentary basins in the Mediterranean and worldwide. 30 31 32

process, however, is often beyond the reach of observations and measurements, particularly in poorly or entirely 48 ungauged basins around the world. all three groups, and reports that these parameters for groundwater recharge were not actually measured in the 12 field and instead conceptual assumptions were used to assign them with weights as variables in a basin-wide 13 transfer function between spatial characteristics and hydrological response.

15
In ungauged basins, where information on hydrological basin responses are missing, physical parameters can be   (Yadav et al., 2007). In addition, the use of so-called 37 hydrological system signatures can help create a link between physical features and basin response and to describe     The WAB is an up to 1000 m thick Upper Cretaceous carbonatic karst aquifer (SUSMAQ, 2002) and 10 conventionally divided into two regional aquifer layers (

12
However, this simplified regional hydrostratigraphy applies only to the Coastal Plain downstream, with its 13 productive abstraction and discharge zone, where the fully confined aquifer acts uniformly and with a low  Table 1, section 2. On a local scale, especially in the phreatic zone 15 upstream, the hydrostratigraphy is far more complex than the above-mentioned bipartite division into Upper and 16 Lower Aquifers.

18
Importantly, whereas the productive Coastal Plain is well developed, monitored and gauged through hundreds of

19
Israeli deep wells, the WAB recharge and accumulation zones in the mountains, slopes and foothills of the West 20 Bank, remain almost untouched, ungauged and unexplored, due to severe Israeli restrictions on Palestinian water

25
So far, only a few authors have attempted the analysis of fully distributed recharge in the WAB (Hughes et al., 26 2008) and no previous study was based on empirical field evidence, measurements and observations. Sheffer

27
(2009) introduced a semi-distributed, partially lumped recharge model, however with a very coarse lithological 28 differentiation into merely two types of rock, either permeable or less permeable. In addition, Sheffer (2009) took 29 his soil model parameters from the general literature and later adjusted them by calibration. In his own words, he 30 focussed and aimed at 'the understanding of temporal influence on recharge processes', rather than on 31 understanding spatial influences (Sheffer et al., 2010;Sheffer, 2009).

33
During the last two decades, other studies of field-based and empirical investigations on sub-catchment, local and 34 plot-scales were conducted. Chloride mass balance calculations were carried out in the adjacent Eastern Aquifer  In the WAB, lumped studies of basin-wide replenishment are widely available, however, mostly based on desktop 43 work. By contrast, distributed recharge quantification has hardly been attempted. Moreover that, the physical form 44 and the spatially variable parameters that rule the recharge process were not observed or measured directly in the  In order to advance the crucial but challenging task of a realistic representation of distributed recharge, this study 3 on Wadi Natuf presents a novel combination of existing techniques that are based on observable processes, 4 parameters and signatures. The assessment adheres to the goal of parsimony and that integrates inductive and 5 deductive steps. The previous paper (Messerschmid et al., 2019) was firmly grounded in field observation, 6 measurements and a forward-calculating location-specific model; now, this current paper extends the findings of 7 the local models in a regionalisation effort to the entire surface catchment area of 103 km 2 .

9
The study aims at generating specific recharge coefficients for every litho-stratigraphic formation in Wadi Natuf

38
According to the old, conventional viewvalid on the regional scalethe regional Upper and Lower Aquifers   Fig. 2b). However, closer scrutiny reveals that this regional 'Middle Aquitard'

43
can be further subdivided. The top forms an aquitard or even aquiclude section of impermeable yellow soft marl 44 (upper Yatta, u-Yat). By contrast, the main (lower) part of this 'regional aquitard' is more carbonatic and in parts

4
Also, the regional 'Lower Aquifer' (LBK & UBK) must be differentiated on the local scale into more aquiferous 5 and more permeable parts (Table 1). Its top is formed by the conspicuous cliff-forming and very permeable reefal 6 limestone of upper UBK (u-UBK), that also acts as a leaky perched aquifer on the local scale (such as in Wadi 7 Zarqa). By contrast, the lower UBK formation (l-UBK) mostly consists of banked, often chalky dolomites (again 8 with intercalations of marl and chert) with a relatively poor aquifer potential. Its top however was found to be more 9 carbonatic but underlain by a twin marl band (Fig. 3c), which hydraulically separates the top from the main, lower 10 part of l-UBK and above which local contact springs align. This top of l-UBK acts as a third local and isolated 11 perched aquifer horizon.

12
By contrast, the regional 'Upper Aquifer' is void of both, perched aquifers and springs, despite the fact that it too

22
These isolated perched hilltop aquifers of central Wadi Natuf stand in contrast to the thick regional aquifers and 23 therefore only incompletely contribute to the deep regional groundwater recharge of the two regional storage and The outcrop areas of all formations, as well as the differences between the local and the regional hydrostratigraphy 1 form one focus of the present study (see Table 1).  Note: The land cover types 4 and 5 are almost completely restricted to Yatta formation outcrops (and in some places parts of 7 the UBK formations). Type 3 is typically found over outcrops of the regional Lower Aquifer. The Upper Aquifer outcrop area 8 is dominated by type 1 (grassland and barren rock). Olives (type 2) can be found in all areas, but are grown on tended terraces 9 mostly in the steep slopes of the Lower Aquifer outcrops in upper Wadi Natuf.

23
Less than 5% of the rural Wadi Natuf landscape are built-up (Messerschmid, 2014). Its typical land forms (Fig. 2) 24 range from rock outcrops and terraces with olives, over grass-and shrublands, arable but currently uncultivated Hebron formation.

13
Typically, in Wadi Natuf, this distribution of LU/LC follows the formation outcrops (geology) with great accuracy, 14 discernible even from aerial photographs. Also soil thickness was measured and found to strongly correlate with 15 lithology and land forms (LU/LC) as discussed in the first part of this series (see Table D1     The classification of distributed physical landscape features and their parameters stands at the heart of this study.

41
Mapping, detection, interpretation and where possible, quantification of their parameters was carried out over a 42 period of more than ten years and over 200 field visits to gain local knowledge on specific field conditions.

43
Accordingly, the landscape characteristics in Wadi Natuf could be attributed to three groups: geology, soil

27
Thirdly, and similar to geology and soil thickness, land use and land cover characteristics, such as relief, natural

37
Conceptually, as already mentioned, the regionalisation in this study comprises of two main steps (rows 2, 3 in Fig Based on the PUB-understanding that physical characteristics control hydrological processes and thus 1 (hydrological) function follows (physical) form, a conceptual framework was set up, as shown in Table 2. The 2 physical features were divided into three groups, such as LU/LC, soil and geology (columns in Table 2), and within 3 each group separately, the different landscape units were divided into distinct classes of recharge potential (lines 4 in Table 2), based on the available geological literature and our extensive field investigations. Then, each 5 lithostratigraphic formation (numbered a, b, c, etc. in the schematic Table 2) was attributed to a distinct recharge 6 class (from low to high in Table 2; as roman numbers I -V in Table 3). As a result, we obtained three independent

16
The general classification framework was then applied to Wadi Natuf and the specific physical features were inserted 17 in Table 2, to obtain a conceptual recharge classification framework, specific for Wadi Natuf (Table 3). Here, in each  (Table 3), resulting in Table 4. However, the modelled RC-values

37
This analysis results in a basin classification framework that categorizes different groups of recharge potential and

38
attributes each formation to one of these classes, shown in Table 3. Each formation is attributed to different classes   Table 4 shows the modelled and the newly attributed and inserted average  Table 4 lists the 15 different outcropping formations in Wadi Natuf in chronological order from the youngest,

21
Yat) and ten more or less aquiferous formations, almost all of which are partially composed of carbonates

22
(including the unconsolidated carbonate gravels forming the shallow alluvial in the Wadis). However, the recharge   transpiration, undergoing a kind of summer dormancy.

32
As described before and applied in the modelling code of the SM-saturation excess and percolation model, the rate the maximum RC-values (instead, the cliff-forming u-UBK formation reaches the maximum here). This is due to 2 the fact that, from a land use and land cover point of view, these two formations had to be grouped into class II of 3 recharge potential (see Table 3), because here, besides the extended grass-and scrub lands, olive groves dominate 4 on the cultivated terraces of l-LBK and on the plains of Jerusalem formation (see LU/LC map, Fig. 2). The other,  (Table 3).

13
Note again that Wadi Natuf comprises of a main part belonging to the WAB, a smaller Eastern portion (in the 14 mountains) belonging to the groundwater catchment of the EAB and reduced outcrop areas, older than and 15 stratigraphically below the bottom formations of the regional Lower Aquifer in both, WAB and EAB. Table 5 16 documents the total recharge in Wadi Natuf (as well as that of the WAB portion only, in brackets and blue colour).

17
The resulting overall area recharge coefficient for the entirety of Wadi Natuf ranges between 39.4 % and 46.1 %,

18
slightly higher for the WAB portion (44.2 % as mean value of the three groups). As can be noted, despite the   Table   25 A1.

23
This study went a step furthercombining deductive (conceptual) and inductive (empirical) approaches to determine 24 spatial variations in groundwater recharge, based on qualitative (dimensionless) and on measured quantitative basin 25 observations alike. Our distribution into distinct classes of recharge potential, we would like to stress here again, was 26 an act of attribution and deduction; it was however, firmly grounded in general physical laws, such as permeability  Table D1. foundby order of importanceprecipitation, soil texture and vegetation cover to be the most meaningful proxies.
2 Therefore, PUB literature concluded that it is necessary to separately control the different main processes at work 3 rather than simply trying to optimise the exact quantification of employed parameters by ever more sophisticated 4 mathematical models. Such multicollinearity of physical expressions was also clearly observed in Wadi Natuf.

5
This is why we tried to avoid problems of multicollinearity and equifinality by testing three conceptual approaches

19
As already mentioned, and by contrast to most earlier studies in the WAB, the focus of our approach was clearly 20 the spatial, not temporal distribution and variability of recharge. This work was based on two assumptions: a) that   (Table 2) and soil depth (Table D1 in

25
In addition, our results confirmed that the temporal distribution of precipitationusually as events of several days 26 durationstrongly affects the percolation rates; a modelling frequency of daily steps was found appropriate under 27 the particular climatic conditions of the WAB recharge areas in the Eastern Mediterranean mountainsides.

29
As the main aim of our research, we thereby obtained a detailed differentiation of the spatial distribution of 30 recharge with formation-specific recharge coefficients for all formations in Wadi Natuf, which is a representative 31 catchment for the recharge area of the WAB. The results of our three-way conceptual analysis and attribution 32 seemed to suggest that indeed, slightly different results of overall recharge rates follow from the three approaches.

33
However, the relative closeness of the three results, e.g. the total WAB recharge in Wadi Natuf of 24, 26 and 28 34 mcm/a, respectively, did suggest that each of the three independent transfer procedures between basin form and 35 response was a realistic representation of the processes at hand. In other words, instead of producing an apparently 36 precise figure for groundwater recharge, our analysis resulted in a less "exact" but more robust realistic and 37 nonetheless close range of recharge quantifications.  Table A1).

26
On the side of spatial differentiation and given the lack of existing hydrological measurements, our approach 27 followed the three-way compromise prescribed by PUB (Beven and Kirkby, 1979) between the advantages of 28 model simplicity, the complex representation of spatial variability of hydrological basin response and the economic 29 limitations on field parameter measurement. This was done by applying the simple cooking recipe of Sawicz et al.