#### 2.1. Study Area

This study was conducted in the Peristerona Watershed in Cyprus. The Peristerona River has an ephemeral water flow and is located on the north-eastern hill slope of the Troodos Mountains (

Figure 1). In its upstream part, the elevation ranges between 1540 and 900 m, with a mean local slope higher than 40% and a mean average precipitation (1980 to 2010) of 750 mm [

19]. The geology of this area is dominated by intrusive rocks (sheeted dykes-Diabase), with smaller areas of cumulate rocks (gabbro and plagiogranites), of the Troodos ophiolitic sequence [

20,

21]. Upstream, the land is covered mostly with sclerophylous forests and mountain agriculture on terraces retained by dry-stone walls. Many of these terraces are abandoned with collapsed walls, and options for rehabilitation are being investigated [

22]. The midstream area ranges between 900 and 500 m above sea level (a.s.l.). The geological formations consist of intrusive rocks of the basal group (over 50% dykes with screens of pillow lavas) and are mainly covered by pine forests, with an annual precipitation of 405 mm. The foothills of the Troodos Mountains are formed by pillow lavas (olivine, pyroxene, and phyric lavas) and outcrops of sedimentary formations (hydrothermal and deep water sediments: umbers, shales, and mudstones) with occasional pockets of highly saline groundwater. The elevation ranges between 500 to 300 m a.s.l., with mean local slope of 20% and 360 mm annual precipitation. Within the downstream Mesaoria plain, the watershed narrows, and the mean local slope is lower than 8%, with an average precipitation of 270 mm. The geology mainly consists of sedimentary formations (mostly alluvium-colluvium: sands, silts, clays, and gravel) from the Pleistocene and the Holocene, which form a shallow unconfined groundwater system in connection with the riverbed [

23]. The Peristerona River recharges the alluvium and the sedimentary Central and Western Mesaoria Aquifer (

Figure 1), which is the largest and most important groundwater reservoir in Cyprus [

24]. Karydas et al. [

25], who applied the empirical erosion model G2 to the Republic of Cyprus, found soil loss rates exceeding 20 tons hectare

^{−1} in the Peristerona watershed.

The study transect can be divided into two zones (

Figure 1): the transient zone and the check dam zone. The transient zone is where water flows without ponding, while the check dam zone is where the inflowing water flow ponds and flows out of the check dam reservoir. Seven groundwater recharge check dams have been constructed along the Peristerona River on sedimentary formations in the Mesaoria plain. The most upstream check dam (named Orounda check dam) was selected for this research (

Figure 1). This structure was completed in October 2011, while the other six were constructed between 1980 and 1990 (not shown).

#### 2.1.1. Transient Zone

The river segment from Panagia Bridge (PB) to Orounda Bridge (OB) is 11 km long and is referred to as the transient zone in this study (

Figure 1). The geology of the transient zone can be divided into two sections: the first section is located between PB and Alluvial Start (AS), with a length of 7.8 km, and its bedrock is composed of intrusive and volcanic rocks. The second section is located between AS and OB, with a length of 3.2 km, and its bedrock consists of alluvium and colluvium. The watershed area up to PB is 77 km

^{2}, up to AS 98 km

^{2}, and up to OB 105 km

^{2}.

#### 2.1.2. Check Dam Zone

Orounda Bridge (OB) is the entrance of the check dam zone, and the Orounda check dam spillway (OC) is the exit. This structure is made of gabions. The check dam, with its reservoir (a) and a cross-section of the check dam (b), is presented in

Figure 2. The check dam wall has five concrete pipes with 0.5 m diameters. The elevation of the inlet of the pipes varies up to 10 cm, and the slope ranges between −0.6% and 0.3%. The spillway is constructed on top of the pipes and is 30 m long (the pipe spacing is about 6 m), 9.5 m wide, and 1 m high. The check dam is constructed on an alluvial riverbed, characterized by coarse gravel, cobbles, and few boulders. A nearby profile borehole (see

Figure 1 for location) shows that the aquifer consist of boulders and gravel to a depth of 6 m, silty sand with fine gravel from 6 to 18 m, fine gravel with sandy marl from 18 to 21 m, sand from 21 to 52 m, and marl from 51 to 334 m (Cyprus Geological Survey Department).

#### 2.2. Field Monitoring

The Water Development Department of Cyprus has measured streamflow continuously at PB with a weir since 1960. In addition to the continuous measurements, instantaneous streamflow measurements were made twice per week. The streamflow velocity was measured with an electromagnetic flowmeter (OTT MF pro; Kempten, Germany) at 0.6 of the depth and multiplied by the cross-sectional area of the stream water profile to obtain discharge (m

^{3} s

^{−1}) using the mid-section method [

27]. The distances between the stations (vertical sections) in each profile were such that no individual station contains more than 10% of the total discharge. The maximum observed streamwater profile was 0.7 m deep and 11 m wide. Measurements were made during the 2014 to 2015 streamflow season (December 2014 to June 2015) at two locations (AS and OB). Between these points, two irrigation channels, one on the east bank and one on the west bank, divert some of the stream water to the downstream agricultural fields. Twice weekly measurements of flow in these channels were made to estimate the amount of water diverted. Diversions to the irrigation canals were irregular. Also, sometimes water was released back to the streambed downstream from the canal inlets and before OB.

For the check dam zone, flow measurements were made at OB and OC (

Figure 1) approximately twice per week during 2014 to 2015. In addition, the groundwater levels in two wells were measured approximately twice per week. The well locations are presented in

Figure 1. Detailed topographic mapping (scale of 1:1000) of the check dam area was made, when the river had no water in Summer 2013. Differential Global Positioning System (DGPS) and a total station (Leica, Heerbrugg, Switzerland) were employed. A Triangular Irregular Network (TIN) model was created from the acquired points using ArcGIS (ESRI, Redlands, CA, USA). With known dimensions, relations between water inflow, depth of the water in the check dam reservoir (h), and open water surface area (A) were established.

The volume and mass of sediment in the ponding area were measured in Summer 2013 and 2015. The depth of the sediment layer was measured with utility poles. Bulk density samples (kg m^{−3}) of the sediment profile were collected and averaged to compute the sediment mass from 24 locations.

#### 2.3. Water Balance Computations

For the transient zone, linear regression relations were derived to link the continuous flow measurements at PB to the instantaneous flow measurements at AS and OB. The obtained relations were then used to estimate the daily flow at AS and OB for the period from 2011/2012 to 2014/2015 (hydrological years). Hydrological years are considered from 1st September to 31th August. The difference in the total flow between the two measurement points was considered groundwater recharge (above zero) or discharge (below zero). An estimate of the evaporation from the transient zone was made with Equation (2) presented below, using data from nearby stations and multiplying it with the surface area of the stream water, estimated from satellite images.

To estimate the groundwater recharge from the check dam, water balance calculations, the main components of which are presented in

Figure 3 and Equation (1), were made. The daily check dam recharge was calculated as:

where

Q_{r} is the recharge from the reservoir area,

Q_{in} is the water inflow,

Q_{out} is the water outflow,

Q_{e} is the evaporation, and

V_{d} and

V_{d−}_{1} are the volumes of water stored in the check dam reservoir area on the current and previous day, respectively. All units are in m

^{3}.

The inflow (

Q_{in}) was estimated with a linear regression model. Evaporation was calculated using an equation that is a combination of the mass-transfer equation introduced by Harbeck [

28] (suitable for lakes with a surface area (A) in the range of 50 m < A

^{0.5} < 100 km located in arid environments) and the equation of Penman and Priestley-Taylor modified by de Bruin [

29]:

where

A is the area of surface water (m

^{2}),

γ is the psychometric coefficient (-),

e_{s} is the mean saturation vapour pressure of the air (kPa),

e is the actual vapour pressure of the air (kPa), Δ is the slope of the vapour pressure curve (kPa C

^{0−1}), and

α is the Priestley-Taylor coefficient defined as:

where unitless

β is the ratio of sensible to latent heat flux with a default value of 0.6 [

30]. The value of the function of wind speed (

f(

u)) is derived from the mass-transfer equation introduced by Harbeck [

28]:

where

u is a wind speed 2 m above ground (m s

^{−1}). Daily meteorological data is obtained from nearby stations.

Actual measurements of flow at OB and OC and measurements of

h (the water height in the check dam) were used to relate, through regression models,

h to

V_{d}(

h),

Q_{out}(

h), and

Q_{r}(

h). To derive

Q_{r}(

h), water balance calculations were performed only with actual measurements. The derived functions were used to calculate the daily potential

Q_{out} and

Q_{r}, which set the daily upper limits. However, the actual

Q_{out} and

Q_{r} values can be lower than the estimated

Q_{out}(

h) and

Q_{r}(

h) because the rate of the change in

h throughout a day is ignored. In order to average this change and reduce the error in the estimates, the

h value in the middle of a day (

h_{mid}) is taken for the calculations;

h_{mid} is the

h of the total water at the end of the previous day in the reservoir plus half of the

Q_{in} of the current day. Summarizing, daily

Q_{r} and

Q_{out} were calculated as:

where

Q_{e}(

h_{mid}) is the volume of water evaporated from an open water reservoir area given a water height

h_{mid}, and

V_{p}_{1} is the volume of the water (m

^{3}) up to the bottom of the pipes.

Considering the amount of annual rainfall and the dimensions of the check dam and the spillway, it is assumed that when the water level reaches an elevation above the spillway bottom, all overflow is exhausted in a single day. Therefore, for V_{d}, Q_{e}, and Q_{r} calculations, h is limited up to the spillway bottom elevation. The calculations were performed with Microsoft Excel^{®} (Redmond, WA, USA).

#### 2.4. Uncertainty Analyses

Uncertainty is presented by calculating the 90% prediction intervals of the estimated values [

31]. These prediction intervals are referred to as uncertainty intervals and are expected to cover measurement errors. Errors in the discharge measurement with an electromagnetic flowmeter are computed as ±6%. This value is the average calculated from six stations (single depth acquisition) with a declared accuracy of the OTT flowmeter of 2% [

27,

32]. All measured streamflow values were less than 6 m

^{3} s

^{−1}. In the literature, comparable error values have been reported for similar measurement ranges [

33]. For the weir measurements, errors are also assumed to be ±6% [

34]. The uncertainty intervals of the evaporation estimates were considered ±5%, based on literature [

35].

The upper and lower bounds of the uncertainty intervals were calculated both for the transient flow and the check dam zone. In addition to the model results with the best estimates (BE) of the components, the water balance was recalculated with the upper and lower bound uncertainty interval values of each component. The aim was to obtain maximum recharge (

Q_{r_max}) and minimum recharge (

Q_{r_min}) estimates as follows:

where

Q_{r_min},

Q_{in}_{_min},

Q_{out_min}, and

Q_{e_min} are the lower bounds and

Q_{r_max},

Q_{in}_{_max},

Q_{out_max}, and

Q_{e_max} are the upper bounds of the uncertainty intervals. Negative lower bounds were set to zero.