Localisation and temporal variability of groundwater discharge into the Dead Sea using thermal satellite data

The authors focus on areas where groundwater discharge can most likely be expected to investigate the influence of groundwater discharge on the thermal distribution of the DS. Those discharge areas are indicated in a previous study during which Mallast et al. (2011) derived groundwater flow-paths. Extending the flow-paths towards the DS reveals six areas on each coast of the DS (Figs. 3, 4) where proposed flow-paths intersect the shoreline and hence where groundwater inflow occurrence is most likely. As the surface-runoff was excluded during the pre-processing step, all thermal anomalies with low ∆T values are assumed to reflect groundwater affected areas (GAA). The visually identifiable extent thereby directly mirrors groundwater discharge volume at the respective location (Mallast et al. 2013). The ∆T value most likely is a measure for the emergence depth, ranging from terrestrial/very shallow to deep spring discharge. Due to the intermittent character of the groundwater discharge and the rather coarse GSD of the SST data, especially small GAAs can visually disappear over different biennial series. To present the most complete result, the authors choose to depict the biennial series, which best reflects the groundwater inflow locations. Additionally, GAAs are classified by continuous coast-parallel extent into: (1) large GAAs (>1 km in length parallel to the coast), (2) small GAAs/single spots (<1 km in length parallel to the coast) and (3) assumed/diffuse GAAs to distinguish discharge magnitudes per GAA. All sites with GAAs are subsequently coded with the area where they occur and a running number.

From the observation of ∆T data along the entire coast, the authors identify two locations representing large GAAs. Both are located at the western coast of the DS (Fig. 3). The first site (A1), located in discharge area A, exhibits ∆T values of 6.9–8.5 °C and a continuous area of ~3.4 km length parallel to the coast and ~300 m orthogonal to it. The second site (C7), located in discharge area C, is characterised by small ∆T values of 6.3–8.5 °C. The extent parallel to the coast is ~2 km with an orthogonal extent of ~200 m. A1 and C7 correspond to the spring areas of Ein Feshkha and Samar, respectively, which are known to have several terrestrial springs (white dots) discharging 82–94 × 10⁶ m³ a⁻¹ for Ein Feshkha and 19–35 × 10⁶ m³ a⁻¹ for Samar (note that the numbers originate from the IHS measurements and represent the minimum and maximum numbers between 2004 and 2011) (Guttman 2000; Laronne Ben-Itzhak and Gvirtzman 2005; IHS 2012). The fact that both thermally detected sites spatially correspond to locations of in situ spring measurements distinctively validates thermally inferred groundwater discharge sites. This observation is further supported by similar ∆T values derived for areas with a stabilised SST through groundwater (Mallast et al. 2013). In fact, as the relationship between in situ groundwater measurements and thermal anomalies appears to be valid, the authors deduce that at both northern ends of site (A1) and (C7) further groundwater discharge occurs. No in situ measurements were conducted at these locations, but ∆T slightly increases to values <9.4 °C over ambient groundwater uninfluenced temperatures (>10 °C). The authors presume that the observation of low ∆T values possibly shows an additional groundwater discharge location of possibly submarine origin.

All discharge areas contain small or single spot GAAs with similar ∆T values of 6.4–8.5 °C as the afore-mentioned large GAAs on the western coast (Fig. 3). Thermally indicated sites B6, D10 and D11 spatially match in situ measurements of spring discharge (note that B6 corresponds to Kane spring area while D10 and D11 correspond to the Qedem spring area). This verifies the spatial information of inferred thermal anomalies for small GAAs also. The GAA extents (B6 and D11: length ~400 m/orthogonal extent ~150 m; D10 length 150 m/orthogonal extent 90 m) suggest less discharge compared to the afore-mentioned group and is verified by in situ measurements (B6 ~ 6–9 × 10⁶ m³ a⁻¹ (IHS 2012 for 2004 to 2005); D11 ~ 10 × 10⁶ m³ a⁻¹ (Shalev et al. 2007 for 2005); D10 0.6 × 10⁶ m³ a⁻¹ (own measurements in 2011). The latter discharge volume reflects only a minimum since several field observations exhibited at least three submarine springs (Fig. 1d) with an estimated sea surface diameter of ~10 m that probably contribute significantly to the thermally indicated extent of site B6.

Further small GAAs are located at sites A2, B5 E14, E15 and F16. All sites generally exhibit smaller extents except for site A2 that is similar in extent to site D10. Hence, taking site D10 as reference, a similar volume can be assumed for site A2 and that the remaining sites (B5, E14, E15, F16) groundwater discharge is lower by an order of magnitude. On the other hand, the groundwater discharge appears to be mainly of terrestrial/shallow origin, applicable to all of these GAAs, since ∆T values (8.1–9.0 °C) are similar to afore-mentioned ∆T values induced by terrestrial springs.

Despite being noticeable from the ∆T data, these locations remain unconfirmed as neither terrestrial springs are reported nor could field observations in the years 2010–2011 directly confirm their existence. Sites E14 and E15 could correspond to the Ein Yesha springs that are mentioned in Guttman (2000) and Vengosh et al. (1991). Site F16 seems also reasonable as Shalev et al. (2009) suggest a NE-orientated hydraulic gradient from the Ye’elim area towards the thermally indicated location at the northern DS basin.

On the eastern coast, all discharge areas except for area F show evidence of small GAAs/single spots. Descriptions of groundwater occurrence from literature are not as exact as on the western coast. This is most likely connected to the steeper coastal morphology and bathymetry which might predominantly result in submarine springs due to the induced increased hydraulic pressure conditions. Closson et al. (2010) and Akawwi et al. (2008) could prove the occurrence of groundwater discharge at some locations that spatially correspond to the thermally indicated small GAAs/single spots with ∆T values of 6.3–8.7 °C at sites G18, G19, H24, I26-27, J28-29, and J34, respectively (see white dots in Fig. 4). The spatial extents of site H24 (160 m parallel and 70 m orthogonal to the coast, respectively) are similar to site D10 on the western coast and thus a discharge volume of similar magnitude (>0.6 × 10⁶ m³ a⁻¹) is assumable. The two low ∆T value spots at site G19 have slightly smaller extents (100 m parallel and 70 m orthogonal to the coast). Consequently, the discharge volume appears to be slightly smaller than for site H24. Site H27 on the other hand has a spatial extent of 300 m parallel and 120 m orthogonal to the coast indicating a higher discharge volume. All three mentioned GAAs (G19, H24, H27) are characterised by ∆T values of 6.9–8.7 °C and a direct connection to the coast. Both aspects commend to a shallow or terrestrial emergence that applies for sites G18, I28 and at least the northern end of site I29 also. Most parts of sites I29 and J34 display similar ∆T values (7.5–8.4 °C) but do not have a direct connection to the coast, which suggests a deeper submarine emergence. For these locations, it is difficult to approximate discharge magnitudes, since neither the required thermodynamic mechanisms and parameters nor the exact emergence depth are known.

Moreover, it should be mentioned that site G17, which falls into the group of small GAA, probably does not reflect a GAA. Instead, it represents discharge from the Jordan River as the coordinates correspond to the location of its mouth.

The first group contains hardly visible GAAs. Sites B3, B4 and C9 are located on the western coast and I30-32 on the eastern coast. All GAAs of this group on the western coast are characterised by extents <4 pixel (pixel size: 30 × 30 m) and ∆T values of 8.9–9.5 °C emphasised by high ambient ∆T values (>10 °C). The slightly higher ∆T values might indicate low groundwater discharge. The small contrast to ambient SST represents an unfavourable condition. Thus, the authors suggest these sites to be allocated to “assumed GAAs” requiring further ground-truth data. This is in contrast to the GAAs on the eastern coast (sites I30-32). Here, GAAs are also characterised by similar ∆T values of 8.1–9.2 °C, exhibiting, however, larger connected extents (>6 pixel). Additionally, all of these sites are verified by already confirmed groundwater discharge locations (Akawwi et al. 2008). Hence, the authors suggest the eastern GAAs of this group to truly reflect groundwater discharge. The emergence is probably submarine with very low discharge magnitudes causing ∆T values of 8.5–9.5 °C.

The second group comprises heterogeneous GAAs with large extents (1–2 km parallel and ~500 m orthogonal to the coast, respectively). Site D12 on the western coast and sites G22, H25, and F38 on the eastern coast belong to this group. Akawwi et al. (2008) report a submarine groundwater inflow for site H25 that verifies our finding. The same applies to site F38, where extensive studies of Closson (2005) and Closson et al. (2010) report landslides, subsidence and sinkhole occurrences that directly or indirectly depend on shallow groundwater flow and dynamics. Both verifications indicate these GAAs to have a groundwater origin. The question remains why these GAAs display slightly increased and heterogeneous ∆T values (8.2–8.9 °C). The missing direct connection to the shore of all (except for site G22) indicates a submarine origin. At all locations (except for site F38), the bathymetry is unequally steep. These GAAs are located at the −450 m msl. isoline on both coasts (Fig. 1), which means a water depth of around <35 m (taking the DS level of −415 m msl. of 2001 and −417 m msl. of 2005). Ionescu et al. (2012) describe the occurrence of submarine springs up to a depth of ~30 m. Hence, from a hydrostatical perspective, this observation indicates these GAAs to be generally reasonable. The slightly higher ∆T values that characterise this group could have two origins. First, as the groundwater ascents due to different densities, a heat exchange between the DS water and the discharging and ascending groundwater occurs. This partly changes the original temperature of the groundwater to the DS water. Most of the ∆T data are recorded in summer months, during which the DS has a temperature of 32−35 °C. The resulting ∆T between DS water and groundwater amounts to 4−7 °C. The density driven buoyancy of the groundwater yields a velocity of ~0.4 m s⁻¹ and hence a travelling time of ~75 s for emerging water at a depth of 30 m (deepest observed submarine spring) (Ionescu et al. 2012; Mallast et al. 2013). Complete temperature equilibrium within the short travelling time is, therefore, unlikely. A certain temperature adaption, however, can be assumed.

Second and probably more decisive is the influence of different limno-physical mechanisms (e.g. currents, temperature diffusion, etc.) on the ascending groundwater. These mechanisms might influence the vertical flow causing the circular pattern at the sea surface to move in horizontal dimensions by some meters over a longer time period. The movement could be repeatedly observed in the field and may explain why springs with submarine origin are thermally unequally clearly visible on the sea surface compared to groundwater that discharges in shallow areas.

The third group encompasses GAAs that are characterised by a near-shore cluster of high ∆T with values of 10.5–12.2 °C, which differ from ambient SSTs. This is in contrast to the aforementioned analysed characteristic that steady groundwater discharge with large discharge volume stabilizes ambient SSTs that subsequently cause low ∆T values.

However, several of these GAAs are observed at both coasts (sites C8, E13, J33, J35 and J36) that match independent groundwater discharge observations and in situ measurement locations (Figs. 3, 4). It is known from the IHS measurements in 11/2004 that at the location of site C8 several springs exist with minor discharge volumes of <0.015 × 10⁶ m³ a⁻¹, while only two springs exhibit discharge volumes of 0.3–0.5 × 10⁶ m³ a⁻¹, respectively. Due to these very small discharge volumes that are distributed along a shore strip of >1 km and mostly along coarse grained wadi fans the authors hypothesise that these rather diffusely discharging springs with low flow dynamics (Fig. 5) and quantity stabilize only a small fringe along the DS coast. This fringe cannot be resolved by the coarse GSD of the SST data. In periods with increased discharge, due to, e.g. the flushing effect, the thermally stabilised area increases and is captured by the SST data subsequently, despite the coarse GSD. As a consequence, the alternation of both discharge situations could lead to higher observed ∆T values.

Akawwi et al. (2008) reported groundwater discharge locations for two of the GAAs on the eastern coast (sites J35 and J36) that validate our findings. Hence, the authors suggest that this thermal pattern potentially indicates groundwater discharge, with mostly diffuse emergence and small discharge magnitudes.

Hence, on the positive side the authors conclude that

Also noticeable are 8 IHS measurement locations between sites B5, B6 and 8 own spring measurements in 2011 south of site E12, where neither low nor distinctive high ∆T values at the same longitude are present. Most of these show discharge volumes of 0.02–0.18 × 10⁶ m³ a⁻¹. Only two springs in discharge area B and two in discharge area E exhibit values of 0.23–0.41 × 10⁶ m³ a⁻¹. From this fact, the authors would argue that the applied thermal analysis based on coarse GSD satellite data has a clear limitation if it comes to the identification of discharge volumes with a magnitude of <0.4 × 10⁶ m³ a⁻¹. If the intention is to identify all groundwater inflow independent of quantity, the authors suggest using airborne thermal imaging as it is applied in the study of Akawwi et al. (2008). For the detection of larger discharge volumes, the applied multi-temporal thermal satellite data prove a successful applicability.

With the exact spring locations identified, the temporal discharge variability is being analysed for the years 2000 to 2011. Due to the variability in discharge volume, some discharge areas display no visually delimitable GAAs. Hence, the focus is set on GAAs that are continuously distinguishable throughout the entire investigation period. These GAAs are located in discharge areas A-C on the western coast and H and J on the eastern coast.

The SST data included in all series are recorded between March and October (Fig. 2). Hence, the results reflect the same hydrological time of the year and imply comparability between biennial series. Within each series, the ∆T values are Gaussian distributed where the groundwater caused lower ∆T values display below the 95 percentile of the normalised ∆T. This threshold and the constraint for the GAAs to have a minimum Euclidian distance of 120 m (4 pixels) to the coast are taken to delineate the GAAs in the respective series. Both parameters allow the determination of the temporal evolution of groundwater discharge quantitatively in which GAA extents are proportional related to groundwater discharge volume.

The GAA extent and hence the groundwater discharge derived from the ∆T data indicate a significant but unsteady decrease in all discharge areas during the observation span (Fig. 6). For discharge area A, which corresponds to the spring area of Ein Feshkha, the behaviour of the GAA extents exhibits a three-steps change. From 2000 to 2001 to 2002 to 2003, the GAA decreases from 2.41 to 0.68 km². It then remains constant for the subsequent 2004–2005 series (0.69 km²). The following two series again show a significant decrease to >0.14 km² before the GAA increases to 0.83 km² during 2010–2011 (Fig. 6). For the years 2004–2011, IHS spring measurements show an identical but less pronounced behaviour except for the last increase that will be discussed later. Mean total measured groundwater discharge decreases almost linearly from 71.0 × 10⁶ m³ a⁻¹ during 2004–2005 to 60.2 × 10⁶ m³ a⁻¹ during 2008–2009 series (Fig. 6).

Discharge area B/C (lower row in Fig. 6) shows a different behaviour. The changeover from 2000–2001 to 2002–2003 demonstrates a similar GAA decrease from 3.24 to 0.08 km² as in discharge area A (Fig. 6). The following series (2004–2005) unlike discharge area A exhibits an increase to 0.42 km². This surface remains stable during 2006–2007, but unlike discharge area A, it decreases to 0.05 km² until the 2008–2009 series. The last changeover to the 2010–2011 series behaves similarly as it exhibits an increase. Comparing the total GAA to the IHS spring measurements reveals an analogous behaviour. The mean total spring discharge (Q _Kane and Q _Feshkha accumulated) of 37.6 × 10⁶ m³ a⁻¹ during 2004–2005 is similar to 41.1 × 10⁶ m³ a⁻¹ during the 2006–2007 series, before it decreases to 30.4 × 10⁶ m³ a⁻¹ during 2008–2009 (Fig. 6). These numbers suggest that if one takes the IHS measured terrestrial spring discharge into account only the always reported rigid discharge ratio of 2:1 between Ein Feshkha (discharge area A) and Kane/Samar (discharge area B/C) is true. In contrast, our results that also include submarine springs suggest that the discharge ratio between both areas can vary inter-annually between 8:1 and 1:3 (Fig. 6).

There is a further interesting observation next to the variation in groundwater discharge rates. Our temporal comparison suggests variable groundwater discharge locations across time. In discharge area A, for example, an inflow is apparent along the entire bay structure during 2000–2001 (Fig. 6). During the following two series inflow only occurs along the northern and central part of the bay. Emergence gradually starts to disappear from the north over the subsequent two series leaving only inflow locations at the bay centre. Magal et al. (2010) report similar findings and describe the gradual drying out of northern springs at the Ein Feshkha Nature Reserve. In contrast, during the last series (2010–2011) the springs seem to reappear along the entire bay. The same phenomenon takes place in discharge area B/C where a shift of groundwater inflow locations from north to south is visible between 2004 and 2007. The dis- and reappearance of springs along this strip of discharge area B/C was also observed during several field observations during the years 2009–2011, which supports the results obtained from the multi-temporal thermal analysis.

The authors hypothesise that this change is not only a result of different recharge amounts and areas. It might additionally indicate the formation of new preferential flow-paths triggered by the lowering of the DS and the location of narrow fault- and folding zones described by Enzel et al. (2000). Groundwater potentials adjust to the dropping of the DS level. This could lead to the activation of a lower lying fault zone as preferential flow-path, which could slightly differ in orientation than overlying fault zones. Or, the internal folding of this zone governs the flow-direction and hence the discharge location.

The fact that the 2000–2001 and 2010–2011 series differ in appearance from the remaining series 2002–2003 until 2008–2009 is also striking. While ∆T data of the latter series show distinct discharge points that mostly correspond to high quantity spring discharge (see red points in Fig. 6), the 2000–2001 and 2010–2011 series show a spatially continuous discharge over large areas. This phenomenon is probably due to the flushing effect (Kiro et al. 2008) that will be discussed later.

Plotting the GAA of the respective series against the calculated recharge (Fig. 7) using the empirical formula given in Guttman (2000) (Appendix 3 in ESM) reveals a similar curve progression with a delayed response time of one series. This similarity is supported by high index of agreement (d) values (index varies between 0 and 1, where 1 indicates a perfect fit of curve progression and 0 indicates no fit (Willmott 1981)) of d = 0.71 for Ein Feshkha and d = 0.66 for Kane/Samar (Appendix 4 in ESM). Recharge curve and spring discharge from IHS measurements for Kane/Samar also display a conform behaviour. In contrast, the recorded Ein Feshkha groundwater discharge rate shows an almost linear decrease and hence a non-conform behaviour with respect to the average biennial recharge. This might indicate a different hydrological system with multiple input sources than only recharge from the Judean mountains.

The GAA curve and recharge curve are dissimilar during 2000–2001 and 2010–2011. It is striking that concurrently, the lake level drops at a maximum of 1.5 m and 1.6 m during these series. Both numbers are above the long-term average drop of 1 m. For this situation with above average drops of the lake level, Kiro et al. (2008) and Yechieli and Sivan (2011) propose an increased discharge due to the contribution of old brines (flushing effect) that could explain the larger GAAs.

Hence, the authors hypothesise that the larger GAAs during both periods are a consequence of the so far unobserved flushing effect. The flushing of old brines is connected to the availability of void spaces to temporarily store water. With the recently exposed Quaternary unconsolidated sandy and clayey lake sediment areas along the entire coastline, this possibility is given and in turn induces an increased discharge over large areas rather than at discrete points. Observations based on the thermal GAA indications show that both discharge areas A and B/C exhibit theses assumed continuous discharge appearance. This supports the hypothesis of an occurring partial flushing mainly during the 2000–2001 and 2010–2011 series.

Unlike at the western coast, the discharge areas H (Wadi Zarka Ma’in and Zara) and J (Wadi Mujib) along the eastern coast do not display a homogeneously decreasing trend. Instead, area H (upper row in Fig. 8) shows a sinusoidal course of decreasing and increasing GAA, whereas the GAA in area J (lower row Fig. 8) increases until 2004–2005 and decreases from thereon. Striking is the GAA in discharge area H in the 2000–2001 series that is with 0.6 km² significantly higher than GAAs of the subsequent series. The same series provides an indication that the flushing effect of old brines occurs on the eastern coast also. The reason is the continuous appearance of GAAs in contrast to remaining series with rather distinctive discharge locations analogously to the discharge areas A and B/C on the western coast during the same 2000–2001 series.

In discharge area J, higher ∆T values can be observed along the entire coastline. As this observation indicates diffuse discharge with minor volumes (see Sect. 4) and as it occurs over a continuous spatial extent along the coast, the authors assume this pattern to be caused by the flushing effect. The topography may be responsible for the low discharge quantities. Unlike the western coast, the eastern coast exhibits steep slopes that extend offshore, with generally non-existent or less developed Quaternary lake sediments (Closson et al. 2010). This, in turn, means less void spaces compared to the western coast, and hence lower quantities of groundwater which are temporarily stored and available for flushing from unconsolidated sediments.

Plotting the derived GAA against the calculated recharge depth (Fig. 9) with the empirical formula given in Guttman (2000) (Appendix 3 in ESM) reveals a similar curve progression and the same response time-shift of one series as on the western coast. The similarity is supported by goodness-of-fit values of d = 0.67 for area H and d = 0.86 for area J (Appendix 4 in ESM). These numbers correspond to those derived on the western coast and indicate that the recharge-GAA relationship is applicable. Figure 9 also reveals the flushing effect to occur in the earliest series 2000–2001 for discharge area H due to the above average drop of the DS level of 1.5 m. As observed on the western coast, a shift of discharge locations is also partly visible on the eastern coast at the example of discharge area J. There, the inflow varies from the tip of the Wadi Mujib fan to the north and to the south. Similar to the western coast, the hydro-geological system may change locally with the drop of the DS level establishing new flow-paths and inflow locations.

The congruency between curve progressions of recharge and GAA extents suggests a direct relationship. Hence, it once again confirms our hypothesis that it is possible to derive information on the temporal groundwater discharge behaviour using thermal satellite data (2013).

Following Ou et al. (2009), who proved a direct relationship between area of discharge plume (in the present case GAA) and discharge quantity, it appears to be possible to derive absolute volumetric quantities based upon the GAAs. However, for the present case, the authors consider the absolute quantification to be unfavourable due to the following reasons:

All points influence a reliable and concrete statement regarding absolute quantities. Hence, the authors suggest applying satellite data at unmonitored locations only semi-quantitatively as pursued during this study.

Our results reveal that