Introduction

Groundwater recharge is an essential component of the water cycle, but it is difficult to quantify. Among the methods in the recharge estimation toolbox, tracer-based approaches are very popular. The recharge rate compilations by Crosbie et al. (2010, 4,386 Australian recharge estimates) and Moeck et al. (2020; >5,000 values globally) are dominated by chemical tracer methods (~80–90% of the values), especially the chloride mass balance method (CMB; Anderson 1945; Allison and Hughes 1978). In the Australian dataset, for instance, ~86% of the estimates are CMB-derived.

The underlying principle is that the conservative ion chloride, originating from atmospheric input, is concentrated by evapotranspiration (ET) so that pore water below the zero-flux plane (ZFP) and eventually groundwater is enriched in chloride. The degree of this evaporative concentration does not only reflect ET, but also its counterpart, groundwater recharge (R). Associated prerequisites are that (1) all chloride is meteoric (no road salt, no leaching from aquifer material, etc.), (2) there is no net change in chloride storage (no plant uptake, no plant cover alteration, no mineral precipitation/dissolution, no sorption/desorption), (3) there is no recycling or concentration of chloride in the subsurface, and (4) there is no (unmeasured) runoff (Allison and Hughes 1978; Edmunds et al. 1988; Wood, 1999). Most studies aim at long-term average recharge values via a steady-state CMB, implying that the system must be in equilibrium (no major change in climate and chloride flux).

Under these conditions, mass conservation dictates that the chloride flux at the surface equals the flux in the subsurface (below ZFP). In the literature, this mass balance, solved for recharge, comes in two main flavors. In the detailed version, dry chloride deposition (e.g., by settling dust) is included explicitly:

$$R=\frac{F_{\textrm{p}}+{F}_{\textrm{d}}}{c_{\textrm{ssf}}}=\frac{P\bullet {c}_{\textrm{p}}+{F}_{\textrm{d}}}{c_{\textrm{ssf}}}$$
(1)

where R is the mean annual recharge, Fp and Fd are the mean annual chloride fluxes via precipitation (wet deposition) and dry deposition, respectively, cssf is the mean chloride concentration in the subsurface (groundwater or pore water below the ZFP), P is the mean annual precipitation, and cp is its chloride concentration (Eq. sensu Edmunds et al. 1988; Scanlon 2010).

The second version is simplified and drops the Fd term:

$$R=\frac{P\bullet {c}_{\textrm{p}}}{c_{\textrm{ssf}}}$$
(2)

When presenting this version, many authors assume that the precipitation samples yielding cp are bulk samples. In this case, collectors are exposed permanently and hence include dry-deposited chloride (Alcalá and Custodio 2008; Appelo and Postma 2004), which dissolves in collected rainwater. This aspect is often made clear when defining the variables used in the equation (Cook et al. 1994; Gee et al. 2005; Wood and Sanford 1995). Occasionally, however, papers define cp as chloride in precipitation/rainfall without mentioning dry deposition or bulk sampling (Al-Ahmadi and El-Fiky 2009; Herczeg and Leaney 2010; Wood 1999) and one may wonder why. Did the authors tacitly assume bulk sampling and trusted that the reader is aware of this aspect? Are the authors unaware of the potential relevance of dry deposition and this component got simply “lost on the way”? Or was this a conscious decision and the authors think that dry deposition is irrelevant, at least in their study area?

Does dry chloride deposition matter?

Apparently, there is no consensus on this question and usually study area characteristics play a role (Appelo and Postma 2004), especially distance to coast (Bresciani et al. 2014; Hutton 1976). Some authors state that there is no dry deposition in their study area, without giving further details (Subyani and Şen 2006). Others neglect dry deposition based on the long distance to the coast and a lack of data (Kisiki et al. 2022). Maréchal et al. (2011) do not consider dry deposition based on corresponding analyses. Besides, a number of authors justify their decision by arguing that “dry deposition as aerosols is in steady-state with [mobilization from] the land surface” (Jin et al. 2015; i.e., deposition = remobilization), at least during the dry season (Edmunds and Gaye 1994; Tewolde et al. 2019). By contrast, in other studies, it is estimated that dry chloride deposition is equal to wet deposition (Szilagyi et al. 2011) or even greater (Murphy et al. 1996). To make matters worse, vegetation has been reported to be a complicating factor because it can enhance chloride deposition by capturing aerosols and droplets. Later, these can be washed off by rain, with a significant effect on CMB (Bresciani et al. 2014; Deng et al. 2013; Eriksson and Khunakasem 1969; and references therein). For perspective, global average values can be considered. Graedel and Keene (1996) assume the dry and wet fluxes from the troposphere to the pedosphere/cryosphere to be equal, and Möller (1990) estimates that the dry/wet ratio is around two over the continents. These numbers suggest that dry deposition matters on a global scale, but the question to what extent this holds true for a specific study area cannot be readily answered.

The method’s popularity and apparent simplicity

The popularity of the CMB method is partly due to the fact that it becomes more precise when recharge is low (Gee et al. 2005)—a feature not shared with other methods (Scanlon et al. 2002). Hence, it is often used in (semi-)arid areas with little replenishment (MacDonald et al. 2021; Scanlon 2010; Scanlon et al. 2006). Interestingly, it is also frequently deemed a low-cost (Edmunds and Gaye 1994; Wood 1999) and simple method (Gee et al. 2005; Subyani and Şen 2006). When looking at the equation (especially Eq. 2), one may get this impression—and of course the principle is indeed straightforward. Yet, on closer examination, some of this perceived simplicity vanishes. The largest source of error is the chloride input (Cook et al. 1994; Wood and Sanford 1995), perhaps especially the dry deposition (Tewolde et al. 2019).

Bulk sampling is thought to reflect total deposition, particularly when installing the collector(s) at the start of the dry season, as recommended by Weaver and Talma (2005). However, also this technique is not without flaws. Bulk collectors are traps, largely preventing remobilization of particles by wind, while such remobilization may happen naturally at the soil surface (Scanlon 2010). Hence, the method could overestimate deposition (Dettinger 1989), but the magnitude of this effect is unknown (Bresciani et al. 2014).

Properly constraining chloride input and capturing its temporal heterogeneity requires a decent number of samples. Most CMB studies target long-term recharge over the residence time of the sampled groundwater and the assumption of steady-state chloride deposition may be a source of uncertainty (Crosbie and Rachakonda 2021). To at least capture modern fluctuations in chloride input (Davies and Crosbie 2018, and references therein), long-term monitoring endeavors are needed (Scanlon et al. 2006). Hence, taking only a few wet samples is a risky business, because rain chemistry can vary between (and within) events (Michelsen et al., 2015) and dry deposition is excluded. Chloride input is also spatially variable, particularly in coastal areas (Alcalá and Custodio 2008; Davies and Crosbie 2018; Hutton 1976; Ordens et al. 2012), where distance to sea and elevation but also the terrain aspect and slope play a role (Guan et al. 2010). Unfortunately, corresponding monitoring networks (e.g., NADP 2023; INDAAF 2021) are rare and chloride (bulk) deposition maps are only available for selected regions (Alcalá and Custodio 2008; Davies and Crosbie 2018). In some areas, the creation of such maps (by interpolation or algorithms considering distance to sea; Hutton 1976) can be complicated by inland salt sources (Crosbie and Rachakonda 2021) such as salt pans (Schulz et al. 2015), which are likely to have an impact on rain chemistry (Michelsen et al. 2015) and dry deposition (Dettinger 1989). To obtain an overview of how researchers deal with some of the listed complications, a systematic literature survey was conducted.

Literature survey

The Web of Science (WoS) database was screened for CMB articles (details provided in the electronic supplementary material, ESM). Aiming at a reasonable sub-sample, the regional focus was on Africa and the Middle East, where the CMB method is particularly popular. One of the obtained 67 articles reports on three sub-studies (different countries), so the total number of studies is considered 69 (Table S1 in the ESM). The papers were searched for information on (1) the applied method to quantify atmospheric chloride input, (2) the corresponding sample number, and (3) the used mean chloride concentration.

The methods sections differ greatly in their level of detail. While some clearly mention the applied sampling strategy, a number of papers do not explicitly state the used method. In such cases, an attempt was made to derive information from keywords, related statements, and context. Where this attempt failed, the method was considered unclear.

The gathered data indicate that 17 studies (about 25%) rely on wet deposition sampling (Fig. 1, Table S2 in the ESM), i.e., the authors seem to ignore a potential contribution by dry deposition. Three studies consider wet and dry deposition separately. In 17 cases (about 25%), bulk sampling was used. In four studies, the authors report other approaches, e.g., mixed sampling strategies (split symbols in Fig. 1). Finally, in 28 cases (nearly 41%), the method remains unclear. In some of these studies, data from the (grey) literature was used, without specifying the corresponding methodology, at times even without mentioning a concrete reference. Multiple studies report that “rainfall samples” were collected, but without making clear how. Was the sampling device only exposed during the rain event (reflecting wet deposition) or was it exposed permanently (implying bulk sampling)? This distinction is important, since the two approaches may yield very different input values and hence recharge estimates.

Fig. 1
figure 1

Map showing the encountered chloride input sampling methods and the corresponding sample numbers in Africa and the Middle East. Hot (semi-)arid climate comprises the Köppen-Geiger classes BWh and BSh (Beck et al. 2018). Salty soil represents Solonetz and Solonchaks (Fischer et al. 2008) as well as salt pans (Schulz et al. 2015). Country codes are defined in the United Nations (2023) website https://unstats.un.org/unsd/methodology/m49/

There is no systematic spatial pattern in applying the various approaches (Fig. 1). Wet deposition was monitored at the coast, but also inland, and the same applies to bulk deposition. Studies with unclear methodology do not cluster regionally either.

This apparent inconsistency in sampling approaches is also reflected by the reported number of collected/considered samples (2–359; Fig. 1; Table S3 in the ESM). In 13 cases, <10 samples were used to characterize chloride input, with two being the lowest sample number. In 14 studies, the sample number was between 10 and 50. In eight cases, >50 samples were reported. The number remained unclear in 34 cases (nearly 50%). These values are hard to evaluate. First, there is no consensus on the required sample number (or monitoring period), which also depends on the studied system. Second, differing sampling strategies make a direct comparison challenging—a monthly bulk sample carries a different weight than a wet-only sample of an individual rain event. Yet, reporting the sample number at all (and the temporal/seasonal distribution) may seem like a reasonable request.

The compiled data include chloride concentrations of the precipitation samples (0.08–61.4 mg/L; Table S1 in the ESM) and hence enable plots considering distances between study areas and potential chloride sources (Fig. 2). Detailed interpretations are hampered by several factors (e.g., continental scale, differing and partly undocumented methods). Nevertheless, the plots seem to show a tendency towards elevated concentrations near (1) the sea or (2) salt sources in general. Interestingly, Fig. 2b) exhibits fewer deviations (at >100 km) from a commonly assumed exponential decay pattern, possibly implying a benefit of considering inland salt sources. The maximum concentration of 61.4 mg/L (>200 km from apparent salt sources; unclear sampling method) is a rather surprising outlier in both subplots.

Fig. 2
figure 2

Relation between chloride concentration in atmospheric input and distance to the a sea and b closest apparent salt source (sea or terrestrial salt source). Seven studies do not mention the used chloride concentration and therefore could not be shown

Finally, it is also noteworthy that most papers do not provide details on sampler designs, site characteristics, mounting heights, etc. In view of studies on height- and sampler-dependent dust collection efficiencies by the atmospheric sciences community (e.g., Goossens and Rajot 2008; Waza et al. 2019), this is somewhat surprising. Whether efforts were made to reduce evaporation from the collection vessel (e.g., by paraffin oil; Bresciani et al. 2014; Deng et al. 2013) is usually not reported either.

Conclusions and a call for transparency

The survey has shown that the encountered studies follow rather different recipes when it comes to quantifying the chloride input for the CMB method. This applies to the general sampling concept (wet, bulk, etc.), sample numbers (2–359), and temporal aspects (single events vs. yearly bulk samples; Table S1 in the ESM). These different approaches probably reflect conscious decisions and one could argue that the authors know their study areas best. Yet, one should not discard dry deposition lightly, just because it was ignored in some other paper relying on wet deposition during a handful of rain events. The possibility of dry deposition should also be considered in inland study areas, especially when terrestrial salt sources may be present. Bulk sampling is thought to reflect total deposition, but also this technique could have downsides. These collectors trap deposited particles, although remobilization by wind may happen naturally at the soil surface. Yet, the degree of the potentially resulting overestimation remains unknown (Bresciani et al. 2014; Dettinger 1989).

Considering all the complications that are relevant for recharge estimations, but also other solute mass balances (Cartwright et al. 2013; Wood 2019), this essay does not provide easy answers. Instead, it is intended to highlight selected issues and represents a call for more transparency in CMB studies, with respect to applied methods and underlying reasons. Frankly, the “maybe” in this essay’s title should not be an option.