Changes In Soil Properties Across A Hydrological Gradient In Saladas From Northeast Spain: Implications For Soil Carbon Stocks, CO2 E ux And Microbial Communities In A Warming World

Andrew Thomas (  ant23@aber.ac.uk ) Aberystwyth University Penglais Campus: Aberystwyth University https://orcid.org/0000-0002-1360-1687 Stephen Tooth Aberystwyth University Penglais Campus: Aberystwyth University S. Lan Chinese Academy of Sciences Thomas Holt Aberystwyth University Penglais Campus: Aberystwyth University Ian Saunders Aberystwyth University Penglais Campus: Aberystwyth University Holly Tarren Aberystwyth University Penglais Campus: Aberystwyth University


Introduction
Throughout the world's drylands, there are numerous permanent and temporary wetlands, such as oodplains, playas and oases (Tooth and McCarthy, 2007). These wetlands in drylands are characterised by diverse hydroperiods (water inundation frequency, duration and depth) but are typically hotspots of biological activity and productivity compared to the surrounding landscape (Millennium Ecosystem Assessment, 2005; Ezcurra, 2006;Tooth et al., 2015). Nonetheless, the healthy functioning and possibly even existence of many wetlands in drylands is threatened by a combination of climate change and human disturbance that will lead to changes in moisture availability ( Xi et al., 2021). Changes in surface water and groundwater availability will be expressed through altered hydroperiods, which will impact on biological and chemical processes and characteristics, including those related to the carbon (C) cycle (Petrescu et al., 2015). Predicting the impacts of changes in moisture availability on the C cycle in wetlands in drylands is challenging but important because many contain large C stocks and may be signi cant sources and sinks of greenhouse gases (Nazaries et al., 2013).
Several distinctive characteristics of the C cycle in wetlands in drylands confound simple attempts to transfer data and concepts from better studied humid region wetlands (e.g. Mitsch and Gosselink, 2015). First, in some wetlands in drylands, the majority of organic carbon (OC) may be generated by bacteria, algae, and aquatic organisms rather than by vascular plants (Domínguez-Beisiegel et al., 2013a). Cyanobacterial photosynthesis has been shown, for example, to contribute OC to playa soils in the Makgadikgadi Basin in Botswana . Although relatively little is known about controls on cyanobacterial photosynthesis in the soils of saline inland wetlands, or in other types of wetlands in drylands, periods of net CO 2 uptake are likely to be short-lived and episodic (Ladrón de Guevara et al., 2014; Thomas et al., 2014) or seasonally controlled (Williams et al., 2014). Second, wetlands in drylands are also sites of inorganic carbon (IC) production and storage (Xie et al., 2009, Yates et al., 2013. Globally, terrestrial IC formation occurs largely in dryland soils, and may be as much as 0.152 Pg C yr − 1 , an amount similar to that buried in ocean sediments (Li et al., 2017). The generalised reaction that leads to IC formation involves CO 3 2− ions, derived from CO 2 dissolved in soil water, which combine with divalent metallic ions to form carbonates ( . Soil carbonate formation is of particular signi cance to the C cycle because carbonate is recalcitrant and less likely to be returned to the atmosphere as CO 2 than is OC. Third, the extreme variations in moisture availability typical of many wetlands in drylands will have a profound effect on microbial communities, greenhouse gas emissions, and ultimately C stocks. Soil microbial activity, and consequently CO 2 e ux, is typically found to be lower at high and low moisture extremes, and higher at intermediate levels where water and oxygen availability are optimal (Orchard and Cook, 1983;Schimel et al., 1999; Moyano et al., 2013).
Despite these previous ndings regarding the distinctive aspects of the C cycle in wetlands in drylands, large gaps in knowledge remain, particularly for saline inland wetlands (e.g. playas, pans, salt lakes). Yet saline inland wetlands are likely to be a signi cant source of CO 2 and play a signi cant role in regional and global C cycles (Duarte et al., 2008).
Indeed, while the IPCC Task Force on National Greenhouse Gas Inventories (IPCC, 2014b) noted that saline inland wetlands are important features in dryland landscapes worldwide, little information was available to assess their role as sources or sinks of greenhouse gases, and consequently the Task Force was unable to provide guidance for these wetland types.
As a contribution towards this critical knowledge gap, we exploit the contrasting hydroperiods of three adjacent playas, known locally as saladas, located near Alcañiz in northeast Spain (Fig. 1). The saladas are situated along a 3.5 km long hydrological gradient from wet to relatively dry and provide an opportunity to test hypotheses regarding how different moisture conditions affect critical aspects of the C cycle. For each salada, we characterised their hydroperiods, determined sediment chemistry (including stocks of OC and IC), calculated CO 2 e ux, and investigated soil microbial communities. Our rst hypothesis (H1) is that relatively wet soil moisture conditions will generate greater OC inputs than in drier soils, due to the controlling in uence of moisture on biological productivity (Schimel et al., 1999). The net effect will be to elevate OC stocks in comparison to drier saladas. The effect of moisture on soil IC is less certain and our second hypothesis (H2) is that IC stocks will be dependent on the dominant carbonate formation processes. If evaporative concentration of solutes is the dominant process, then the drier saladas will have greater soil IC stocks than wetter saladas. If, however, IC formation is largely due to the supply of cations in in owing waters and associated biomineralisation, then the wetter saladas will have higher soil IC stocks than drier saladas. Our third hypothesis (H3) is that very high soil moisture that is close to, or at saturation levels, will suppress microbial activity and CO 2 e ux (Moyano et al., 2013). Our nal hypothesis (H4) is that each of the three saladas will contain different soil microbial communities due to their different inundation periods and soil moisture content. The empirical data collected to test these hypotheses provides the basis for an improved conceptual understanding of the impact of climate or human-induced hydrological changes on the C cycle in the saladas. Ultimately, such knowledge may contribute to more effective management and protection of these important but threatened wetlands in drylands (c.f.

Study Region
The town of Alcañiz is located in the southern part of the central Ebro basin (also sometimes referred to as the Ebro depression) in the Teruel Province, northeast Spain. The Ebro River, located ~ 22 km to the north of Alcañiz, bisects the basin as it ows from the northwest towards the Mediterranean Sea in the southeast. The river receives much of its discharge from snowmelt and rain-fed tributaries originating in the Pyrenees to the north and maintains perennial ow but the central basin has one of the highest annual water de cits in Europe (Herrero and Snyder, 1997

Study saladas
Data were collected in May 2018 at Salada Grande, Salada Pequeña and Salada de la Jabonera de las Torrazas (hereafter shortened to Salada Jabonera), located ~ 5 km west of Alcañiz (Fig. 1a). The area covered by Salada Grande, Pequeña and Jabonera is ~ 1.24, 0.19 and 0.19 km 2 , respectively. The saladas have developed in an area of sub-horizontal Miocene siltstones, mudstones and ribbon or sheet sandstones and microconglomerates (sensu Friend et al., 1979). Preferential weathering of the less resistant argillaceous strata has tended to give rise to the depressions, leaving some of the more resistant calcareous sandstone and microconglomerate bodies in positive relief in the form of sinuous, inverted palaeochannels on the salada margins (Sánchez et al., 1998;Gutiérrez et al., 2005). All three saladas form part of a 7 km 2 Natura 2000 site (Saladas de Alcañiz, site code ES2420114), which was designated on account of being one of the most important endorheic areas on the Iberian Peninsula, with plant communities to adapted to saline conditions and a rich fauna linked to seasonal inundation (European Commission, no date). There is a topographic gradient from the surface of Salada Pequeña at ~ 359 m above sea level (masl), through Salada Grande at ~ 351 masl, and to Salada Jabonera at ~ 340 masl. Although topographically the highest, Pequeña is the wettest of the three saladas ( Fig. 1a-b). Shallow (< 0.5 m deep) but frequently prolonged standing water occurs over much of the salada surface, partly because of rainfall and runoff but also because of contributions from groundwater and a small stream that periodically in ows from the southwest (Fig. 1a). Salada Grande is drier and Salada Jabonera drier still, but both typically experience shallow standing water over at least part of their surfaces for some of the year as a result of rainfall and runoff, and possibly some limited groundwater contributions. In all three saladas, the ephemeral nature of the standing water, combined with the widespread dissolution of halite and gypsum and its subsequent evaporative concentration, leads to a high soil pH (Gutiérrez et al., 2005). Analysis of water extracts from saturated pastes of Salada Grande sediments indicates that soils are dispersive when wet but prone to pelletisation when desiccating, which promotes aeolian de ation during dry periods (Gutiérrez et al., 2005). The surfaces of Pequeña and Grande are mostly unvegetated, except at the fringes where soils are elevated above the salada surface by a few metres ( Fig. 1ab). Salada Jabonera, the driest of the saladas, has a patchy cover of vascular halophytic shrubs and grasses across the surface ( Fig. 1a- For each image, surface water area was manually digitised using ArcGIS 10.5. Mapping is deemed accurate to +/-1 pixel (20 m for Sentinel and 30 m for Landsat MNDWI) resulting in a maximum uncertainty of +/-12%, although in reality, uncertainty is much lower owing to the distinct boundaries between the water surface and adjacent bare soil surface in the true-colour composites.

Soil moisture and temperature
The moisture content of salada soils to 5 cm depth was recorded in a grid pattern every 10 m across the surface of each salada using a DeltaT SM150 probe (n = 100, 120 and 180 on Saladas Pequeña, Grande and Jabonera, respectively). The grids corresponded approximately to the area of the sampling sites for CO 2 e ux determination ( Fig.   1a), although additional soil moisture readings were taken with the DeltaT probe adjacent to the chambers used to determine CO 2 e ux (Sect. 2.6).

Soil physical and chemical properties
On each salada, four pits up to 0.5 m deep were excavated at equal distances within the area of the soil moisture sampling grid. From each pit, up to ve soil samples were collected from a range of depths (see Supplementary   Information), then air dried and bagged prior to analyses. H + ion activity was determined in a 1:2.5 soil-water mixture using a pH probe. Approximately 5 g of each sample was dried at 105°C to a constant weight and heated at 430°C for 16 hours in a mu e furnace to determine organic matter by loss-on-ignition. Bulk density was determined after weighing the oven dry mass of a soil sample collected from the subsurface using a stainless steel tube with an internal volume of 99 cm 3 .
An elemental analyser (vario PYRO cube, Elementar UK Ltd.) was used to determine the total C, N and S content of 10-30 mg sub-samples in tin capsules (the mass was dependent on the amount of loss-on-ignition). C stable isotope measurements were made on the same sub-samples using a coupled mass spectrometer (visION, Elementar UK Ltd.).
Organic C content and isotopic compositions were determined from separate sub-samples, after removal of inorganic C with 50 µL of 10% HCl in silver capsules ( where R sampleC indicates the ratio of 13 C/ 12 C in the sample, and R VPDB is the ratio of 13 C/ 12 C in the VPDB. Isotope ratio measurements were calibrated to the VPDB scale using commercially available standard reference materials (B2205 EMA P2, B2153 low organic content soil, B2151 high organic content sediment and B2159 sorghum our, all from Elemental Microanalysis, UK). Further elemental analysis on each sediment sample was performed with a Nito XL3t 950 GOLDD + portable X-ray uorescence spectrometer (pXRF). Data in the results section of this article are reported as depth averages, with depth speci c data reported in the Supplementary Information (S1).

Soil CO 2 e ux
To quantify soil CO 2 e ux, respiration chambers were positioned at 12 locations in a grid across the water-free surface of each salada (Fig. 1a). Maximum distances between chambers were c. 50 m on Salada Pequeña, c. 120 m on Salada Grande and c. 70 m on Salada Jabonera. Comprehensive details of the chamber design and method can be found in Thomas (2012) and Thomas et al. (2018) but, in summary, are made from white uPVC and comprise two parts: i) a lower chamber that when pushed 3 cm into the surface forms an air-tight seal; and ii) a screw-on lid that enables soil gases to accumulate inside the chamber and to be sampled during measurement cycles. The chamber lids contain a sampling port covered with a Suba seal for gas extraction and a two-way valve to ensure any pressure differences between the chamber and atmosphere are minimal and rapidly equilibrated. Chamber surface area is 83 cm 2 and in this study chamber volume ranged from 0.48 to 0.52 litres depending on insertion depth. Heat sinks mounted through the chamber walls ensured the internal air temperatures were not elevated above ambient. Three of the 12 chambers were equipped with a sensor (USB502, Adept Science, UK) to record the air temperature and humidity inside the chamber at 2 minute intervals.
Measurements were taken three times each day to capture a range of temperature and light conditions, giving a total of 36 measurements at each of the three saladas. Following previous sampling protocols to determine soil CO 2 e ux (see for example Thomas, 2012; Thomas et al., 2018), the lid was placed on the chamber and 12 ml of gas was immediately extracted through the sample port using a syringe and hypodermic needle secured with a leur lock. After approximately ve minutes, another syringe was gently pumped to mix the air within the chamber before a second sample was collected. Both sample CO 2 concentrations were determined immediately after each of the three measurement cycles using an EGM-4 infrared gas analyser (PP Systems, Amesbury, USA). Mass CO 2 ux in mg m − 2 hr − 1 was determined from the changes in CO 2 concentration normalised to mean temperature and pressure during measurement (Kutzbach et al., 2007). To correct for the effect of any diffusion suppression owing to the accumulation of CO 2 inside the chamber, a diffusion correction factor was applied (for details see Thomas, 2012). At the time of sampling, surface soil temperature and moisture adjacent to each respiration chamber were determined (n = 3) using an infra-red thermometer and a soil moisture probe (SM150, Delta-T Devices Ltd., Cambridge, UK).

DNA extraction and sequencing
At two locations on each salada, one gram of soil was collected from three depths ( The raw sequencing data were quality ltered (< Q20) using a Trimmomatic trimmer, and merged using FLASH (Magoč and Salzberg, 2011). The chimera were eliminated using UCHIME, and the high-quality sequences were classi ed into different operational taxonomic units (OTUs) at a 97% similarity cut-off using UPARSE (Yuan et al., 2018). The taxonomic information from all the sequences was annotated by an RDP classi er using the Silva database (SSU132), with a con dence threshold set at 70%. Shannon index was analysed by Mothur (V1.30.1), with the details for this index calculation described at https://mothur.org/wiki/shannon/.

Statistical analyses
Statistical analyses of the soil chemical and CO 2 e ux data were performed using SPSS (IBM v. 25). One-way analysis of variance (ANOVA) was used to test whether mean values of the dependent factors (moisture, temperature, total C, total N, S, pH, δ¹³C) were signi cantly different between saladas. For CO 2 e ux, where multiple readings were taken at each site, repeated measure ANOVA was used. The Levene's F statistic was used to test equality of variance. Although ANOVA can tolerate inhomogeneous variance, where these conditions were not met the more robust Welch and Brown Forsythe tests of signi cance were used. Tukey's HSD post-hoc test was undertaken to determine whether saladas were signi cantly different with a probability of p < 0.05.

Salada hydroperiods
Over the six year monitoring period, the saladas had contrasting surface water inundation regimes (Table 1). Salada Pequeña had water covering up to 50% of the surface for 86% of the time, with > 50% water coverage for 47% of the time, and > 80% water coverage for 10% of the time. Saladas Jabonera and Grande were both drier. On Salada Grande, water covered up to 50% of the surface for only 26% of the time, with > 50% water coverage for only 16% of the time.
Salada Jabonera was drier still, with water covering up to 50% of the surface for only 19% of the time, and water coverage never exceeding > 50%. The soil moisture conditions at the time of eld sampling (May 2018) re ected the longer term differences in surface water inundation regimes (Table 1), with signi cant differences evident in the mean soil moisture content for the three saladas (F = 87.4, df = 2, p < 0.001) (Fig. 2a). Nevertheless, on all three saladas, large within-site variations were evident (Fig. 2a), with moisture typically increasing with distance from the fringes towards the salada centres. Mean soil temperatures on Salada Pequeña were signi cantly warmer than on Saladas Grande and Jabonera (df 2, f 7.0, p = < 0.001) but variations in soil temperature within each salada were also large (Fig. 2b).

Soil Properties and C Stocks
All salada soils were alkaline (pH 8.8-9.1) but differences in sediment chemistry between the saladas are evident ( Table 2). Concentrations of K, Ca and organic matter were lowest in the soils of the wetter Salada Pequeña, intermediate in Salada Grande and signi cantly higher in the drier Salada Jabonera ( Table 2). Concentrations of S were lowest in the soils of the drier Salada Jabonera and highest in the soils of the wetter Salada Pequeña. Fe concentrations were very similar in the soils of the drier Saladas Jabonera and Grande and signi cantly lower in the soils of the wetter Salada Pequeña (Table 2). Total C stocks were greatest on the drier Salada Jabonera and least on the wetter Salada Pequeña (Fig. 3). Total C stocks on all saladas were dominated by the inorganic fraction (Table 3, Fig. 3), comprising 75% of the total C stock on the wetter Salada Pequeña and 90% on the drier Saladas Grande and Jabonera. Organic C concentrations were highest on the wetter Salada Pequeña and lower on the drier Saladas Grande and Jabonera (Table 3, Fig. 3). For all three saladas, the δ 13 C VPDB of total C was less negative than the organic C fraction (Table 3). There were also signi cant differences in the δ 13 C VPDB of the soil, with the wetter Salada Pequeña having the most negative δ¹³totalC and the drier Saladas Grande and Jabonera having less negative δ¹³OrgC (Table 3). Table 3 Bulk density, and total, organic and inorganic C concentrations of salada soils, and the δ 13 C VPDB of total C and organic C fractions. Inorganic C calculated from difference between total and organic C.  (Table   4 and Fig. 2c). Soil CO 2 e ux varied signi cantly between saladas (F = 18.3, df = 2, p = < 0.001), with the mean value from Salada Grande signi cantly higher than the mean values from Saladas Pequeña and Jabonera. Nevertheless, there was also signi cant variation in soil CO 2 e ux within each salada in both space and time (i.e. through the measurement day) (Fig. 2c), and is quanti ed in Table 4 as the coe cient of variation. Spatial variations in CO 2 e ux were lower on Salada Pequeña (58.0%) and Salada Grande (30.5%) and higher on Salada Jabonera (95.6%) ( Table 4).
Temporal variations in CO 2 e ux were lowest on Salada Pequeña (38.0%) and Salada Grande (58.4%) and higher on Salada Jabonera (71.1%) ( Table 4). The combined CO 2 e ux data from all three saladas are not well predicted by soil temperature (Fig. 4a). This can be explained by the contrasting CO 2 e ux-temperature relationships on each salada (Fig. 4b-d). Soil temperature alone explains very little of the variance in CO 2 e ux on the wetter Salada Pequeña (Fig. 4d) or on the drier Salada Jabonera (Fig. 4b). On Salada Grande, however, 40% of the variance in soil CO 2 e ux is explained by temperature (Fig. 4c). Soil moisture content is a slightly better single variable predictor of CO 2 e ux from the three saladas than temperature, with a tendency for CO 2 e ux to be higher at intermediate moisture levels of between 40 and 60% v/v (Fig. 5a). At higher moisture contents (> 75% v/v) where the soil is saturated, and where soil moisture is low (< 25% v/v), CO 2 e ux tends to be lower and less variable (Fig. 5a-d). On Salada Pequeña, the initially high moisture content means that e ux tends to decline with increasing moisture (Fig. 5d) but on the drier Saladas Grande and Jabonera, the initially lower moisture contents mean that e ux tends to increase with moisture (Figs. 5c-d).

Microbial Communities
For soils from the three saladas, Shannon index values ranged from 0.3 to 4.4 for bacteria, 1.6 to 3.7 for eukaryota, and 0.7 to 4.0 for archaea (Fig. 6). The bacterial and eukaryotic diversities represented by the Shannon index were not signi cantly different between saladas (P > 0.05; Fig. 6), but the archaeal diversity in Salada Grande was signi cantly higher than in Saladas Pequeña and Jabonera (P < 0.05; Fig. 6). At phylum level, the composition of bacteria, eukaryota and archaea were similar in the saladas, but their relative abundances varied between the saladas (Fig. 7). Bacterial communities were dominated by Proteobacteria and Firmicutes, and together these phyla accounted for 83.4% of bacterial sequences (Fig. 7a). The Proteobacteria were more abundant in Salada Pequeña, while Firmicutes were more abundant in Saladas Jabonera and Grande. The most abundant eukaryotic sequences belonged to Chloroplastida (30.5%), Amoebozoa (8.5%) and Fungi (8.0%) (Fig. 7b). The Chloroplastida and Fungi were more abundant in Saladas Jabonera and Pequeña, while Amoebozoa were more abundant in Salada Grande. The archaea were dominated by Euryarchaeota (53.6%) and Thaumarchaeota (41.2%) (Fig. 7c). The Euryarchaeota were more abundant in Salada Grande, while Thaumarchaeota were more abundant in Saladas Jabonera and Pequeña.

Interpretation And Discussion
The three saladas near Alcañiz form a hydrological continuum, and our ndings show how their contrasting hydroperiods are associated with different soil properties, C stocks, CO 2 e ux and microbial communities. Changes in these soil properties and functions in response to increasing or decreasing wetness are not necessarily linear, but many statistically signi cant differences between the wetter and drier extremes are nonetheless evident. By integrating our interpretations of the hydrological controls on these soil properties, functions and microbial characteristics, we can use an ergodic approach (space-for-time substitution) to assess how these saladas, and potentially other similar saladas in the central Ebro basin and farther a eld, may respond to 21st century climate or human-induced hydrological changes.
Previous studies in South Africa and Australia have successfully used this approach to assess potential changes to oodplain wetlands that occur across wetter-to-drier hydroclimatic gradients (Grenfell et al., 2014;Larkin et al., 2017Larkin et al., , 2020).

Soil Organic Carbon (OC)
We hypothesised (H1) that relatively wet soil moisture conditions would stimulate biological productivity and elevate OC stocks in comparison to drier saladas. On each of the three saladas, soil OC is likely derived from various sources, including microbes, aquatic organisms, birds, insects, plants, and soil washed or blown in from the surrounding landscape. The proportion of soil OC derived from these sources will change with water availability. The highest concentrations of soil OC were found in the wetter Salada Pequeña (Table 3, Fig. 3). The wetter (frequently inundated and/or near-permanently saturated) condition of Salada Pequeña (Fig. 1b, Table 1 Fig. 1a), which will also add organic material to the water column and sediment. Plants are entirely absent from the salada surface, presumably because the saturated, saline conditions inhibit their establishment, and are only present on the higher surfaces elevated above the level of inundation (Fig. 1a). In contrast, at the time of sampling on the drier Salada Jabonera (Fig. 1c, Table 1), there was little evidence of aquatic or bird life, but there are patches of halophytic shrubs across much of the salada surface and a greater density and diversity of shrubs and grasses around the margins. The plants will contribute spatially localised organic matter inputs. Salada Grande, with its intermediate surface water and soil moisture conditions (Fig. 1b, Table 1), has characteristics both of Salada Pequeña and Salada Jabonera; at the time of sampling, it was largely dry with little evidence of aquatic or bird life, and while the salada surface is largely devoid of plants, patches of halophytic shrubs and grasses grow in a narrow band at the margins (Fig. 1b). These different combinations of OC sources may explain why the δ 13 C VPDB of the soil OC fraction from the drier Salada Jabonera is signi cantly less negative than in soils from Saladas Pequeña and Grande (Table 3). Typical δ 13 C values for C3 plants (i.e. generally shrubs and trees) range from − 22 to − 38‰ and for C4 plants (generally grasses) range from − 8 to − 15‰ (Yeh and Wang, 2001), which suggests that a greater proportion of soil OC on Salada Jabonera may be derived from plant inputs, namely the shrubs and grasses on the salada surface and margins, as well as plant litter and soil inputs associated with overland ow from the steeper northern and eastern margins.

Soil Inorganic C (IC)
Compared to the soil OC concentrations, the higher concentrations of soil IC in all three saladas suggests that they may be active sites of carbonate formation due to a combination of evaporative concentration and the reaction of CO 2 with available ions such as Ca 2+ , Mg 2+ and Na + Zhao et al., 2016). For all three saladas, the δ 13 C VPDB of total C was less negative than the OC fraction (Table 3), re ecting the controlling in uence of these carbonates and dilution of the signal from plant and animal material. However, because many parts of the central Ebro basin are underlain by carbonate-bearing lithologies (e.g. limestone, calcareous sandstone) and other evaporites we cannot discern from our data whether the IC is from geological sources (i.e. derived from bedrock at the base of the soil pro le, or supplied by overland ow or aeolian inputs) or whether it is being formed in situ through near-surface pedogenic processes. Nevertheless, although the exact source of soil IC remains unclear, Xie et al. (2009) reported rates of non-biological CO 2 adsorption by alkaline soils of 0.3-3.0 µmol m − 2 s − 1 and concluded that this mechanism is likely to be associated with a hugely underestimated global terrestrial C sink. In Xie et al.'s (2009) study, the intensity of non-biological CO 2 absorption increased with soil salinity and soil alkalinity, and it is reasonable to assume that the saladas are also sites of active and signi cant carbonate formation. Our hypothesis (H2) was that drier saladas would have greater soil IC than wetter saladas if the evaporative concentration of solutes was the dominant process of carbonate formation. In our study, the soil IC in the wetter Salada Pequeña was lower than IC concentrations in the drier Saladas Grande and Jabonera, and also comprised a lower proportion of total C (

CO 2 e ux
At the time of the measurements, the salada soils were a net source of CO 2 to the atmosphere (Table 4)  soil moisture was > 75% v/v or < 25% v/v (Fig. 5). These data support our initial hypothesis (H3) regarding the nature of the relationship between CO 2 e ux and soil moisture (Orchard and Cook, 1983;Schimel et al., 1999;Moyano et al., 2013), whereby high soil moisture contents greatly reduce gas diffusion, and the metabolic activity of aerobic microorganisms is impeded by oxygen de cit (Moyano et al., 2013). Nevertheless, soil CO 2 e ux is highly spatially and temporally variable both within and between the saladas (Table 4, Fig. 2c). In saturated or near-saturated soils, like those in Salada Pequeña, temperature has very little effect on CO 2 emissions (Fig. 4d). In drier soils, like those in Salada Jabonera (Fig. 4b), CO 2 e ux also can become insensitive to temperature (c.f. Thomas, 2012).
The complex spatial and temporal patterns of soil CO 2 e ux from the three saladas illustrate the pitfalls of making single variable predictions using either soil moisture or temperature, something Rey et al. (2017) conclude could lead to considerable errors if using these variables to predict the response of soil CO 2 e ux to climate change in drylands. The very high spatial and temporal variability in CO 2 e ux demonstrates the need for extensive sampling, for single point measurements of e ux in time and space are unlikely to provide a reliable estimate of CO 2 emissions from the saladas.

Microbial Diversity and Composition
Our soil microbial data (Figs. 6 and 7) re ect the broad microbial genetic diversity observed in water samples from 11 shallow, saline lakes in the Monegros region (Casamayor et al., 2013). In their study of Bacteria, Eukaryota and Archaea, Casamayor et al. (2013) found that water salinity levels had little or no impact on microbial ecological diversity (Shannon-Weaver index) or genetic diversity (novelty level). In our study, we hypothesised (H4) that there would be different soil microbial communities in each of the three saladas as a direct, or indirect consequence of their contrasting inundation regimes and soil moisture content. From our data, salada surface water regimes and soil moisture conditions appear to have no effect on bacterial and eukaryotal diversities, although archaeal diversity was signi cantly higher in Salada Grande with its intermediate moisture conditions (Fig. 6). There are, however, different microbial community structures in each salada. These results suggest that different microbial species may be involved in the C metabolisms and cycles depending on the water availability in the salada. Consequently, these distinct community structures will likely lead to differences in C storage between different saladas (Vikram et al., 2015;Maier et al, 2018). Prokaryotic Proteobacteria and Firmicutes (Fig. 7a), together with eukaryotic Fungi (Ascomycota) and Amoebozoa (Protosporangiida) (Fig. 7b)  Nevertheless, Chloroplastida (Chlorophyta), a common dryland eukaryotic photosynthetic microbe (Bates and Garcia-Pichel, 2009;Zhang et al., 2018) is also widespread in the salada soils (Fig. 7b), which suggests that it may generate microbial OC in the salada soils. However, this remains an area of uncertainty as other studies have shown that periods of microbial photosynthesis and net CO 2 uptake are intermittent, unpredictable and likely to be only short-lived (Thomas, 2012. Euryarchaeota were detected in all the salada soils and they are known to include methanogens (Watanabe et al., 2011). Given the saturated condition of some of the salada soils, at least for part of the year, CH 4 may also be generated and released, but we do not have the data to test this notion.

Changes to the C cycle in saladas with 21st century warming
Many of the world's drylands are predicted to warm by 3 to 4°C and experience a reduction in precipitation of 5 to 30% by the end of this century (Lin et al., 2015), including those in Spain (Drobinski et al., 2020). These changes will have a profound effect on multiple ecosystem characteristics and functions within many wetlands in drylands (Huang et al., 2016). For example, for many riverine wetlands, increasing aridi cation, coupled with greater hydroclimatic variability, may induce major changes to channel-oodplain wetland structure and thus patterns of water and sediment dispersal (Tooth, 2018;Larkin et al., 2017Larkin et al., , 2020. To date, such studies have not considered the implications of such physical and associated ecological changes for the C cycle, although many aspects of the C cycle are likely to be affected. For saline inland wetlands (including saladas) in particular, the impacts of climate change on C cycles are essentially unknown (IPCC, 2014b).
With 21st century warming, major structural changes to the saladas near Alcañiz or more generally across the central Ebro basin are unlikely, with the features likely to remain as topographic depressions in an increasingly arid landscape. Nonetheless, these climatic changes will impact on regional hydrology (e.g. lower and more variable rainfall, greater evaporation) and likely will result in altered salada hydroperiods (e.g. less frequent, shorter inundation phases, and more frequent, longer drying phases). Hence, for many saladas there may be a temporal shift along the hydrological continuum from wetter to drier conditions (i.e. a shift from a Pequeña-type situation through a Grande-type situation to a Jabonera-type situation). These hydrological changes may be exacerbated or compounded by human impacts (e.g. river diversion or water abstraction that results in groundwater table lowering). In rarer instances, some saladas may experience a temporal shift from drier to wetter conditions; for example, where excess irrigation runoff or deliberate damming leads to raising of local groundwater tables and establishment of permanent lakes, as has happened at La Estanca, a former salada located ~ 5 km to the west-northwest of Alcañiz (Gutiérrez et al., 2005), and also at La Were et al., 2019), few studies have investigated the C cycle in wetlands in drylands. Consequently, many knowledge gaps remain, particularly regarding how climate changes will affect the mechanisms and rates of IC formation, the residence times of OC, and the interaction between OC and IC. These gaps contribute to uncertainty surrounding the size of the C stock in wetlands in drylands as well as the processes that lead to C uptake and release, and to CO 2 emissions. What is clear is that the saladas near Alcañiz, and possibly similar features across the wider central Ebro basin, are currently sites of C storage, contain active microbial communities and typically release CO 2 to the atmosphere. Although the impacts of aridi cation will depend on the initial hydration status of the saladas, our ndings suggest that we can perhaps expect a reduction in OC stocks, as the OC sources most vulnerable to declines in water availability are those associated with stream inputs and aquatic fauna (e.g. crustaceans, amphibians, bird life).
By contrast, aridi cation of the saladas may increase soil IC formation through evaporative concentration, at least in the short term. Over longer time periods, desiccation will reduce the supply of cations and bicarbonate to the saladas, thus limiting the capacity to form carbonates. In addition, a reduction in sediment moisture content in initially very wet saladas is likely to lead to an increase in CO 2 e ux as conditions become more favourable for aerobic microbial activity in the soil. This may reduce soil OC residence times as it is more likely to be respired. By contrast, further drying of saladas where soil moisture content is already typically < 50% v/v may inhibit microbial activity and lead to a reduction in CO 2 e ux. The foregoing trends suggest that any future reduction in the extent and duration of inundation and/or soil saturation on the saladas may shift the composition of C stocks to less organic-rich, more carbonate-based stocks, at least in the short term, with an overall decline in the total C stock.
Caution needs to be exercised in extrapolating the ndings from the saladas near Alcañiz and those in the the wider central Ebro basin to other saladas in Spain or to saline inland wetlands more generally (e.g. Zhang et al., 2020). Combined with the aforementioned process uncertainties, different playas may respond to changes in hydroperiods in different ways depending on the dominant ions in solution. Evaporation of water from saladas with appreciable concentrations of calcium bicarbonate or calcium sulphate is likely to result in the precipitation of calcrete (and/or gypsum) and thus increase IC stocks (Eugster and Jones, 1979;Tooth and McCarthy, 2007). Where saladas are dominated by sodium bicarbonate, sodium chloride, or sodium sulphate, evaporation is more likely to result in the development of increasingly saline surface water that is toxic to many biota and lead to a reduction or change in OC inputs (Eugster and Jones, 1979;Tooth and McCarthy, 2007).
The potential in uence of changing hydroperiods on the soil microbial community in saladas and the implications for OC and IC formation, storage, and release also remains poorly understood. Our data (Fig. 7) suggest that prevailing moisture conditions affect the soil microbial community structure through changes to relative abundances. A reduction in salada soil moisture may lead to lower relative abundances of Proteobacteria, Amoebozoa and Euryarchaeota, ultimately in uencing C assimilation, metabolism and storage. Nevertheless, drying of saladas may also cause greater aerobic microbial activity (at least initially), which will probably reduce soil OC residence times and further affect the microbial community structure (Swenson et al. 2018). These uncertainties notwithstanding, our data can serve as a framework to guide future investigations.

Conclusions
Three adjacent saladas near Alcañiz with contrasting surface water regimes and soil moisture conditions provided an opportunity to investigate, for the rst time, how water availability affects their C stocks, CO 2 emissions, and microbiology. Frequent inundation and/or near-permanent soil saturation promotes higher soil OC stocks owing to input from a greater range of OC sources, especially biological. Drier salada soils, however, have higher soil IC stocks owing to evaporative concentration and formation of carbon-bearing mineral compounds. CO 2 e ux is highest at intermediate moisture conditions, but the spatial and temporal variability in CO 2 e ux on salada surfaces was very high, demonstrating the need for intensive sampling regimes to provide realistic estimates of their contribution to atmospheric CO 2 exchanges. There were also different microbial community structures in each salada, suggesting a possible link to moisture availability. These three saladas, and numerous other similar features across the central Ebro basin, are important areas of C storage and likely play an important role in local and regional C cycles. They also host many rare and threatened plants and animals (Conesa et al., 2011; European Commission, no date), and their C cycle and microbial community characteristics provide an additional reason to monitor changes resulting from 21st climate changes and protect these vulnerable environments from anthropogenic disturbances.   Data from the three saladas, illustrating within-and between-site variations in: (a) moisture; (b) temperature; and (c) CO2 e ux. Individual values shown with mean (adjacent larger marker) and standard deviation (line). Horizontal displacement of the data points for each salada are partly for clarity but partly schematic as they correspond to the shorter lines of sampling locations in each of the sampling grids (see Figure 1a). p values indicate signi cance of differences.