Using ✿✿✿✿✿✿✿✿✿ water ✿✿✿✿✿✿✿✿✿ stable ✿ isotopes to understand evaporation, moisture stress and re-wetting in catchment forest and grassland soils of the summer drought of 2018.

isotopes to understand evaporation, moisture stress and re -wetting in soils 2018.” presents an interesting contribution to our understanding of ecohydrological processes in a mixed land cover catchment (forest and agricultural), especially under the influence of climate anomalies. The authors conducted a case study in North-East Germany in the Demnitzer Millcreek catchment. They highlight the use of isotopic tracers together with conventional hydrology to understand the effect of drought progress, the recovery of soil water storage and the memory effect of drought evaporation when the catchment could no longer hold streamflow and crop production and further mixing with fresh precipitation. Abstract. In drought sensitive lowland catchments, ecohydrological feedbacks to climatic anomalies can give valuable insights into ecosystem functioning in the context of alarming climate change projections. However, the dynamic inﬂuences of vegetation on spatio-temporal processes in water cycling in the critical zone of catchments are not yet fully understood. We used ✿✿✿✿ water ✿ stable isotopes to investigate the impacts of the 2018 drought on dominant soil-vegetation units of the mixed land-use Demnitzer Mill Creek (DMC, NE Germany) catchment (66 km 2 ). The isotope sampling was carried out in conjunction with 5 hydroclimatic, soil, groundwater, and vegetation monitoring. Drying soils, falling groundwater levels, cessation of stream ﬂow and reduced crop yields demonstrated the failure of catchment water storage to support “blue” and ✿✿✿✿✿✿✿✿✿✿✿ (groundwater ✿✿✿✿✿✿✿✿ recharge ✿✿✿ and ✿✿✿✿✿✿ stream ✿✿✿✿✿✿✿✿✿ discharge) ✿✿✿ and ✿ “green” ✿✿✿✿✿✿✿✿✿✿✿✿✿✿✿✿ (evapotranspiration) ✿ water ﬂuxes. We further conducted monthly bulk soil water isotope sampling to assess the spatio-temporal dynamics of water soil storage under forest and grassland vegetation. Forest soils were drier than the grassland mainly due to higher interception and transpiration losses. However, the forest soils also had more 10 freely draining shallow layers, and were dominated by rapid young (age < 2 months) water ﬂuxes after rainfall events. The grasslands soils were more retentive and dominated by older water (age > 2 months), ✿ ; ✿ though the lack of deep percolation produced water ages ∼ ✿✿ > 1 year under forest. We found the displacement of any “drought signal” within the soil proﬁle limited to the isotopic signatures and no displacement or “memory effect” in d-excess over the monthly time step, indicating rapid mixing of new rainfall. Our ﬁndings suggest that contrasting soil-vegetation assemblages communities have distinct impacts on ecohydrological partitioning and water ages in the sub surface ✿✿✿✿✿✿✿✿✿ sub-surface. Such insights will be invaluable for developing sustainable land management strategies appropriate to water availability and build resilience to climate change.

Ecosystem response to this climatic anomaly is investigated by using water stable iso-tope data of precipitation and throughfall, stream-, groundwater and especially from soil water profiles. Monthly soil profile samples in 6 different depths down to 1 m under two different land-use types were taken from September 2018 to February 2019. Soil water isotopes were analysed using direct vapor equilibration laser spectrometry (DVE-LS). These data were used to estimate mean transit times (MTT) in the soils at the different depths as well as young water fractions, using a fitted sine-wave method. Based on collected meteorological and sap flow data ET-pot was calculated. Soil moisture was monitored in three different depths at both sites. Drought severity was quantified with the SPI, based on long-term precipitation data from the DWD.
It could be shown that the forest soils were dominated by rapid young water fluxes after rainfall events whereas the grassland soils were more retentive and dominated by older water. It is concluded that implications for blue and green water management should be investigated in a greater range of representative vegetation/ soil units and that further research efforts on climate change and management adaptations in the critical zone of drought sensitive ecosystems is needed.
Overall, the manuscript is well structured and nicely written.
The topic fits well to the scope of the journal and appears to be of interest for the readers; I only suggest moderate revisions prior to acceptance and publication in Hydrology and Earth System Sciences.
We thank the reviewer 1 for the encouraging comments on our manuscript. We are grateful for this very detailed and careful review of our work.

Fig. 1
There is enough space to put the overview beneath the detailed mapsmakes the layout of the figure a bit clearer. Right part: please replace "Landuse" by 'Soils' Legend title was adapted.  Corrected Table 4 Deltas are missing in header Are 5th and 95th percentile of d18-O identical for precipitation and throughfall? Stream: 5th and 95th percentile of d18-O both -8.6? Please double-check.
We double-checked table 4 and precipitation and throughfall numbers are correct. Stream 95th percentile was corrected. Thank you! Table 5 Caption: 'soil water isotope samples' instead of "soil isotopes samples" Thank you. We changed 'soil isotope samples' to 'bulk soil water isotope samples' to clarify.
Global changes I would prefer "and" instead of "&" (e.g. l. 62, 112, 136) The study shows an important work with a logical structure and is clearly written, in my opinion, it deserves to be considered for publication in the HESS after some minimal revisions. Most of my editing comments match those of Referee 1 and have already been addressed by the authors.
We are grateful for the constructive comments of reviewer 2 on the manuscript. We appreciate this thoughtful and stimulating review of our work. Throughout the revision of the manuscript, the authors have adapted the terminology used to improve clarity for the discussion and key messages. For the specific comments, we have addressed them accordingly. Responses are given below and will be incorporated in the revised manuscript.
I recommend the authors to be careful when using the terms "blue and green water", as it is broad and varied in the literature, so I suggest that they highlight in the introduction section what they specifically refer in this study. We clarified the use of terms "blue" (as groundwater recharge and stream discharge) and "green" (evapotranspiration) water fluxes in the introduction section. (L. 7-8). We would likerespectfullykeep these terms as they are important and widely used terms in the literature.
I'm a little concerned about the limited availability of soil water isotope samples (monthly basis) used to drive such a conclusion based on tentative MTTs. The manuscript would benefit for a wider discussion and to clearly state this limitation. In order to reaffirm the credibility of these results, I suggest widening the context of the study by comparing it with similar drought cases in nearby sites or with comparable geographical regions.
We agree, and recognise of the limitations of monthly destructive samples and we are careful to be circumspect about the inferences. Still, as the work getting such samples is so labour intensive, not many such data sets exist. To assess spatial variability (replicates) and enable ongoing sampling in the limited site area beyond the study period, we had to limit the temporal resolution of the sampling. Nevertheless, the insight in subsurface processes was invaluable and demonstrated the efficiency of this method for a first approximation. We will expand the critical discussion but cannot widen the context of the actual study (obviously) as this would be a different paper. To our knowledge, no other bulk water isotope samples exist from nearby sites (or at least are not published yet). Though we make comparison to chloride related water ages from lysimeters for another site in Brandenburg. (L. 295-296, 356-358) Further, an extended amount of literature pointed out that MTT (based is a gamma distribution with two parameters and derived MTTs concept) is only a qualitative indicator of catchments for a first screen and basic comparison, however a bit critical when the evolution of water ages is involved. With the available information, I firmly believe that it would be possible to obtain better and accurate results by including more elaborate and non-stationary criteria in the analysis.
We agree with your concern and are well aware of the limitations of this method. We tried to emphasise the tentative nature of these results. Further this concern lead to the additional consideration of Young water fractions as an unbiased indictor of water ages. This largely supported the MTT results and helped underpin our conclusions. The basic nature of this analysis is further highlighted in the manuscript (L. 178, 403) Also, we added a reference to a process-based ecohydrological modelling approach (considering isotopes) at these two plot sites which also estimated the water ages in the conclusion (recently published by Smith et al.;L. 412).This is also broadly consistent with the more basic approximations reported here. However, conducting such complex non-stationary analysis would be beyond the scope of the paper.
Finally, please improve figure 4, the size of the symbols and the colours used make it difficult to identify isotopic signatures.
We adapted color-codes and sizes to make it easier to identify.

Editor comments
As you can see the two reviewers were rather positive on your study. The comments they raised are well addressed by your reply. So I invite you to make these changes accordingly.
Thank you. We will include the changes in the manuscript.
Additionally, I have some minor comment too: -L187-188: unit should be mm/d Thank you. We changed it from mm to mm/d. (L. 204-205) -L212: an interception percentage of 7% is rather low, so likely the canopy is quite open, resulting in a significant amount of throughfall that stems from direct rainfall. This might explain the little enrichment in throughfall water. It would also be nice to refer to https://doi.org/10.1002/wat2.1187, where a nice overview is given on the effect of interception on throughfall. water fluxes. We further conducted monthly bulk soil water isotope sampling to assess the spatio-temporal dynamics of water soil storage under forest and grassland vegetation. Forest soils were drier than the grassland mainly due to higher interception and transpiration losses. However, the forest soils also had more 10 freely draining shallow layers, and were dominated by rapid young (age < 2 months) water fluxes after rainfall events. The grasslands soils were more retentive and dominated by older water (age > 2 months), ✿ ; ✿ though the lack of deep percolation produced water ages ∼ ✿✿ > 1 year under forest. We found the displacement of any "drought signal" within the soil profile limited to the isotopic signatures and no displacement or "memory effect" in d-excess over the monthly time step, indicating rapid mixing of new rainfall. Our findings suggest that contrasting soil-vegetation assemblages communities have distinct impacts 15 on ecohydrological partitioning and water ages in the sub surface ✿✿✿✿✿✿✿✿✿ sub-surface. Such insights will be invaluable for developing sustainable land management strategies appropriate to water availability and build resilience to climate change.
Copyright statement. TEXT 1 Climate change provides an urgent impetus for an improved understanding of ecohydrological interactions in areas where 20 water is limited (Wang et al., 2012). Increasing temperatures and reduced rainfall in the growing season are affecting extensive regions (Tetzlaff et al., 2013); in some cases, causing natural vegetation communities to adapt by changing their composition, distribution and physiology (Wookey et al., 2009). Where vegetation is managed for forestry and agriculture, land use strategies may need to adapt to build resilience towards newly evolving climate regimes. This includes choice of species, crop rotation cycles and sustainable production targets (Stoate et al., 2009). As well as constraining biomass productivity, such strategies will 25 also have implications for the residual water available to maintain groundwater recharge, river flows and associated ecosystem services. In summer 2018, an exceptional drought over much of Europe set new, long-term meteorological records causing significant loss of agricultural production, water shortages and low river flows over extensive areas (Imbery et al., 2018). This drought previewed the warmer, drier conditions that climate change is expected to deliver across much of Central Europe as the 21 st Century progresses. For a future where water resources might become less reliable, conceptualisation of the dynamic 30 interactions between vegetation, soils and water fluxes from stores in various ecosystem compartments needs to be improved and is a major focus of current "critical zone" science (Penna et al., 2018). Understanding local environmental factors, like how atmospheric water demand is modulated by vegetation cover, is a prerequisite to better managing the effects of droughts (Mishra and Singh, 2010). Stable isotopes in the water molecule have been successfully used to trace water fluxes in the soilplant-atmosphere-continuum , and can reveal important process insights into ecosystem water cycling 35 (Dubbert and Werner, 2019). Stable isotopes of hydrogen and oxygen are often seen as ideal tracers as they are an integral part of the water molecule itself. Isotopes are conservative tracers and not altered by chemical reactions, but only by mixing and fractionation. Numerous studies have applied stable water ✿✿✿✿ water ✿✿✿✿✿ stable ✿ isotopes to constrain water sources and fluxes in the unsaturated zone. Complementary to hydroclimatic monitoring, stable isotopes as environmental tracers can provide insights into ecohydrological processes in the "critical zone" (Grant and Dietrich, 2017). They have been used to investigate evaporation 40 (Allison and Barnes, 1983;Barnes and Allison, 1988), groundwater recharge (Koeniger et al., 2016), weathering influence on flow paths (Bullen and Kendall, 1998) as well as water ages (Tetzlaff et al., 2014;Sprenger et al., 2019b), plant water uptake (Rothfuss and Javaux, 2017) and the partitioning of evapotranspiration (Kool et al., 2014;Xiao et al., 2018). Highly seasonal dynamics of soil water ages and their dependency on soil water storage have further been investigated via ✿✿✿✿ water ✿ stable isotope modelling (Sprenger et al., 2018). Isotopes were also used to examine the influence of vegetation on soil bulk (Oerter and 45 Bowen, 2019; Sprenger et al., 2017) or other components like throughfall  often supplemented with xylem water isotope data (Geris et al., 2015;Brooks et al., 2010;Goldsmith et al., 2019 (Sprenger et al., 2015). However, there are still unresolved problems related to sampling (Sprenger et al., 2015) or extraction (Orlowski et al., 2016(Orlowski et al., , 2018 of water from complex matrices like soil or plant tissue. Laboratory routines for the direct-equilibrium method (Wassenaar et al., 2008) use the state of isotopic equilibrium between liquid and gaseous water in a closed system to determine the isotopic signature of the liquid soil water. They have been developed (e.g., Hendry et al., 2015) and applied successfully ✿✿✿✿✿✿ adapted ✿✿✿ and ✿✿✿✿✿✿✿✿✿✿✿ successfully ✿✿✿✿✿✿ applied ✿ in several studies (e.g., Klaus et al., 2013;Sprenger et al., 2017;55 Stumpp and Hendry, 2012). On-going efforts are being made to solve various problems associated with the different methods for soil water isotope analysis, and results have to be interpreted accordingly (Gaj and McDonnell, 2019;Gralher et al., 2018;Sprenger et al., 2015). This study focuses on the long-term monitoring site Demnitzer Millcreek catchment (DMC), a mixed land use catchment located south-east of Berlin in Brandenburg, Germany. As the exceptional drought developed in summer 2018, we monitored moisture dynamics in drying soil profiles under different land cover types, falling groundwater levels and 60 decreasing stream flows. Crucially, we used stable isotopes from different waters in the latter stages of the drought to address the specific objectives of this study ✿✿ to: 1. To assess the development and progress of the drought and the subsequent recovery on soil water storage.
2. Explore ✿✿✿✿✿✿ explore, using bulk soil water isotopes, the evolution of the evaporation signal of the drought and its "memory" effect following infiltration & ✿✿✿ and ✿ mixing with new precipitation during re-wetting.

Discuss
✿✿✿✿✿✿ discuss the implications of ecohydrological processes for the response times and recovery of hydrological stores in the DMC catchment by combined use of hydroclimatic and isotope data.

Study site
Our study was based ✿✿✿✿✿✿ located ✿ in the 66 km 2 Demnitzer Millcreek catchment (DMC) in Brandenburg, north east Germany (52°23'N 14°15'E), 55 km south-east of Berlin. This long-term study site is a tributary of the River Spree and one of the 70 few headwaters in the region that does not originate in a lake but in a network of agricultural drainage channels. Catchment orientation is NNE -SSW with elevation from 38 to 83 m above sea level and a low average slope of less than 2 %. Located in the Northern European Plain, the geology of the catchment is strongly influenced by the Pleistocene glaciation.
The catchment outlet is situated in the Berlin glacio-fluvial valley near Berkenbrück, where the DMC surface runoff drains into a small lake (Dehmsee) and subsequently into the River Spree. The geology of the upper catchment is dominated by 75 unconsolidated sediments of ground moraine material. Important factors for nutrient cycling in this landscape are kettle hole lakes (Nitzsche et al., 2017) and wetlands (Smith et al., 2020a). The stream network is embedded in fluvial and periglacial deposits surrounded by basal tills with intermittent riparian peat fens in valley bottom areas. The northern catchment is mainly characterised by eutric soils and silty brown earths. Next to the stream, sandy gleysols or peaty histosols are dominant. DMC has a seasonal and strongly continental climate, with cold winters (mean air temperatures in January and July are 0.2°and 19°C, 80 respectively). Precipitation is dominated by convective summer events and low intensity winter rain, with generally less than 10 % of the annual total occurring as snowfall. Potential evapotranspiration (PET) commonly exceeds average precipitation and runoff coefficients are typically < 10 % of annual precipitation (Smith et al., 2020a). Non-irrigated arable land (mainly winter cereals, maize) dominates the upper catchment and contributes 58 % of the area. Further downstream, the cover of mixed coniferous and deciduous forests increases. The stream traverses several peat fens that were used as pasture. Manmade 85 3 connections of disconnected glacial hollows to the stream network altered ✿✿✿✿✿✿✿✿ increased the total channel length from 20 km in 1780 to 88 km at the present day (Nützmann et al., 2011) to supply mills and gain new arable land by draining. The catchment has been subject of various studies investigating e.g. CO 2 saturation (Gelbrecht et al., 1998), influences of wastewater treatment (Gücker et al., 2014), other historical anthropogenic impacts (Nützmann et al., 2011) and the impact of beaver re-colonisation (Smith et al., 2020a). This study focuses on two plots with contrasting landcover in close spatial proximity to each other 90 (∼400 m) and the stream (Figure 1). The two experimental plots are forested (FA) and covered by grassland (GS). FA is dominated by mature oak trees (Quercus robur), and includes other tree species such as Scots Pine (Pinus sylvestris) and red oak (Quercus rubra). GS is pasture that is harvested once a year. Distance to stream differs between GS (∼15 m) and FA (∼90 m). GS has eutric arenosol (humic, transportic) soil whereas the FA soil is a lamellic brunic arenosol (humic) according to the World Reference Base (WRB) classification. GS is characterized by higher clay contents in the upper soil, a higher pH, 95 and narrower C/N ratio than FA. There is also a shift in pH at FA due to the presence of calcite at the lowest layer.

Data and methods
An automatic weather station AWS (Environmental Measurement Limited, UK), located in the NW of the catchment, was used to record meteorological data (e.g. net radiation, air temperature, precipitation, ground heat flux, relative humidity) every 15 minutes. To monitor transpiration rates from trees at FA, a sap-flow measuring system with 32 sets of Granier-type (Granier,100 1987) sensors (Thermal Dissipation Probes, Dynamax Inc., Huston, USA) was installed in 13 trees during the growing season from 21.4.18 to 23.10.18. Sensors were installed at approximately 1.3 meter above ground. The tree diameter was also measured at this height (DBH; mean: 76 cm; SD: 35 cm). All sensors consisted of two thermometers installed in the sapwood in 4 cm vertical distance from each other and were shielded from external sources of temperature change (e.g. radiation). The upper thermometer was heated and differences in temperature were collected hourly with a CR1000 data logger (Campbell 105 Scientific, USA). The difference in temperature was used to calculate flux velocity and combined with the sapwood area to calculate a flux rate. Conditions of zero transpiration were determined from daily maximum temperature differences. The resulting flux rate per unit sapwood area was adjusted to the plot using a ratio of sapwood area to forest area that was established with ten trees. Data from the AWS were used to estimate potential evapotranspiration with the FAO Penman-Monteith equation (Allen et al., 1998). To facilitate comparison, the sap-flow derived transpiration and FAO PET were normalized by subtract- Umwelt-Geräte-Technik GmbH, Müncheberg, Germany) at three depths (20, 60 and 100 cm) with six replicates per site. The probes recorded with a 15 minute frequency and a precision of ±3 % for volumetric soil water content and ±0.2°C for soil temperature. The data was averaged and aggregated to daily values to estimate soil storage in the first meter ( Figure 2) from volumetric soil moisture by weighting the upper sensors to represent 40 % of the first meter each and the lowest 20 %. One 120 soil pit (depth > 100 cm) per plot was excavated and the profile was described following common pedological procedures. Soil cores and composite samples were taken to determine further physical and chemical characteristics in the laboratory of the Technical University of Berlin (Table 1) Water samples in autosamplers and rain gauges were protected from evaporation by a paraffin oil layer of a thickness > 0.5 mm (IAEA/ GNIP precipitation sampling guide V2.02 September 2014). Samples were extracted with a syringe from below the paraffin and filtered (0.2 µm, cellulose acetate) in the field and cooled until stored at 8°C in the laboratory. in the topsoil (0-10 cm) to gain sufficient water (3 ml; Hendry et al., 2015) for the lab analysis. Soil bulk water isotopes were analysed using the direct-equilibrium (DE) routine of Wassenaar et al. (2008). Soil samples were immediately stored in sampling bags and sealed with ziplocks instantaneously. All samples were then stored sealed and thermally isolated until being weighed, inflated with the headspace gas and heat sealed in the laboratory. We used diffusion-tight metalized sample bags ( ✿✿✿✿✿✿✿✿✿✿✿✿ CB400-420siZ, ✿ Weber Packaging, Güglingen, Germany) as established in other direct equilibrium studies (Sprenger et al., 155 2015). Synthetic dry air was utilized as inflation atmosphere to enable a posteriori correction of biogenic gas matrix changes in the headspace . After introducing dry synthetic air as headspace  The gamma function was fitted by maximizing the objective function using the Kling-Gupta Efficiency statistic (KGE; Gupta et al., 2009) within predefined parameter ranges for the shape factor (α; 0.5 to 5) and the scale parameter (β; 2 to 50); these were set to avoid MTTs shorter than the sampling frequency. The shape of the gamma distribution enabled us to represent short-and long-term tracer input contributions (e.g., Kirchner et al., 2001) to soil bulk water. We further excluded the first two sampling dates in the upper 10 cm of both plot soils for this analysis to avoid implausible results due to tracer enrichment 185 introduced by soil evaporation. We calculated young water fractions using the fitted sine-wave ✿✿✿ sine ✿✿✿✿✿ wave method described by Von Freyberg et al. (2018), adjusting the topsoil values for the two first sampling occasions to the precipitation input for the same reasons. The young water (Figure 2) represents the estimated fraction of water in the sampled soil depth that fell as precipitation within the last 2-3 months.

Hydroclimatic situation
Exceptional climatic conditions during the study period, with low precipitation and high temperatures, are reflected by the Standardized Precipitation Index (SPI, Table 2). Values varied between "moderately wet" (1.0 to 1.49) to "extreme drought" (-2 and less). The different SPI time windows indicate the progression of the 2018 drought in different temporal contexts and therefore represent drought impacts on different compartments of the catchment water cycle. We found short-term monthly and c) are given as the geometric mean (m g ) of accumulated daily values. The forest soil was notably drier than the grassland; and overall, the grassland soil showed much less variability and a lower drought effect on soil moisture 210 than the forest site. The upper forest soil moisture content showed rapid responses to precipitation inputs. Further progression of wetting fronts to depth was damped and lagged, which resulted in decreasing SD with depth (Table 3) (Figure 2 d). Flows had ceased earlier in the summer (20.7.18) as groundwater levels fell, though there was a brief response to the July rainfall that resulted in temporary discharge (27.7.18).

Dynamics in
The isotopic samples obtained from different water cycle compartments are displayed in two-dimensional isotope space ( Figure   3) supplemented by the global meteoric water line (GMWL), (Craig, 1961). Statistical characteristics are summarised in Table   225 4. Daily precipitation showed the highest range from being depleted in heavy isotopes (δ 18 O= -18.3 ‰; δ 2 H= -140.2 ‰) in winter to less depleted and even enriched in oxygen relative to the VSMOW standard in summer (δ 18 O = 0.3 ‰; δ 2 H = -7.3 ‰).
Throughfall samples showed a smaller range in δ 18 O (min = -17.0 ‰, max = -1.1 ‰) and δ 2 H (min = -129.5 ‰, max = -14.9 ‰). The forest and grassland soil samples exhibited substantial damping in isotopic variability relative to precipitation. They also displayed deviations from the GMWL at the more enriched end of their dual isotopic spectrum. We found that water in forest soils had a heavier isotopic composition (Table 4)   Heatmaps ( Figure 5) show the changing bulk soil water isotope signatures for both landuse types, which are strongly influenced by precipitation inputs (shown in Figure 4). The geometric mean (m g ) of the replicate samples is displayed as a colour code for all sampled depths. Sampling started in September 2018 after several months of high air temperature and low precipitation when the severity of the drought became clearer (Figure 2). Under both land use types, the upper 20 cm showed highest dates at 2.5 cm. The SD decreased with depth ( Drought impacts on the soil bulk water by evaporation are shown by more negative values of d-excess as this metric was 260 originally presented as an index for non-equilibrium conditions (Dansgaard, 1964).  Estimates of water ages and young water fractions (i.e. % of water younger than 2-3 months) are displayed in Figure 6, and show striking differences with depth between FA and GS soil profiles. The FA young water fraction was > 75 % in the upper 5 soil depths and dropped dramatically to < 5 % at 90 cm depth. GS young water fractions declined more gradually with depth.
Values ranged from > 75 % at the upper most soil layer down to ∼20 % at 90 cm. The method returned significant sine wave fits (Table 6) Table 1) and ecohydrological feedbacks (Figure 2). The drought was characterised by low precipitation input (Figure 2 a), high temperatures (a), low soil water storage (b, c), declining groundwater levels (d), and stream flow ceasing 290 ✿✿✿✿✿✿✿ ceasation ✿ (d). Only the heavy rainfall in July prevented the drought being ✿✿✿✿✿ much more severe, given the persistence of dry warm weather into the late autumn. We found that local observations in DMC were consistent with other, large scale observations on the characteristics of the drought (Imbery et al., 2018). We observed differences of the drought dynamics under the FA and GS plots which were expected considering the importance of vegetation on water cycling in the critical zone (e.g., Dubbert and Werner, 2019). Generally, soils under the forest were drier than the grassland, which likely reflects the greater interception and 295 transpiration losses under forest that have been observed elsewhere in Brandenburg (Douinot et al., 2019). However, it is clear that differences in soil properties also result in greater moisture storage and retention in the more clay-rich upper profile of the grassland soil. Dynamics of PET and transpiration rates of oak trees (Figure 2) imply that the trees were able to successfully sustain transpiration throughout this drought, as transpiration rates did not appear to be reduced relative to atmospheric demand (indexed by PET). Low soil water availability under the forest during drought conditions raise the question of the origin of the 300 water transpired by oaks. Clearly, the high rainfall inputs in July, which replenished storage in the top ✿✿✿✿✿ upper soil were likely critical in enabling transpiration to continue through July and August via rapid recycling. However, oak trees may also be able to access deeper water sources (e.g., David et al., 2004) near the water table via deep roots (which were present at 1 m in the FA sampling plot). In addition, ecophysiologically based water-limitation-tolerance has been observed in various oak species by Hahm et al. (2018). Isotopes were key tools used in this study to assess the impact of the drought on different soil-vegetation assemblages. The direct equilibrium method applied used monthly destructive soil sampling to return stable isotope ratios in the soil bulk water 310 molecules. This provided further insight in ✿✿✿ into ✿ spatio-temporal patterns of water movement and mixing in the unsaturated zone ( Figure 5). Using heat maps, we were able to visualise qualitative patterns of site-specific advection and dispersion of the isotopic input signal from the soil-atmosphere interface down to 1m. Further, the evaporation signal from the drought 2018 was apparent in the soil bulk water d-excess profiles at the forested and grassland site in September and October. Summer soil evaporation, in combination with precipitation deficits during the drought, were likely the driving processes leading to 315 the temporarily enriched (compared to recent precipitation) isotope signatures in the topsoils ✿ Both sites showed a subsequent displacement and mixing of the bulk soil isotope signatures with incoming precipitation as re-wetting progressed in the autumn and early winter. However, the evolution of isotopic and 320 d-excess signal in the unsaturated soil storage indicates differences between sites ( Figure 5). The forest soil being sandier and having little direct ground vegetation cover, allowed a deeper penetration of the evaporation front and therefore more negative d-excess values in depths up to 30 cm. Limited transfer of d-excess effects to depths observed in this study is in accordance with findings of Sprenger et al. (2016). Lower soil moisture (and therefore water storage) in the forest (Figure 2) likely led to ✿ a stronger Rayleigh fractionation effect on the remaining bulk soil water. Mixing with incoming precipitation resulted in moisture 325 replenishment and rapid over-printing of the isotopic drought signal (d-excess) at both sites. We did not see a strong "memory" -effect ✿✿✿✿ effect ✿ of the d-excess in individual soil depths on a monthly time step over the entire study period. The apparent contradiction between our findings and recent findings by (Sprenger et al., 2019a) who reported consistently disjunct soil water pools, likely simply reflects different soil and climatic properties, leading to different hydrological pathways. (Sprenger et al., 2018) described the bound soil water as water under different matric potentials. This has implications for the interpretation of 330 interactions speeds between different water pools in the soil and their partitioning into green and blue water fluxes. Findings by Bowers et al. (2020) support the relatively fast time-dependent isotopic mixing of water held under different tensions in the matrix of sandy soils. A study in a controlled ecosystem (Evaristo et al., 2019) further highlights the importance of spatiotemporal dynamics in soil water for its partitioning and interpretation of resulting patterns.

335
We were also able to use the isotope time series to provide a first approximation of the travel times of water in the soil during the re-wetting phase. Despite the short data time-series, fitting simple sine wave and gamma models ( Figure 6) provided useful insights into the spatial differences in the young water fraction and MTTs between, and within, the soil profiles of the two sites. As we sampled bulk soil water at only monthly intervals, the resulting values can only be regarded as indicative, but they capture the shift from dry to wet conditions. The upper soils ✿✿✿ soil ✿✿✿✿✿ layers ✿ are dominated by younger water (less than a month 340 ✿✿✿ old) at both sites and age increases with depth ( Figure 6). Likely causes of the poor model fits of TTD in the upper soil are evaporative fractionation and the sampling frequency being too coarse to capture the temporal resolution of these processes in the sandy soils. Furthermore, the bulk soil water sampling likely underestimates the effects of preferential flow, especially in the more heterogeneous upper horizons. Nevertheless, similar ages and differences were reported by Smith et al. (2020b) from a physically based tracer-aided ecohydrological modelling approach; this increases confidence that the results are instructive. 345 We found a steady increase in age with depth at the grassland site and generally higher soil moisture content. These patterns were represented by higher α-values in comparison to the FA site (Table 6) (Tetzlaff et al., 2014) in freely draining podzols in the Scottish Highlands. This leads to ✿✿✿✿✿✿✿✿✿✿✿ down-profile ✿✿✿✿✿✿ change ✿✿✿✿✿✿ reflects ✿ a ✿ more consistent moisture flux to depth, which is characteristic of the grassland site. In contrast, the FA soil 350 was more freely draining, and drier, and hence younger water could have a greater influence following rainfall events, even in the deeper soil layers. Substantially older soil water at 80 to 100 cm depth indicates more irregular groundwater recharge at the FA site, though the poor model fit here highlights the low variability in the isotopic composition of the deep forest soil and resulting uncertainty. These tracer-based inferences were supported by the soil moisture dynamics (Figure 2 b & ✿✿✿ and c) which also indicated more frequent percolation to depth under grassland and a temporally limited (to winter) groundwater recharge 355 under forest. Deeper soil bulk water that mainly represents groundwater recharge was older under the forested plot. This is in accordance with findings from tracer-aided modelling of water age dynamics under forest and grassland at another site in Brandenburg from Douinot et al. (2019). The displacement of the isotopic signal down the soil profile was not observed in d-excess, with limited "memoryeffect" ✿ " ✿✿✿✿✿ effect ✿ of the drought 2018 with time and depth in both soils. But, despite the rapid recovery of the d-excess signal in the isotopic composition of bulk soil water, this was not indicative of drought recovery. With 360 SPI values still low for longer averaging periods, the effects of the drought impact were still evident in the catchment. Stream flow did not recover until the beginning of 2019, and even then, flows were subsequently much lower than ✿ in ✿ the previous winter ( Figure 2). The dual-isotopic characteristics of groundwater ( Figure 3) suggest a well-mixed storage. The concurrence of the isotopic composition of this storage with the stream signal indicates that the DMC is a groundwater fed stream. This is also consistent with recent analysis of the catchment flow data (Smith et al., 2020a), and is further supported by the temporal 365 synchrony of stream flow reoccurrence and groundwater recovery (Figure 2 d). The drought intensity value (Table 2) still indicates a substantial deficit in soil and groundwater stores ✿✿✿✿✿✿ storage reflecting incomplete rewetting of the catchment at the end of the study period.

Wider implications
Sandy soil properties and weather-dependent farming make landscapes like the DMC vulnerable to droughts in a continental 370 hydroclimate where dry, hot summers are likely to become more common. Understanding and managing soil moisture in the unsaturated zone of the catchment will be fundamentally important to developing land use strategies that will be resilient in the face of climatic warming. Both crops and trees (Amin et al., 2019) primarily rely on shallow soil moisture storage and there is a clear trade-off between biological productivity, water use and the maintenance of other ecosystem services such as groundwater recharge and river flow generation. It is important that management of crop lands and forests is based on an understanding of 375 how water is partitioned into green and blue water fluxes, so that evidence-based decisions can be made that prioritise water use for the most important, sustainable societal needs. The ephemeral character of the DMC stream -which has been perennial in the past -during the drought, underlined the failure in supporting blue water fluxes that can be of importance for habitat structure, connectivity and water quality. For example, recent recolonization of the catchments by beavers (Castor fiber) has had a major impact on flow regimes and water quality; changes that might not be sustained if the stream becomes ephemeral 380 for longer periods (Smith et al., 2020a). The ✿✿✿✿✿ urgent ✿ need for greater understanding of water security is given urgency with ✿✿✿✿✿✿✿✿✿ underlined the by alarming climate projections for the area (Lüttger et al., 2011). Climate change impacts are already crosssectorally perceived by local land use managers in the North German Plain (Barkmann et al., 2017). The catchment failed to maintain normally expected green and blue water fluxes throughout the growing season and drought of 2018 which made ecosystem services and agricultural land use unsustainable. To improve system understanding and management strategies, 385 further research is needed. The range of different land use types / soils has to be broadened to capture large scale heterogeneity.
Further, the obtained datasets have to be integrated in models to enable quantitative estimations, upscaling to larger areas and extrapolation in time and climatic contexts of these processes. On the basis of our work here, we would argue that insights from ✿✿✿✿✿ water stable isotopes could play a key role in this process. To exploit the potentials of stable water ✿✿✿✿ water ✿✿✿✿✿✿ stable isotopes to gain process insight, studies on finer spatial-temporal resolutions are fundamental. Those field based assessment in a natural 390 setting are the basis to further evolution of ✿✿✿✿ water ✿ stable isotopes as tracers in hydrology. We propose that these studies should be hand in hand with methodological and model development.

Conclusions
We presented an assessment of the 2018 drought and associated ecohydrological feedbacks in a lowland headwater ✿✿✿✿✿✿✿✿ catchment in North East Germany (DMC) using hydroclimatic data in conjunction with isotope-based techniques. The study focused on 395 two plot sites with differing vegetation/soil communities during a period of water stress, when the catchment could no longer sustain blue water fluxes (e.g. stream flow) or green water needs (e.g. crop production) and the subsequent recovery. In general, the forest site "used" more water and was more freely draining and hence had drier soils. Nevertheless, the transpiration dynamics of Oak trees showed some resilience towards drought conditions and appeared to meet atmospheric moisture de- relatively recent rainfall (< 2-3 months age), with age increasing with depth to > 6 months. The deeper forest soil horizons appear to have only old water (∼ ✿ 1 year old). In contrast to the individual isotopic signature of soil water, which took some time to be erased by mixing with winter rainfall, no "memory" -effect    29 Table 6. Soil bulk water MTT estimates, including their α -and β -value, from best fits of gamma function and young water fraction ✿✿✿✿✿ (YWF) from best sine wave fit and associated p-value.