Groundwater extraction reduces tree vitality, growth and xylem hydraulic capacity in Quercus robur during and after drought events

Climate change is expected to pose major direct and indirect threats to groundwater-dependent forest ecosystems. Forests that concurrently experience increased rates of water extraction may face unprecedented exposure to droughts. Here, we examined differences in stem growth and xylem hydraulic architecture of 216 oak trees from sites with contrasting groundwater availability, including sites where groundwater extraction has led to reduced water availability for trees over several decades. We expected reduced growth and xylem hydraulic capacity for trees at groundwater extraction sites both under normal and unfavourable growing conditions. Compared to sites without extraction, trees at sites with groundwater extraction showed reduced growth and hydraulic conductivity both during periods of moderate and extremely low soil water availability. Trees of low vigour, which were more frequent at sites with groundwater extraction, were not able to recover growth and hydraulic capacity following drought, pointing to prolonged drought effects. Long-term water deficit resulting in reduced CO2 assimilation and hydraulic capacity after drought are very likely responsible for observed reductions in tree vitality at extraction sites. Our results demonstrate that groundwater access maintains tree function and resilience to drought and is therefore important for tree health in the context of climate change.


Fig. S1: Location of the study region (a), the three study regions near Freiburg, Emmendingen and Lampertheim (b), and locations of the investigated forest stands in each region (Lampertheim, Emmendingen and Freiburg) (c, d and e).
Stand IDs in (c), (d) and (e) are acronyms denoting the study region and site type (first and second letter, respectively) of each investigated stand. F, E, L for Freiburg, Emmendingen and Lampertheim, respectively. W indicates sites without groundwater extraction, C upland sites and E, S and N are extraction sites (E for the extraction sites in Emmendingen and Freiburg, S and N for the Southern and Northern extraction sites, respectively, in Lampertheim). Numbers in Stand IDs indicate the average tree age in each stand. The maps (a-e) were created using QGIS version 3.14-pi (https://qgis.org/en/site/). Satellite images in c-e were obtained using the QuickMapServices Plugin (https://plugins.qgis.org/plugins/quick_map_services/) in QGIS.  Colours indicate site types (red for groundwater extraction sites, blue for noextraction sites and grey for upland sites). Bars with light colours indicate the maximum lifespan of trees per stand while the darker colours indicate the period during which vessel variables where measured. Vertical dashed lines coloured in red indicate the onset of groundwater extraction at the extraction sites. Stand IDs are acronyms denoting the study region and site type (first and second letter, respectively) of each investigated stand. F, E, L for Freiburg, Emmendingen, and Lampertheim, respectively. W indicates sites without groundwater extraction, C upland sites and E, S, and N are extraction sites (E for the extraction sites in Emmendingen and Freiburg, S and N for the Southern and Northern extraction sites, respectively, in Lampertheim). Numbers in Stand IDs indicate the average tree age in each stand.

Supplementary Methods
Crown vitality assessment: Tree vitality was assessed based on the crown conditions of target trees as proposed by Roloff [1]. Vital trees with full crowns and vital fine twigs were assigned to crown class 0; trees assigned to class 1 showed first signs of crown degradation and twig abscission; trees of class 2 were noticeably weakened already; and seriously weakened and dying trees with dead main branches and disintegration of the entire crown were assigned to crown class 3. Selected trees covered all vigour classes present in the studied stands.
Statistics used to assess the quality of developed chronologies: Mean gleichläufigkeit (synchronicity) (Mglk) was used to assess similarity of detrended tree-ring series [2]. The glk() function from the dplR package was used to compute glk (gleichläufigkeit) which performs pairwise comparison of all possible combinations between series. The expressed population signal (EPS) is an indicator of how well a chronology represents a theoretical infinite population [3]. Low values of EPS (commonly <0.85) indicate that the chronologies are dominated by individual tree signals rather than a consistent regional signal [4]. Rbar is the mean correlation between series within a chronology and is a measure of common signal strength of detrended chronologies. The signal to noise ratio (SNR) is a measure of the desired signal in each chronology versus the amount of unwanted information and random variation [4,5].

Climate data and calculation of the Standardized Precipitation
Evapotranspiration Index (SPEI): Monthly resolved temperature and precipitation data were acquired from the German Weather Service (Deutscher Wetterdienst) using the meteorological stations closest to study sites (<20 km). For sites near Freiburg and Emmendingen, data (from 1921 to 2016) from one station located within 10 km from the sites were used. For the Lampertheim region, meteorological data were available for the period between 1897 and 2016. Based on these meteorological data, we calculated the Standardized Precipitation Evapotranspiration Index (SPEI) [6] using the SPEI package in R [6]. An accumulation period of 6 months was selected for the calculation of SPEI because we found in an earlier study that this accumulation period correlated best with inter-annual tree-growth variations across the investigated sites [7].

Drought event identification:
For the identification of drought events at our study sites we used the SMI1.8 for the time-period between 1977 (8 years after the onset of groundwater extraction in Lampertheim) and 2015. Years were classified based on SMI1.8 averages of the vegetation season into: "extreme drought", "severe drought", "moderate drought" and "not dry" using a percentile approach (see also [8]) to classify years based on their historical frequency. Accordingly, extremely dry years are events with a likelihood of occurrence of < 5% of the time (between 1977 and 2015).
Severely dry years have a 5-10% and moderately dry years a 11-20% chance of occurrence. With the percentile approach we identified two extreme drought events for each region (Supplementary Fig. S.4).