More extreme and frequent drought periods reduced crop production and altered stable isotope ratios of C and N in plants

2.1. Experimental set-up

The Austrian Agency for Health and Food Safety (AGES) lysimeter facility consists of 18 filled gravitation lysimeters representing the three main soil types of the Marchfeld, namely, calcaric Phaeozem (Ps), calcic Chernozem (Ch), and gleyic Phaeozem (Pg) (Gerzabek and Krenn 2005; Stenitzer and Hösch, 2005). The gleyic Phaeozem, a groundwater-influenced soil type, had become dry and experienced unnatural shrinking and crack formation. The main soil chemical properties of the upper 10 cm of lysimeter soils were redetermined in 2019 and are presented in the Supplemental Table 1. Further soil properties can be found in Tataw et al. (2014) and Watzinger et al. (2023). Rainfall was simulated with an overhead irrigation system since 2011 (Tataw et al., 2014, 2016). For the “current” scenario, half of the lysimeters were irrigated according to current precipitation patterns, as calculated by the 30-year mean of rainfall (amount and distribution) in Groß-Enzersdorf, Marchfeld (517 mm). The other half of lysimeters in the “predicted” scenario were watered according to the precipitation pattern predicted for a dry year (20 percentile) for the period 2071–2100 in the Pannonian region (436 mm) (Tataw et al., 2016). From 2018 onward, an adapted and improved model, the historic (1961–2000) and the hot and dry model scenario presented in Seidl et al. (2019), was used to calculate the “current” scenario and the “predicted” very dry year (5 percentile) scenario (531 and 386 mm precipitation per annum, respectively) (Watzinger et al., 2023). The lysimeter station was covered by a plastic tunnel and irrigated during the growing season, and it remained uncovered in winter when it received ambient precipitation. The period of coverage and irrigation treatment followed the precipitation scenarios for dry (2017) and very dry (2018, 2019) years, except that a heavy rainfall event was simulated on the 25^th of April 2018 and 60 mm of precipitation was applied additionally. The amounts of precipitation and irrigation applied in the years 2017–2019 are illustrated in Figure 1.

Figure 1.

The crop rotation and crop residue management throughout the experiment followed the common agricultural practices in the Marchfeld region (Table 1). Before sowing and after the harvest, the soil was manually tilled. Fertilizer was applied according to Austrian standard recommendations (BMLFUW, 2017) (Table 2). N fertilizer was added as calcium ammonium nitrate (NAC 27 N, Borealis LAT GmbH, Austria) with a stable isotope composition δ¹⁵N of −1.2‰. Visual controls for pest infestations were done on a regular basis to identify potential insect pests and to allow for appropriate measures to be taken. No significant insect infestations were detected for the duration of the experiments described. However, weeds were quite abundant and were removed manually.

Aboveground plant biomass was harvested at the end of the vegetation period for crop and weed biomass determination and for further elemental and isotopic analyses in the laboratory. Plants were dried at 60°C for dry biomass determination. Cereal crops were separated into three different plant components – straw, grain, and chaff. On 06.07.2017, the second leaves after the flag leaves were sampled, pooled, and analyzed. Analysis of plant biomass and C and N isotope composition was conducted by Elemental Analyzer – Isotope Ratio Mass Spectrometry (EA-IRMS; elemental analyzer coupled to Delta V Advantage; Thermo Scientific, Bremen, Germany).

2.2. Statistical analysis

Two- or three-way analysis of variance (ANOVA) type III sum of squares analyses using soil type, precipitation treatment, and/or year as the main factors were run, and Tukey HSD (Honestly Significant Difference) tests were conducted. Generally, no transformation of raw data was necessary to meet the assumptions of normality and variance homogeneity. For statistical analysis, the R package Jamovi was used (R Core team, 2020; The jamovi project, 2021, Gallucci, 2019).

3.1. Response of plant biomass production to drought

More frequent and intense drought periods significantly decreased the total biomass and grain yield of crops and the harvest index (proportion of grain biomass to total aboveground biomass) in our lysimeter experiment (Tables 3 and 4, Figure 2, Supplemental Figure 1, and Supplemental Table 2). Moreover, earlier plant maturation was observed under the predicted precipitation scenario. Grain yields were highest in the gleyic Phaeozem, followed by the calcic Chernozem and the sandy calcaric Phaeozem. However, the impact of drought was strongest in both Phaeozem soil types compared to Chernozem (40% vs. 30% biomass reduction in spring barley and 40% vs. 10% biomass reduction in winter wheat). The grain yield declined most strongly in the sandy Phaeozem (60% reduction). These results are in line with current knowledge of the response of cereals to water shortage as observed in our experiment. The soil water content in the upper 10 cm determined during several points throughout the vegetation periods frequently approached the wilting point (−1.5 MPa) (Watzinger et al., 2023). It has been repeatedly reported that water limitation during grain filling reduces photosynthesis and accelerates leaf senescence in cereals, resulting in reduced availability of assimilates and a shortening of the grain-filling period (Dupont and Altenbach, 2003; Barnabás et al., 2008; Mariem et al., 2021). The lack of photoassimilates as well as reduced activity of starch-synthesizing enzymes inhibit grain starch production, which is one of the leading causes of reduced crop yields under water deficit (Worch et al., 2011). Higher water-holding and nutrient provision capacity of Chernozem likely buffered the impact of drought, as already reported by Tataw et al. (2016). Formation of cracks in the gleyic Phaeozem, which has a comparable water holding capacity to Chernozem, enabled preferential water flow, and hence likely hindered the storage of water. Our results were well in accordance with those presented by Tataw et al. (2016) from the same experimental site, who found significant reduction of grain yield and harvest index.

Figure 2.

In contrast to crops, weeds were less affected, which would underpin current knowledge that a genetically diverse community can adapt to environmental stress easier than a single species (Klaus et al., 2016) (Tables 3 and 4, Supplemental Table 2). Also, mustard biomass, which grows as a catch crop in winter, was not significantly affected by the soil type and precipitation scenarios. In contrast, the biomass of mustard even increased by 40% in Chernozem. It appeared that plants growing over the winter period (mustard and winter wheat) were not negatively affected by the previous drought periods when the current and predicted precipitation manipulation treatments were not in place, that is, both treatments received the same amount of precipitation. Especially plants in the Chernozem even profited from this, which was also observed by Evans and Burke (2013). We suspect that nutrients in general and N specifically accumulated in the soil due to reduced plant uptake in late spring and summer. Accumulation of N under the predicted precipitation regime in this experimental site was indicated using a stable isotope labeling approach (Watzinger et al., 2023).

The variation of plant growth between the lysimeters was generally high due to variations within the irrigation system. This was caused by differences in the water pressure along the irrigation lines and the drizzlers (2018: current 464 ± 38 mm, predicted 383 ± 20 mm; 2019: current 401 ± 32 mm, predicted 284 ± 14 mm) (Figure 1). Moreover, replications of lysimeter × treatment combinations were low (n = 3). In addition, the intensity of irrigation was high (2017: 86.4 mm h⁻¹ and 2018 onward: 76.2 mm h⁻¹), mimicking rather extreme rainfall events with intermittent dry spells. Moreover, some lysimeters were affected by animal activities (bioturbation by mice, burrowing activities by rabbits, dogs, and crows), especially in 2017, and there was an island effect, as lysimeters were surrounded by bare soil. These experimental issues might have led to the observed high variability and lower plant growth than expected under comparable field conditions.

Our data has to be contextualized within the constraints of unaltered temperature and CO₂ concentration manipulation and bearing in mind that we had mimicked rather extreme drought years.

Cereal production models including temperature and CO₂ concentration, besides precipitation, simulated lower reduction of yield or even increased yield for the same region in an average climate change year: spring barley 2% to >10% and winter wheat 0%–10% reduction (Eitzinger et al., 2013); winter wheat around 10 % yield increase and spring barley 18%–21% (Alexandrov et al., 2002; Eitzinger et al., 2008); and winter wheat 18%–28% yield increase (Thaler et al., 2021). Yield increase was mainly attributed to CO₂ fertilization. Without CO₂ fertilization, the yield of winter wheat is expected to decrease by around 12% in 2080 (Eitzinger et al., 2008). In addition, the predicted precipitation pattern chosen for our lysimeter study simulated the driest year out of five (20 percentile) in 2017 and the driest year out of 20 (5 percentile) in 2019. Precipitation manipulation was only commenced during the growing season; thus, the predicted higher precipitation during winter season was dismissed in our experimental set-up. Thaler et al. (2021) have presented data from the 10 percentiles and found that the increase of yield was less pronounced (11%–23%).

3.2. Alteration of the plant δ¹³C value

The increase of plant δ¹³C in the predicted precipitation scenario verified increased stomatal resistance and reduced photosynthetic C assimilation (Table 5, Supplemental Tables 2–5). Delta ¹³C values were also higher in sandy Phaeozem, followed by gleyic Phaeozem and Chernozem, which was also likely related to the soil water-holding capacity. Straw δ¹³C values were in line with those reported for leaves by Tataw et al. (2016) on the same experimental site. δ¹³C values were generally higher in grain than in leaves as reported earlier (Craufurd et al., 1991). The difference between straw and grain δ¹³C values was only significantly affected by the year/crop species. The difference was highest in winter wheat in 2019 (up to 3‰) and lowest in spring barley in 2018. This difference might have either derived from physiological differences of the plant species (Domergue et al., 2022) or be explained by a higher difference in water availability during leaf and grain development in 2019 than in 2018, when plants had received an additional 60-mm heavy rainfall event in late April.

3.3. The plant N content and its delta¹⁵N values

Grain N content of spring wheat was significantly elevated under drought in 2017. In the following years, the N content did not change significantly between the current and predicted precipitation scenarios, indicating no loss in the quality of grain (Supplemental Tables 2–5 and Supplemental Figure 2). Most importantly, N content in straw and mustard was higher under drought, which was likely driven by an accumulation of available N in the soil system due to lower plant N uptake, while N fertilization remained unchanged and N mineralization was less reduced (Watzinger et al., 2023, Evans and Burke, 2013) (Supplemental Tables 2–5 and Supplemental Figure 3). Higher straw N content likely facilitated the translocation of N into grains. In addition, the mobilization and translocation of proteins to cereal grains over the translocation of starch is favored under drought stress (Mariem et al., 2021), which might have also come into play in our experiment, thus balancing the grain quality between the treatments.

Crop δ¹⁵N values were significantly elevated in gleyic Phaeozem and Chernozem in comparison to the sandy calcaric Phaeozem soil, with a significant decrease in 2019 (Figure 3, Supplemental Tables 2 and 4 and Supplemental Figure 4). This likely resulted from a higher N uptake in the Pg and Ch soil and the higher percentage uptake of soil-born N with a higher δ¹⁵N value (6.9‰) than mineral N fertilizer (−1.2‰). Plant δ¹⁵N decreased in 2019, which can be attributed to three applications of N fertilizer in comparison to only one fertilization event in the previous years. Grain was ¹⁵N enriched by around 1‰ in comparison to straw. Enrichments are related to relocation of N from leaf to grain during grain filling (Kanstrup et al., 2011; Zhou et al., 2013), while provision of mineral N fertilizer during grain filling leads to lower fractionation due to lower relocation of N (Fuertes-Mendizábal et al., 2018). Accordingly, ¹⁵N enrichment in grain was less in winter wheat, which received three applications of mineral N fertilizer, in contrast to only one mineral fertilizer application in April and May during spike formation in spring wheat and spring barley in 2017 and 2018. Significant differences in plant δ¹⁵N values between precipitation scenarios were rare and observed in leaves in 2017 and grain in 2019. Then, δ¹⁵N values were lower under drought, possibly indicating higher percentage of mineral fertilizer N usage (Bol et al., 2005; Choi et al., 2017) and higher N remaining from soybean plant residues (N-fixing plant species with low δ¹⁵N values) incorporated in 2016. Decreased plant δ¹⁵N values have also been attributed to more closed N cycling, that is, low N input/output versus N turnover in soils. A closer N cycle has been seen in natural ecosystems and increasing mean annual precipitation (Amundson et al., 2003), while the use of mineral fertilizer might trigger the loss of N under drought due to asynchronicity of fertilizer application and plant uptake (Ullah et al., 2020). However, as we have already documented lower soil N losses but similar N turnover rates under drought in this experiment (Watzinger et al., 2023), we concluded that δ¹⁵N values in the plant were rather driven by the source coinciding with the accumulation of mineral N fertilizer.

Figure 3.