Rooting depth and specific leaf area modify the impact of experimental drought duration on temperate grassland species

Due to climate change, temperate grasslands are being exposed to increasingly severe droughts. Concurrently, land-use intensification is altering grasslands’ functional composition by promoting fast-growing, resource-acquisitive species with high specific leaf area (SLA).

How SLA affects the ability of grassland species to resist and recover from increasingly severe droughts and if deep roots improve their drought performance remains unclear. To investigate this, we established a common garden field experiment including temperate grassland species with SLAs of 17.9–39.3 mm2g-1 and maximal rooting depths of 16–252 cm. After 1.5 years, we simulated droughts for 0, 79, 134, 177 and 220 days.

Drought effects on plant performance increased with drought length, reducing the survival of green tissue and annual biomass by up to ~50% across all 32 species considered. As plant-available water remained in deep soil layers by the end of all treatments, deep roots mitigated the negative effect of increasing drought length on productivity in the later stage of drought and favoured productivity after long drought. The low-to-high SLA trait gradient among the 16 graminoid species seemed to represent alternative survival strategies ranging from dehydration tolerance to dehydration avoidance, rather than drought sensitivity. Variable drought responses along the SLA gradient of forbs imply that multiple other traits are related to drought resistance across evolutionarily distant species.

Synthesis: Our results suggest that deep roots are beneficial for temperate grassland species subjected to longer periods without rainfall when plant-available water is lacking in shallow soil layers but remaining in deep soil layers. In the face of increasing drought severity, we thus recommend (1) fostering deep-rooted species in intensive grasslands on deep, productive soil and (2) directing further studies towards identifying management practices that support deep rooting in semi-natural grasslands.

Methods
We selected 48 plant species from four functionally different seminatural or agricultural Swiss grasslands. In our common garden field experiment, these study species were allocated to four subplot types according to their prevalence in the reference grasslands. This should avoid that species chosen from the more productive reference grasslands would overgrow species from the less productive grasslands. In March 2021, which was 1.5 years after establishement, we covered the whole experimental site with a greenhouse plastic and simulated droughts for 0, 79, 134, 177 and 220 days by omitting the regular watering of the respective plots. 

We measured specific leaf area (SLA) on 12–14 relatively young but fully expanded leaves without obvious signs of herbivory, pathogens, or senescence of 6–12 adult individuals per species growing in control plots in May 2021. For species with both rosette and stem leaves, we measured stem leaves if available and we included all petioles and stipules into the measurements. We checked and completed our SLA measurements in June 2022. Further plant traits were compiled from literature and databases (see references in the README file)

To measure plant performance under and after drought, we measured the biomass of each species in each subplot in three harvests in the year of drought and in one harvest in the post-drought season (file Biomass_data). We further recorded the abundance of each species in each subplot at the start of the experiment, at the start and the end of the respective drought treatment and in spring after the drought (Abundance_data). For these abundance records, we used a frame with 24 15 × 15 cm grid squares at fixed positions and recorded for each species in each subplot the grid squares in which the species rooted.

Out of the 48 selected species, eleven species had only poorly established, for two species adequate maximal rooting depth values were not available, for two species we could not measure SLA adequately and one species had to be weeded when it started to overgrow other species in the subplot. Our analyses are thus based on the 32 species with sla values in the file Species_traits_data