Satellite Observations Reveal a Positive Relationship Between Trait-Based Diversity and Drought Response in Temperate Forests

Abstract

Climate extremes such as droughts are expected to increase in frequency and intensity with global change. Therefore, it is important to map and predict ecosystem responses to such extreme events to maintain ecosystem functions and services. Alongside abiotic factors, biotic factors such as the proportion of needle and broadleaf trees were found to affect forest drought responses, corroborating results from biodiversity-ecosystem functioning (BEF) experiments. Yet it remains unclear to what extent the behavior of non-experimental systems at large scales corresponds to the relationships discovered in BEF experiments. Using remote sensing, the trait-based functional diversity of forest ecosystems can be directly quantified. We investigated the relationship between remotely sensed functional richness and evenness and forest drought responses using data from temperate mixed forests in Switzerland, which experienced an extremely hot and dry summer in 2018. We used Sentinel-2 satellite data to assess aspects of functional diversity and quantified drought response in terms of resistance, recovery, and resilience from 2017 to 2020 in a scalable approach. We then analyzed the BEF relationship between functional diversity measures and drought response for different aggregation levels of richness and evenness of three physiological canopy traits (chlorophyll, carotenoid/chlorophyll ratio, and equivalent water thickness). Forest stands with greater trait richness were more resistant and resilient to the drought event, and the relationship of trait evenness with resistance or resilience was hump-shaped or negative, respectively. These results suggest forest functional diversity can support forests in such drought responses via a mixture of complementarity and dominance effects, the first indicated by positive richness effects and the second by negative evenness effects. Our results link ecosystem functioning and biodiversity at large scales and provide new insights into the BEF relationships in non-experimental forest ecosystems.

Keywords: Sentinel‐2; biodiversity–ecosystem functioning (BEF); ecological monitoring; functional diversity; plant traits; remote sensing.

FIGURE 1

Study area of canton Aargau (west) and canton Zurich (east) and location in Switzerland (top left). Highlighted on the map is the Sihlwald site, where we validated the drought response results. The true color composite shows the study area in summer 2017, based on June/July Sentinel‐2 data. Map lines delineate study areas and subregions and do not necessarily depict accepted national boundaries.

FIGURE 2

Calculation of diversity metrics from traits within the calculation area (top left). Shown is an example translation of the 60‐m radius (blue circle) neighborhood area to a mask for the calculation (bottom left). The numbers indicate the weighting of each pixel in calculating the value of the center pixel. Concepts of diversity metrics (right) in three‐dimensional trait space. Richness (Ric) (top right) and evenness (Eve) (bottom right). The traits considered include chlorophyll content (CHL), carotenoid/chlorophyll ratio (CCR), and equivalent water thickness (EWT).

FIGURE 3

Trend of the mean normalized difference water index (NDWI) in the study area between 2017 and 2020. The numbers in the legend represent the mean percent changes for the three drought response measures (change 2017 to 2018 resistance, change 2018 to 2019 recovery, and change 2017 to 2020 resilience) across the entire study area in northern Switzerland.

FIGURE 4

Average diversity of 21 subregions with the plot on the left showing their median richness and evenness. It is important to note that the variation within the regions is large, and the differences between regions are comparatively small (see Figure S6). The subregions shown on the right were obtained by grouping the forests of the study area according to the intersection of (1) canton (Aargau [AG] and Zurich [ZH]), (2) geographical regions (Central Plateau [Eastern and Western], Rhine plains, Jura, and Pre‐Alps), and (3) four, respectively seven, cantonal forest districts. Blue‐green colors represent canton AG, and red‐yellow colors represent canton ZH. The color gradients range from southern to northern regions within the cantons.

FIGURE 5

Subregion‐corrected drought resistance, recovery, and resilience (left to right) in relation to functional richness and evenness. These data were binned into 20 bins along richness and evenness and into the 21 subregions, resulting in 8400 bins. We then fitted subregion to correct for subregion differences and finally related the thus corrected drought responses to functional richness and evenness using multiple regression (as described in Section 2). Resistance and resilience increased with richness. Resistance showed a hump‐backed relationship with evenness, while resilience decreased with evenness.

FIGURE 6

Variance explained by the linear model combining the influence of the diversity metrics richness and evenness on resistance (change in NDWI during the drought 2017–2018, Table S2), recovery (change in NDWI after the drought 2018–2019, Table S3), and resilience (change in NDWI after the full 2‐year observation period 2017–2020, Table S4). The bars from top to bottom in each panel are the contributions to the r ² values of linear richness (ric), log‐transformed richness (logric), evenness (eve), evenness squared (eve²), the 21 subregions (REG), and interactions of the diversity metrics and subregions (ric × REG, eve × REG). Note that all contributions are significantly larger than zero. The formulae for the fitted linear models are listed in R notation, with N representing the number of pixels per bin.