Biodiversity loss has been repeatedly demonstrated to decrease community resistance to colonisation, increasing their sensitivity to invasive species. Based on these results, Elton’s hypothesis proposed that biodiversity loss reduces the competitive ability of plant communities by emptying ecological niches and, thus, increasing the availability of unused resources. Yet, direct evidence for the effect of diversity loss on resource use and the consequences for the community competitive ability remains scarce, especially under natural conditions.

We created a species richness gradient (1 to 27 species on average) simulating a realistic species loss through the long-term (6 years) removal of rare and subordinate species in a species-rich oligotrophic wet meadow. To assess the effects of species richness on the competitive ability of communities, we transplanted phytometers of two species (the grass Holcus lanatus and the forb Plantago lanceolata) into plots varying in species richness and into gaps where plant competition was eliminated. We then compared phytometers’ performance (relative growth rate) and functional traits responses (vegetative height, leaf area, leaf nitrogen and phosphorus content, specific leaf area, and leaf dry matter content). We also tested the effect of targeted species richness on local light interception, soil nutrient content, soil water content, and soil surface temperature and whether they could reliably explain phytometers’ performance and trait responses.

Both phytometer species showed a much lower relative growth rate in the vegetation than in gaps demonstrating strong responses to plant competition. However, the effects of species richness were surprisingly weak as all plant communities ranging from monocultures of a dominant species to the richest communities (27 species on average) exerted a similar competitive pressure. Surprisingly, almost none of our measures of local resource depletion explained trait or growth rate responses.

Synthesis. Phytometers effectively quantified competitive pressure exerted by natural plant communities. However, realistic loss of rare and subordinate species had only limited effects on community competitive ability, and these effects were not explained by changes in local resource depletion. This highlights the complexity of diversity-competition relationships and cautions against inferring invasion resistance from species richness alone.

Design of long-term removal experiment

The study was conducted in an oligotrophic wet meadow in South Bohemia region, 10 km southeast of České Budějovice, Czech Republic (48.953 N, 14.593 E, 510 m a.s.l.). The mean annual temperature of this locality is 9°C with the mean annual precipitation 725 mm (data averaged across years 1990 – 2021 available from: https://www.meteoblue.com, accessed November 20, 2024). It is a very species rich meadow with over 30 species per 1 m² dominated by Molinia caerulea, Danthonia decumbens and Betonica officinalis (alliance Molinion caeruleae). This meadow is nutrient-poor with intermittently wet soils, partially drying out especially during the summer period. This meadow is traditionally extensively managed by mowing once a year at the peak of the vegetation season.

We used a long-term diversity removal experiment established in 2016 in the study site which simulates a realistic species loss scenario. In addition to control plots corresponding to the intact vegetation, four species richness levels were created by the manual removal of rare and subordinate species, keeping the 1, 3, 6 and 12 most abundant species. These 5 treatments were replicated 5 times in 1 × 1 m plots organised in a Latin square. Plots have been maintained twice a year (mid-May and the end of August) by manually removing the aboveground biomass of non-targeted species with the efforts to remove also their roots. More details about the experimental design could be found in Lisner et al. (2023).

When the phytometer experiment started in 2021, all the plots had recovered from the disturbance associated with the initial species removal – as demonstrated by the relatively stable biomass production in all plots (see Lisner et al. 2024) – and the maintenance of the species richness levels required the removal of only a few plant individuals representing a small coverage of the vegetation.

The phytometer experiment

We used the grass Holcus lanatus and the forb Plantago lanceolata as phytometers. Both species are perennial hemicryptophytes, resident and typical for the study site. Both species are typical especially for European temperate meadows and mesic pastures. They are generalists to temperature, moisture and nutrients and resistant to intermediate disturbances (www. pladias.cz, accessed June 12, 2026; Chytrý et al., 2021). Moreover, they are known to be successful in germinating and establishing in natural communities and are thus ideal model species (Dietrich et al. 2013; Janíková et al. 2024; Tammaru et al. 2021; Wilfahrt et al. 2021). Seeds from a commercial supplier (Planta Naturalis, Czech Republic) were sown on March 8, 2021 into peat jiffy pots (4 cm of diameter) and grown in a climatic room at 20°C under a 12 h :12 h day-night cycle. While seeds of P. lanceolata started to germinate on March 10, H. lanatus seeds started to germinate on March 14 but in very low amount, thus, we repeated the sowing of H. lanatus on March 18 with more success in germinating seeds on March 24.

On April 14, 2021, we transplanted four individuals of each species into every plot of the long-term removal experiment. P. lanceolata seedlings were 5 weeks old while H. lanatus seedlings were 3 weeks old. Hence, all transplanted individuals were approximately the same age and reached a similar development stage in each species. Phytometers were planted with their jiffy pots by coring a hole of the appropriate size, one individual of each species in each of the four inner edges per plot. In addition, we transplanted phytometers into 0.3 × 0.3 m competition-free gaps artificially created next to each row of the experiment (2 gaps per block), where above- and below-ground parts of the vegetation were removed. In each gap, we planted two phytometers per species (thus, we had also four individuals per species for this type of plot per block).

On June 8, 2021 (one week before the biomass harvest of the long-term removal experiment), we harvested aboveground biomass of planted phytometers (100% individuals survived) and measured their functional traits after 55 days in the field (spring period hereafter). As most phytometers resprouted from their belowground parts (92% and 47% individuals of P. lanceolata and H. lanatus, respectively), we conducted a second harvest on August 9, 2021, after two months (62 days) of growth (summer period hereafter). Overall, we specifically followed the species growth patterns during their first year to follow the establishment of new individuals under different species richness contexts.

Phytometers’ relative growth rates

The relative growth rate of phytometers was calculated as:

where M₀ and M_t are the aboveground initial and final dry mass, respectively, for the spring period (t = 55) and for summer period (t = 62).

To account for the differences in size between individuals at the time of transplantation, we predicted M₀ in spring using non-destructive traits. Calibration curves were produced using a training dataset corresponding to a selection of 30 individuals, representative of all plant sizes at the time of the transplantation. For H. lanatus, we used plant height in cm (h), the cumulative length of all leaves in cm (l), and their interaction as predictors of M₀ in mg. A multiple linear regression model based on the selected individuals provided the following estimates:

For P. lanceolata, the measured M₀ was predicted using h, number of leaves (n), and their interaction. Similarly, a multiple linear regression model based on the selected individuals provided the following estimates:

The predictive models were validated using an independent data set composed of 15 extra individuals that were not used to estimate the parameters. The high goodness of fit (r² = 0.94 and 0.75 for H. lanatus and P. lanceolata, respectively) confirmed the quality of our models (see Fig. S1).

For summer, we used the measured aboveground biomass at the spring harvest as M₀. Therefore, the value calculated in summer cannot be directly interpreted as a measure of relative growth rate. Nevertheless, we hypothesized that individuals with large aboveground biomass were also larger belowground, allowing a better resprout after the harvest.

Phytometers’ traits

During each harvest, we measured six functional traits of phytometers which are expected to respond to light, soil nutrient and water depletion and can be affected by temperature. Vegetative height was measured as the highest non-stretched leaf. The area of a leaf (LA) was based on scans of two undamaged and fully developed leaves per individual analyzed in ImageJ 1.x (Schneider et al. 2012). Specific leaf area (SLA, the ratio between LA and dry mass) and leaf dry matter content (LDMC, the ratio between the dry and water saturated masses) were measured on the same leaves. We also measured the nitrogen and phosphorus contents in leaves (LNC and LPC, respectively). To ensure enough material for chemical analyses, we combined the leaf biomass of the four individuals growing in each plot. Leaf chemical analyses were conducted at the research centre laboratory in Chomutov of the Crop Research Institute, Czech Republic. Total nitrogen [% in dry matter] were determined by dry combustion according to Dumas method (Dumas 1826), and total phosphorus [% in dry matter] after mineralizing using Mehlich III extractant solution (Mehlich 1984).

Estimates of resource depletion and temperature

We determined soil nutrient content by collecting a soil-core of 1.5 cm in diameter and 20 cm deep in each species richness plot at the time of the biomass harvest (i.e., in mid-June 2021). Soil samples were air-dried, sieved through a 2 mm mesh and milled. Samples were then analyzed at the research centre laboratory in Chomutov of the Crop Research Institute, Czech Republic, to measure the soil nitrogen and phosphorus contents. The Mehlich III extractant solution (Mehlich 1984) was used to measure the dissolved soil phosphorus content, while the total soil nitrogen content was determined according to Kjeldahl method (Kjeldahl 1883).

The light interception by the canopy was calculated from data measured by the SS1 SunScan Canopy Analysis System device (Delta-T Devices Ltd). The device compares the incident to the transmitted photosynthetic active radiation (PAR) at the ground level using 64 sensors embedded in a 1 m long probe. We conducted two orthogonal measurements per each species richness plot and repeated measurements four times in each plot in spring (April 21, May 11, May 26, and June 9, 2021) and summer (June 16, June 29, July 22, and August 9, 2021). The light intercepted by canopy was calculated as:

The values for each period were averaged for subsequent analyses.

Soil water content and ground-level temperature were measured using TMS dataloggers (TOMST) equipped with temperature and soil moisture sensors. These dataloggers collected data every 15 minutes, thus we averaged this data for each species richness plot. The temperature that we used was measured in °C at the level of the soil surface. Although the temperature is not a real resource that can be depleted but rather an environmental variable, we were using it together with the estimates of resource depletion hereafter. Water content received from dataloggers is not the volumetric water content but signal values which should be calibrated according to appropriate soil type. For our use, we did not calibrate the data and use only the signal values from the dataloggers because we looked at the relative differences and not at the absolute values.

Data analyses

All statistical analyses were conducted on R version 4.3.3 (R Core Team 2024).

            For each species, we calculated the response of RGR and traits to different levels of plant species richness as:

 where X_plot is the mean value of the four individuals collected in each plot varying in species richness (1, 3, 6, 12, and control plots) and X_gap is the mean value of the four individuals growing in the competition-free gaps in the same block. A response of 0 means no change compared to gaps while a negative, respectively positive, response means a decreased, respectively increased value compared to gap. To test whether RGR and trait responses differed from 0, we used linear mixed-effects models (LMM) in the R package “lmerTest” (Kuznetsova et al. 2017) with a t-test on the intercept (Table S1) and considering the block as a random factor to account for the variation caused by possible heterogeneity in the experimental plots.

            We run LMM in the R package “lme4” (Bates et al. 2015) with Type II Wald Chi-square test of analysis of variance (ANOVA) from the R package “car” (Fox & Weisberg 2019) to test firstly the effect of targeted species richness on RGR and functional traits responses of phytometers and secondly the effect of targeted species richness on measures of resource depletion. We used log-transformed targeted species richness as the expression of diversity, i.e., 1, 3, 6, and 12 species for the corresponding removal treatments, and 27 species for control plots (i.e., the mean species richness observed in control plots during the monitoring of the vegetation in June 2021). Targeted species richness was treated as a continuous variable. The block identity was considered as a random factor. The analyses were run separately for RGR and each trait response, each phytometer species, and each season, as well as for each measure of resource depletion. For data visualisation (Figs 2 and 3), we used the R package “ggplot2” (Wickham 2016).

            To test the direct and indirect effects of targeted species richness on phytometers’ responses through the measures of resource depletion, we constructed structural equation models (SEM) using the R package “piecewiseSEM” (Lefcheck 2016) for each phytometers’ response significantly responding to targeted species richness in the previous LMM analysis. For each species and season, we established a common a priori model where the targeted species richness is hypothesised to affect measures of resource depletion, which in turns, affects a given phytometers’ response (Fig. S2). These models were constructed based on the results from previous studies (Table 1). Each phytometers’ response was analysed separately. Individual structural equations (i.e., links) were fitted by a LMM with the block identity as a random factor. The ability of the SEM to represent well the data was tested by Fisher’s C statistics.