Long-term protection in grasslands enhances soil carbon storage via reduced disturbance and community trait diversity-environment adaptations
Abstract
Understanding how ecological disturbance and plant community traits regulate soil carbon storage is critical for predicting ecosystem feedbacks to global change and designing sustainable land-use strategies. However, the processes by which disturbance regimes mediate the trade-offs between species preservation and soil carbon storage are difficult to predict due to their complexity and remain debated, particularly in comparison to protected systems.
We employed a paired-site design, sampling 30 long-term managed grasslands and their paired nature reserve counterparts across a gradient of three grassland types (wet, mesic, and dry) in Central Europe (Czechia). At each site, we quantified soil carbon stocks, characterized soil chemical properties, measured aboveground biomass production, and assessed plant community composition through species diversity and functional trait analyses.
We found that protected grasslands store significantly more carbon than their conventionally managed counterparts, driven by reduced anthropogenic disturbances, which promote ecosystem stability and enhance nutrient retention, coupled with plant community traits that favour carbon storage.
Our results show that dry grasslands accumulated more carbon than mesic or wet types, likely due to trait-mediated stabilization and constrained microbial activity under aridity. Despite higher biodiversity in protected areas, soil carbon stocks were uncorrelated with species richness, revealing a potential indirect decoupling of species richness and carbon sequestration. Notably, biomass-carbon correlations persisted in managed grasslands but vanished in protected systems, indicating divergent dynamics of productivity from storage under undisturbed conditions. The environmental indicators predicted carbon stocks in protected grasslands, whereas managed systems relied on community characteristics and acquisitive traits.
This study demonstrates that protected grasslands support both richer biodiversity and larger soil carbon stocks compared to commercially managed sites. The findings that underscore the ecological benefits of sustained protection and limited disturbance, providing valuable insights for land management and restoration strategies within existing conservation programs that integrate biodiversity protection with carbon sequestration, such as the Natura 2000 Network.
Dataset DOI: 10.5061/dryad.jq2bvq8q7
Description of the data and file structure
This dataset was generated as part of a study investigating the effects of long-term ecological protection versus conventional management on soil carbon storage, biodiversity, and ecosystem functioning in Central European grasslands.
The research employed a paired-site design across 30 grassland pairs (60 sites total) in the Czech Republic. Each pair consisted of one long-term protected grassland (Nature Reserve) and one neighboring commercially managed grassland, stratified across three moisture-based vegetation types: Dry, Mesic, and Wet. Data collection took place during the 2024 season.
The dataset integrates field measurements of soil properties (carbon stocks, nitrogen, pH, electrical conductivity), aboveground biomass, vegetation community composition, and derived plant functional traits and ecological indicator values. The goal was to identify the mechanisms (disturbance history, environmental filtering, plant functional strategies) driving differences in soil carbon sequestration between protected and managed ecosystems.
The associated manuscript, "Long-term protection in grasslands enhances soil carbon storage via reduced disturbance and community trait diversity–environment adaptations," provides the full ecological context, hypotheses, and interpretation of these results.
Files and variables
Primary Data File: Data.xlsx
Sheet Name: Data
Key Variables and Column Descriptions:
- Site Identifiers & Classification:
- Code_Site, Site_no, Site: Unique identifiers for each sampling site.
- District: Geographic region (Palava, Bile.Karpaty, Jihocesko).
- Grassland_Type: Management regime (Nature.Reserve or Managed).
- Pair: Pair identifier (P1 to P30), linking a protected and a managed site.
- Vegetation_type: Moisture category (Dry, Mesic, Wet).
- Aboveground Biomass:
- Biomass_Total_N (%), Biomass_Total_C (%): Nitrogen and Carbon concentration in aboveground biomass.
- Biomass_[g/m^2^]: Aboveground dry biomass per square meter.
- Disturbance History:
- No_Disturbance_Records: Number of recorded disturbance events (e.g., mowing) from historical land-use analysis.
- Since_last_Disturbance_[year]: Years since the last major disturbance.
- Soil Carbon & Nitrogen Stocks (by depth):
- Soil_C_Stock_05 (kg/m2), Soil_C_Stock_15 (kg/m2), Soil_C_Stock_30 (kg/m2): Soil organic carbon stock at 0-5 cm, 5-15 cm, and 15-30 cm depths.
- Soil_N_Stock_05 (kg/m2), Soil_N_Stock_15 (kg/m2), Soil_N_Stock_30 (kg/m2): Soil nitrogen stock at 0-5 cm, 5-15 cm, and 15-30 cm depths.
- Soil_Total_N_[%], Soil_Total_C_[%]: Total nitrogen and carbon concentration in the soil profile.
- Soil_C_Stock (kg/m2), Soil_N_Stock (kg/m2): Total soil carbon and nitrogen stock integrated over 0-30 cm depth.
- Soil_CN ratio: Soil Carbon to Nitrogen ratio (unitless).
- Basic Soil Properties:
- pH: Soil pH measured in a 1:5 soil:water solution.
- EC (mS): Soil Electrical Conductivity (in millisiemens per Centimeter), indicating salinity/soluble salts.
- Plant Community Ecological Indicators (Ellenberg Indicator Values - EIV): (No unit, just a value)
- Community-weighted mean (CWM) values reflecting species' ecological optima (unitless, just a value):
- EIV_Light, EIV_Temperature, EIV_Moisture, EIV_Reaction (pH), EIV_Nutrients.
- Plant Functional Traits (Community-Weighted Means - CWM):
- Aboveground: CWM_LDMC (Leaf Dry Matter Content)[mg/g], CWM_SLA (Specific Leaf Area) [mm^2^/mg].
- Belowground (Clonal & Root Traits): CWM_Primary_root [presence or absence], CWM_Persistence_of_connection [years], CWM_Clonal_spread [cm], CWM_Mean_Root_diameter [mm], CWM_Root_dry_matter_content [g/g], CWM_Root_N_concentration [mg/g], CWM_Root_mycorrhizal_colonization [%], CWM_Specific_root_length [m/g].
Code/software
All statistical analyses were performed using R statistical software version 4.4.2 (R Core Team, 2024). Differences in carbon stock and other soil properties (pH, EC, total nitrogen, and C: N ratio) between grassland types across distinct soil profiles were analyzed using ANOVA (with function aov), with grassland type and soil profile as fixed effects and the identity of grassland pairs as a random effect. We further conducted separate ANOVAs for carbon stock across all soil profiles, where grassland type was treated as a fixed effect and grassland pair identity as a random effect. The effect of vegetation type (wet, mesic, dry) on total carbon stock was also assessed using ANOVA, with vegetation type and grassland type as fixed effects and grassland pair identity as a random effect.
Additionally, we applied a linear mixed-effects model (LMM) using the function aov to test the effect of grassland type and plant biomass (fixed effects) on total carbon stock while accounting for grassland pair identity as a random effect. Within each grassland type, total carbon stock was also correlated with plant biomass using the lm function.
To explore further drivers of total carbon stock, we performed separate linear regressions for each grassland type using either ecological indicator values or functional trait composition (expressed as community-weighted means, CWM). Full regression models (with all EIV values or with the CWM of all the above-mentioned traits) were first tested for overall significance. If significant, stepwise selection using the step function was applied to identify the most parsimonious model, starting with backwards selection and allowing for the potential re-inclusion of previously excluded variables. The significance of the final selected predictors was evaluated using the summary function. For visualization of these relationships, we displayed standardized regression coefficients of the reduced model, but for the non-significant variables, we showed standardized regression coefficients from the full model to provide a full list of relationships for all variables. Residuals of all tests were visually checked for normality, and in the case of ANOVA, also tested for homoscedasticity by the Bartlett test. These assumptions of the used tests were met in all cases.