Rewetting does not return drained fen peatlands to their old selves

The study addresses whether rewetting drained temperate fen peatlands restores them to near-natural conditions in terms of biodiversity, ecosystem functioning, and land cover characteristics, and how quickly this occurs. Peatlands, which store ~30% of global soil carbon on ~3% of land, have been widely drained for agriculture, forestry, and peat extraction, leading to CO2 emissions, land subsidence, eutrophication, and biodiversity loss. Rewetting is promoted for climate mitigation and biodiversity restoration, with the Paris Agreement implying rewetting of ~500,000 km² of drained peatlands by 2050–2070. However, profound peat physical and geochemical changes caused by drainage, combined with elevated nutrients and altered hydrology, may hinder rapid recovery. The authors quantify restoration outcomes across temperate Europe by comparing 320 rewetted fen sites with 243 near-natural sites, contextualized by 10,000 additional fen vegetation plots, to test for convergence towards near-natural conditions and to identify typical trajectories after rewetting.

Background literature shows intact peatlands provide major ecosystem services, including large carbon storage, water regulation, and specialized biodiversity. Drainage causes peat oxidation responsible for ~5% of anthropogenic GHG emissions, nutrient release, and subsidence. Rewetting can promptly reduce net carbon loss and may reestablish carbon sequestration, but drainage alters peat hydraulic properties (increased bulk density, reduced porosity and conductivity) and elevates nutrients, potentially causing unstable water tables, episodic inundation, and methane peaks. Recovery towards pre-drainage states occurs primarily where disturbance was weak. Prior work emphasizes the sensitivity of rewetted systems to weather extremes and suggests that management strategies from natural systems may not directly apply to rewetted fens. The concept of novel ecosystems may capture the trajectory of heavily altered fens following rewetting.

Design: Comparative analysis of 320 rewetted fen peatland sites and 243 undrained, near-natural sites across the major temperate fen regions of Europe. Rewetted sites had a documented rewetting action (e.g., drain blocking) resulting in mean annual water tables ≥ −25 cm. Mean time since rewetting was 9 years (range 1–54) at sampling. Pre-rewetting land use was agriculture (80%), forestry (10%), or peat extraction (10%). Near-natural sites had no known drainage history based on expert knowledge and field/remote sensing verification.

Response clusters and data collection:

• Vegetation: Complete vascular plant and bryophyte species lists with percent cover in plots (median 16 m²; range 12–25 m²). Total of 539 species; mean richness 15.1 species per plot.
• Hydrology: Water table depth relative to surface from piezometers with loggers or manual readings (≥1 full year, biweekly/monthly; >80% measured >2 years; average duration 2.3 years). Metrics: annual median, minimum, maximum, and amplitude of water table.
• Geochemistry: Pore-water pH and electrical conductivity (0–60 cm), and topsoil (0–30 cm) bulk density and organic matter content, sampled in summer alongside vegetation surveys. Geochemistry available for a subset (57 sites) spanning the wider geographic range.
• Land cover characteristics: 208 spectral–temporal metrics from Sentinel‑2 A/B optical time series (2018) across 10 spectral bands/indices (e.g., MNDWI), using all cloud- and shadow-free observations (median 45 clear-sky observations per pixel). Metrics averaged over 3×3 pixels around plot centers; tests with other aggregations gave similar results. Inclusion of Sentinel‑1 metrics did not affect outcomes.

Comparative framework: Sites providing ≥2 response clusters were included. Vegetation data were contextualized with 10,000 randomly selected plots classified as fen/mire EUNIS habitat types from >90,000 plots in the European Vegetation Archive.

Analyses:

• Ordinations: Non-metric multidimensional scaling (NMDS) for each response cluster. Dissimilarity: Bray–Curtis for vegetation (with step-across for no shared species), Euclidean for hydrology, geochemistry, and land cover metrics. Group differences tested with ANOSIM.
• Variance: Compared pairwise distances within groups; significance via permutation tests (diffmean, 1000 permutations) due to non-independence and potential heteroscedasticity.
• Pairwise counterpart analysis: For each rewetted site, assigned the spatially closest near-natural counterpart of similar peatland origin (hydrogenetic/ecological type), biogeographic region, and (for mountains) altitude; pairs used to compute dissimilarity versus time since rewetting. LOESS (span 1.4) and linear models assessed temporal trends; near-natural-to-near-natural pairs provided a baseline dissimilarity (horizontal dashed lines).
• Software: R with vegan, labdsv, simba, and FORCE framework for EO processing.

Quality control: Taxonomic standardization for vegetation (Euro+Med, bryophyte checklist), exclusion of incomplete species lists, and verification of site histories and hydrological data coverage.

• Rewetted fen peatlands differ significantly from near-natural counterparts in vegetation composition and geochemistry, with smaller but significant differences in hydrology and land cover characteristics (ANOSIM: vegetation R=0.35, p<0.001; geochemistry R=0.13, p=0.040; hydrology R=0.02, p<0.001; land cover R=0.02, p<0.001).
• Many rewetted sites fall outside the near-natural range: 63% (vegetation), 44% (geochemistry), 20% (hydrology), 21% (land cover) outside the near-natural 95% confidence ellipse in NMDS.
• Helophytisation is widespread post-rewetting: higher relative frequency of EUNIS tall helophyte beds (25.5% rewetted vs 6.2% near-natural; χ² p<0.001), and 66% higher cover sum of tall helophytes in rewetted sites. Indicator species include Typha latifolia and Phalaris arundinacea.
• Plot-scale diversity is lower in rewetted sites across Hill numbers; Shannon diversity: 1.46 ± 0.04 (rewetted) vs 1.75 ± 0.04 (near-natural), p<0.001.
• Brown mosses with high peat-formation potential are largely absent where tall helophytes dominate.
• Hydrology and soil physical differences consistent with drained-phase degradation: median water table higher (rewetted mean median 5.0 cm above surface vs 1.5 cm below in near-natural), greater amplitude (+15%), and higher maximum water tables; topsoil bulk density increased (+61%) and organic matter content decreased (−18%) in rewetted sites.
• Land cover metrics show greater variability in MNDWI and near-infrared reflectance in rewetted sites, reflecting more open water and altered vegetation structure.
• Rewetted sites are generally more variable than near-natural sites across all response clusters, indicating reduced stability and potential sensitivity to climatic extremes.
• Little to no temporal convergence toward near-natural conditions up to ~30 years post-rewetting across all clusters. Linear models show significant intercepts (p<0.001) but non-significant slopes (p>0.05) for hydrology, geochemistry, land cover; vegetation shows a weak slope (p=0.013) but predicted Bray–Curtis dissimilarity remains high (0.85 after 5 decades; 95% CI 0.64–1.07).
• Restoration success depends on more than maintaining a water table close to the surface; some sites with near-surface water tables remained highly dissimilar from near-natural counterparts.

The findings demonstrate that rewetting drained temperate fens frequently leads to persistent alternative states dominated by tall graminoid wetland plants, with long-lasting differences in vegetation, hydrology, geochemistry, and land cover relative to near-natural fens. This outcome addresses the central question by showing that simply restoring high water levels does not generally return systems to their pre-drainage biodiversity and functioning, at least within several decades. Elevated nutrients and altered peat physical properties from prior drainage likely promote helophyte dominance and suppress bryophytes and low vascular plants via light competition. Rewetted sites exhibit greater variability and heightened sensitivity to weather extremes, implying reduced self-regulation compared to near-natural peatlands. The lack of temporal convergence suggests that restoration outcomes are set early; sites either align with near-natural conditions promptly or follow divergent trajectories for decades, consistent with the concept of locally novel ecosystems. These trajectories have implications for greenhouse gas dynamics, as helophyte aerenchyma can either increase or suppress methane emissions, and the absence of peat-forming brown mosses may affect long-term carbon accumulation. Therefore, management goals should acknowledge the likelihood of novel ecosystem states and focus both on climate benefits from halted peat mineralization and on developing process-based strategies to steer ecosystem functions under these new conditions.

Across a continental-scale dataset, rewetted temperate fen peatlands generally do not revert to near-natural states in biodiversity and ecosystem functioning within three decades. Rewetting often induces helophytisation, altered hydrology, and geochemistry, leading to stable alternative community and function states with greater variability. The study highlights that restoration success hinges on more than water table elevation and that early post-rewetting dynamics may determine long-term trajectories. For policy and practice, this implies adjusting expectations toward novel ecosystem outcomes while prioritizing climate benefits from stopping ongoing carbon losses. The authors call for interdisciplinary, process-based research—including microbiome and biogeochemical assessments—and coordinated data sharing to better predict and manage rewetted peatlands. Future research should identify determinants of immediate resemblance vs divergence, quantify full greenhouse gas balances across vegetation states, and test management interventions that may promote desired functions (e.g., peat formation, biodiversity goals).

• Geochemistry analyses were based on a subset of sites (57), limiting statistical power for this cluster.
• Near-natural sites are among the least disturbed available but may not be pristine; site histories rely partly on expert judgment and remote/field verification.
• Pairwise dissimilarity matrices involve non-independent observations; permutation-based tests mitigate but do not eliminate this issue.
• The study spans temperate European fens; results may not generalize to bogs or tropical peatlands.
• Time since rewetting covers up to ~30 years; longer-term convergence cannot be excluded but was not detected.
• The study did not directly measure greenhouse gas fluxes across all sites; inferences on GHG dynamics are indirect via vegetation and hydrology proxies and referenced studies.