A new study published in Hydrogeology Journal offers the most mechanistically complete answer yet. Led by Omar Ashraf Nimr of the University of Oulu in Finland, together with colleagues at Flinders University in Australia and collaborators across the Finnish research community, the research used a fully integrated three-dimensional physics-based model called HydroGeoSphere to simulate how rewetting propagates through a boreal fen called Matorovasuo, located in Kittilä in Northern Finland roughly 170 kilometres north of the Arctic Circle. Unlike the sparse point measurements that dominate conventional restoration assessments, the model resolved water-table dynamics across the entire 224-hectare mire and its 556-hectare headwater catchment at daily resolution over a full hydrological year, generating millions of spatially explicit data points that no field campaign could ever collect.

The study site is far from an arbitrary choice. Matorovasuo is a groundwater-fed aapa mire, a gently sloping, patterned fen typical of the northern boreal zone, designated in 2026 as a UNESCO ecohydrology demonstration site. Forestry drainage beginning in the late 1960s carved an approximately 49-kilometre network of ditches across the peatland, in places spaced as little as 10 metres apart, altering around 60 percent of the area. Restoration under the EU LIFE PeatCarbon programme, one of the northernmost peatland restoration efforts ever attempted, was carried out in 2024. Workers removed encroaching trees guided by comparisons with a 1957 aerial photograph, filled roughly 6.8 kilometres of ditches with logs placed longitudinally and another 4.1 kilometres with compacted peat, and constructed 92 transverse dams, seven of logs and 85 of peat, concentrated along two major north-south collector ditches that had historically intercepted and diverted water away from the mire’s interior.

The modelling framework itself represents a methodological leap. HydroGeoSphere solves surface and subsurface flow simultaneously within a single implicit system, coupling three-dimensional variably saturated groundwater flow described by the Richards equation with two-dimensional overland and channel flow governed by the diffusion-wave approximation of the Saint-Venant equations. The computational mesh contained 114,147 triangular elements, refined down to half-metre resolution near ditches and restoration structures, while the vertical domain resolved peat into three sublayers capturing the steep decline of hydraulic conductivity with depth, plus layers of glacial till extending down to bedrock mapped with ground-penetrating radar. Calibration drew on twelve IoT-enabled monitoring wells transmitting data via LoRaWAN, eddy covariance measurements of evapotranspiration from a tower within the mire, and daily meteorological forcing from two nearby weather stations of the Finnish Meteorological Institute. Performance metrics were solid: a Kling-Gupta efficiency of 0.61 for water-table depth, 0.62 for evapotranspiration, and 0.78 for snow depth, with an uncertainty analysis confirming that the peat hydraulic conductivities controlling the headline results were constrained by the calibration data with 95 to 98 percent variance reduction.

The central result is striking. Simulated restoration elevated the domain-scale median water table by approximately 23 centimetres. More remarkable still, areas classified as undisturbed, the wet open sedge fens at the heart of the mire where no direct interventions occurred, gained roughly 22 centimetres on average, with local rises exceeding 40 centimetres. This finding demolishes the implicit assumption that restoration benefits are confined to visibly ditched zones. Historical drainage, the authors show, casts a hydrological shadow far beyond the ditch lines themselves, and blocking those ditches re-establishes lateral connectivity that allows recovery to propagate deep into areas previously assumed to be beyond help. In a sloping fen system where marginal ditches cut across the slope and isolate downslope peat from up-gradient water sources, restoring that lateral flow is arguably the single most important hydrological outcome.

Evidence for this reorganisation came from an unexpected analytical direction: variogram analysis, a geostatistical technique borrowed from mining and petroleum engineering that quantifies how quickly values become dissimilar with distance. Under drained conditions, the effective correlation range of mean annual water-table depth was 345 metres; after simulated restoration it stretched to 459 metres. The effect was most pronounced along the principal drainage direction, where the correlation length expanded from 265 to 471 metres. In plain terms, the drained peatland was a fragmented landscape of locally coherent drawdown zones decoupled from one another at longer ranges, while the restored peatland behaved as a hydraulically connected whole. The water table, once sliced apart by the ditch network, had regained the spatial coherence characteristic of an intact mire.

The study also revealed that the reach of any given restoration structure depends critically on what it is made of and how it is built. Peat-filled ditches, with their very low permeability, act as intense hydrological plugs: they generated strong water-table rises immediately adjacent to the fill, but the effect decayed exponentially, losing about 70 percent of its magnitude within roughly 40 metres, a relationship the exponential model fitted with an R-squared of 0.98. Dams told a different story. By creating discrete zones of impoundment that raise upstream water levels and drive lateral head gradients through connected peat, dams produced broader, plateau-like responses extending up to 100 metres, though with greater spatial variability. Log-filled ditches fell somewhere in between, displaying a non-monotonic pattern: a weak near-field response within 10 metres, followed by recovery to a stable plateau of about 25 centimetres beyond 30 metres, reflecting the superposition of local attenuation and broader catchment-scale redistribution. A single leaky log-fill accomplishes little on its own, but a network of them accumulates roughness and slows water across the landscape, generating what the authors describe as a synergistic retention state.

Geomorphology emerged as a third decisive control. Water-table rise varied quadratically with peat thickness, peaking at roughly 30 centimetres where peat was about 1.8 metres deep and declining on either side of that tipping point. Too thin, and there is insufficient storage capacity for restoration-induced head increases to persist; too thick, and internal hydraulic damping disperses the signal over a larger volume. Distance from the peat-mineral soil boundary followed an exponential growth relationship with a fit of R-squared 0.92: near the margins, thin peat over mineral substrate acts as a hydraulic sink that dissipates rewetting gains, while the peatland interior, beyond about 150 metres from the boundary, retains water and allows restoration signals to accumulate. Peat formation slope showed an inverse quadratic relationship with an optimum near 3 percent, where the peatland is already well buffered and gains from rewetting are limited. Even seasonality mattered: wet-period responses nearly doubled those of dry subarctic winters, with the water-table difference between restored and drained scenarios peaking at about 33 centimetres during spring snowmelt and autumn rainfall, and dipping to around 18 centimetres in late winter.

The authors are careful to frame their numbers as order-of-magnitude benchmarks rather than universal constants, conditioned on the hydrogeology, restoration design, and climate of a single site. The restored configuration represents a long-term hydrological equilibrium projection rather than a prediction of the near-term transient that monitoring wells are currently recording, and processes such as peat swelling, so-called mire breathing, and progressive Sphagnum regrowth remain outside the model’s scope, likely causing it to overestimate short-term responsiveness. Yet the identified mechanisms, connectivity-driven redistribution, scale-dependent attenuation, and geomorphic control, arise from general hydrological principles and should transfer across comparable boreal fens. For a field where restoration projects are typically funded for four to five years and evaluated with a handful of piezometers, the study provides something genuinely new: a mechanistic, design-oriented framework for anticipating where, when, and why rewetting will work. As nations race to restore degraded peatlands under binding climate commitments, knowing that the choice between a peat plug and a dam can change the radius of recovery by a factor of two and a half is not academic detail. It is the difference between spending restoration budgets where water will actually stay, and spending them where it will quietly drain away.

Subject of Research: Integrated surface–subsurface modelling of water-table dynamics and rewetting propagation in a restored boreal fen peatland

Article Title: How does rewetting propagate through restored peatlands? An integrated surface–subsurface modelling analysis of water–table dynamics

Article References: Nimr, O. A., Marttila, H., Batelaan, O., Partington, D., & Ala-Aho, P. (2026). How does rewetting propagate through restored peatlands? An integrated surface–subsurface modelling analysis of water–table dynamics. Hydrogeology Journal. https://doi.org/10.1007/s10040-026-03173-5

Image Credits: AI Generated

DOI: 10.1007/s10040-026-03173-5

Keywords: peatland restoration, rewetting, boreal fen, water table, groundwater modelling, HydroGeoSphere, ditch blocking, hydrogeology, carbon emissions, variogram analysis, LIFE PeatCarbon, Finland