Climate warming and projected increase in summer droughts puts northern peatlands under pressure by subjecting them to a combination of gradual drying and extreme weather events. The combined effect of those on peatland functions is poorly known. Here, we studied the impact of long-term water level drawdown (WLD) and contrasting weather conditions on leaf phenology and biomass production of ground level vegetation in boreal peatlands. Data was collected during two contrasting growing seasons from a WLD experiment including a rich and a poor fen and an ombrotrophic bog. Results showed that WLD had a strong effect on both leaf area development and biomass production, and these responses differed between peatland types. In the poor fen and the bog, WLD increased plant growth, while in the rich fen, WLD reduced the growth of ground level vegetation. Plant groups differed in their response, as WLD reduced the growth of graminoids, while shrubs and tree seedlings benefited from it.

In addition, the vegetation adjusted to the lower WTs, was more responsive to short-term climatic variations. The warmer summer resulted in a greater maximum and earlier peaking of leaf area index, and greater biomass production by vascular plants and Sphagnum mosses at WLD sites. In particular, graminoids benefitted from the warmer conditions.

The change towards greater production in the WLD sites in general and during the warmer weather in particular, was related to the observed transition in plant functional type composition towards arboreal vegetation.

The study site was located at Lakkasuo in southern Finland (61°47’N, 24°18’E). Based on its ecohydrology, Lakkasuo can be divided into a mesotrophic fen, an oligotrophic fen and an ombrotrophic bog, hereafter called rich fen, poor fen, and bog. In 2000–2001, a water level drawdown (WLD) experiment was initiated to mimic the long-term effects of lower water tables (WT) on the ecosystems. Within the three sites, an intact control area and a WLD area were established, located 20 m apart. The drainage treatment lowered the WT by an average 6.4 cm in 2017 and 7.5 cm in 2021 when compared with the control areas. To disentangle the potential effects of short-term changes in growing conditions from the long-term influence of peatland drying/drainage, we compared observations collected during two growing seasons with contrasting weather conditions, i.e., summers of 2017 (cold and wet) and 2021 (warm and dry). 

The seasonal development of leaf area index (LAI: m² m^-2) was estimated for each vascular plant species. Sampling was done throughout both growing seasons from early May to late August–September at regular intervals, six times in 2017, and five times in 2021. Tree seedlings and saplings (up to 0.5 m in height) were considered part of the ground vegetation cover and were included in the LAI estimations. In each sample plot, the number of leaves per vascular plant species were counted from five permanent representative subplots (0.08 × 0.08 m). However, large plant species with infrequent or uneven distribution were counted at the plot-level. For the rich fen WLD area, which had smaller circular plots, leaf number was counted at the plot-level. Simultaneously, the mean area per leaf was measured for each plant species. Representative leaves (varying in age and size) were collected adjacent to the plots and scanned with a portable leaf area meter (LI-3000, Li-Cor, UK). Leaf area values were multiplied by the leaf counts to determine the LAI value per species per plot. Finally, the species-wise LAI estimates were summed to provide an estimate of the total plot-level LAI estimate of all leaves present in the plot, which was used in subsequent analysis. To assess the impact of WLD and the two contrasting growing seasons on the phenological development of LAI, we used a nonlinear mixed effects model with a log normal unimodal function with parameters that described the leaf area phenology in each plot. 

Growth over the growing season was defined for each Sphagnum species using the cranked wire method. Cranked wires were installed into intact homogenous patches of each species within each area to serve as a fixed reference point. Sphagnum growth was measured as the increment of individual Sphagnum stems along the wire from spring to autumn. In addition, shoot density and the weight (g) of 1 cm of stem were defined in Sphagnum patches adjacent to the measurement plots in July in both years. To estimate Sphagnum biomass production, the increment (cm) was multiplied by relative coverage and dry weight per cm stem to calculate the annual biomass production of each Sphagnum moss species (g per growing season) and were subsequently summed per plot. Leaf biomass production of vascular plants was calculated by converting LAI to biomass using specific leaf area (dry mass per leaf area, g m^-2). The specific leaf area was defined for five replicates per species per site per WLD/control area in July of both years. Species-specific biomass production was quantified as the difference between minimal biomass at the start of the growing season and maximal biomass at peak growing season and was subsequently summed to calculate total and PFT-specific leaf biomass production. Specifically, the PFTs were graminoids (sedges and grasses), forbs, shrubs, and tree seedlings (up to 50 cm height). The effects of WLD and year on biomass growth were analysed using linear mixed effects (LME) models, in which we included sample plot as a random effect to account for repeated measurements. For biomass production we fitted three LME models with either the biomass production of the total ground level vegetation, the Sphagnum mosses, or vascular plants as the response variable, and site, year and WLD treatment and their interactions as fixed effects. The composition of vegetation in terms of vascular PFTs was highly variable between sites; this was verified by testing biomass production of each vascular PFT with an LME model with site as a fixed effect. Because of the strong influence of the site, we continued to analyse the biomass production of each vascular PFT with site-specific LME models, with year, WLD treatment and their interaction as fixed effects. Normality and homogeneity of variances were verified with Kolmogorov-Smirnov and Levene’s Tests, respectively, and with diagnostic plots. Where needed, variance structures were applied (VarIdent) or the data were log normal (ln) transformed.