Site selection and flooding data
To select sites with variable flooding regimes at small and large spatial scales, we used previously collected data on inundation frequency for the Rastmyran wetland (Åhlén, Jarsjö and Hambäck 2023). In that study, inundation frequency (the proportion of days when each 2x2 m2 pixel was under water) was estimated across the wetland during the normally unfrozen period from the end of March to the end of October using a combination of water table measurements in the field and extrapolation through a digital elevation model. Water tables were estimated with a HOBO water table logger placed at the in- and outlets of the fen and the digital elevation model was constructed based on data from the Swedish Mapping, Cadastral and Land Registration Authority (for details see Åhlén, Jarsjö and Hambäck 2023). For the purpose of this study, we selected sites that included both homogeneously dry, homogeneously wet and mixed wet/dry conditions. As a measure of inundation heterogeneity during the selection process, we contrasted the inundation frequency in the central pixel (2x2 m2) and peripheral pixels (within 6x6 m2) (Fig. 3a). Please note that wet and dry periods are somewhat separated in time, with wetter conditions during spring and drier conditions during summer and autumn.
Dry pixels included those with inundation frequency (IF)<0.45, which corresponds to a dry situation during most of summer, whereas wet pixels included those with inundation frequency>0.65. The distribution of IF-values across the dry and wet classes was then per definition bimodal in the central pixel (wet=high IF>0.65, dry=low IF<0.45; Table 1, column 2), but this bimodality did not always translate to peripherical pixels outside the central pixel, which exhibited a more heterogeneous character (Table 1, columns 3-5). Thus, the inundation frequency<0.45 in homogeneously dry sites both in the central pixel and for most pixels within 6x6 m2, and correspondingly >0.65 for homogeneously wet sites. In dry-wet sites, the central pixel had an inundation frequency<0.45 whereas most peripheral pixels had an inundation frequency>0.65, and correspondingly for wet-dry sites. Finally, we included dry-wet/dry and wet-wet/dry sites, meaning that peripheral pixels had variable inundation frequencies with some wet and some dry. Thus, we selected 6 categories with different dry-wet conditions and 10 sites per category (Table 1).
Post selection, to further investigate scale effects, we estimated the mean inundation frequency in successively larger areas (10x10 m2 … 26x26 m2, Fig. 3b) surrounding the central pixels (not including the central pixel). We also calculated the proportion of dry pixels within each scale to estimate the occurrence of dry refuges in the vicinity. The inundation frequencies and proportion of dry pixels (IF<0.45) were strongly correlated (r>0.97) and we therefore dropped the inundation frequency from analysis. Finally, a heterogeneity index was calculated for each site from the proportion of dry to non-dry pixels using the Shannon Diversity Index (\(H^{{}^{\prime}}=-\sum{p_{i}*ln(p_{i})}\),pi = proportion of dry pixels), where low values indicate lower heterogeneity and large values indicate high heterogeneity.

Environmental variables

In each site, we collected data on vegetation height (August) and soil organic content (June). Vegetation was measured within each central pixel by recording the maximum vegetation height at three randomly selected points. Soil samples were collected using a soil core sampler, placed in bags and transported to a laboratory freezer until further processing. The soil organic content was determined using the loss on ignition method. Samples were dried in an oven at 105°C for 24h. After drying, the samples were sieved using a 2mm sieve and weighed on a microscale to obtain the initial weight. The samples were then placed in a high-performance oven at 550°C for 4h. After this, the samples were reweighed, and the loss of ignition was calculated using the formula: [(initial weight-after weight)/initial weight] x100. From this, the soil organic content was derived.