The coated seeds were produced by Arevo and SweTree Technologies under the name SeedPAD (afterwards SP). SP consists of a single seed of Scots pine (Pinus sylvestris) covered in vermiculite and wrapped in polysaccharide foil with a diameter of 35 mm and thickness of 3.5 mm (Fig. 1). In contact with water, the polysaccharide dissolves, attaching the vermiculite to the underlying mineral layer, thus immobilizing and protecting the seed. The SPs used in the following experiment were either fertilized or unfertilized. Fertilized SPs included the commercial fertilizer arGrow in the coating of vermiculite material, i.e., the fertilizer adds 10 mg nitrogen in the chemical form of the amino acid arginine and 5.5 mg phosphorus as phosphate to each SP.
Both fertilized and unfertilized SPs were deployed in May–June of 2017 to 12 clearcut forest sites across Sweden between the latitudes of 59°N and 67°N (Fig. 2, Table 1). The sites were chosen because of their dry characteristics, locations where Scots pine is commonly planted. Considering the large latitudinal gradient, deployment time had to be adapted to local growing conditions. Therefore, southern sites were planted during May and northern sites during June after the snow melted and at the onset of growing season. Before deployment, mechanical site preparation was performed on all sites in the form of disc trenching. This is a standard scarification process which involves making long furrows and exposing the mineral soil underneath. The SPs were deployed on the exposed mineral soil and marked with sticks next to the deployed pads. Deployment, scarification and marking were performed by forest landowners, i.e., forest companies or private forest owners who were members of a private forest owners’ association. Orchard seeds used in the SPs were proprietary to each company and had an average of 98% germination capacity percentage. Non-research parts of clearcuts were regenerated with nursery-grown seedlings.
Table 1 Description of the 12 experimental forest clearcut sites listed from north to south. Mean annual precipitation across the sites was in the range 500–700 mm (2010–2019). Sites with plot type “both” contained plots of both circular and rectangular type On each of the 12 regenerated clearcut sites, survey plots were established in the form of circles or rectangles (both 100 m2) with the location completely randomized (Table 1). Plots were sufficiently away from clearcut edges (> ~ 20 m) to avoid edge effects and thus did not include a buffer zone. Within each of the circular survey plots either 35 fertilized or 35 unfertilized SPs were placed directly on the mineral soil that had been cleared of entire organic dominated layer by the process of soil scarification. The rectangular plots consisted of two parallel 50 m rows, one with 50 fertilized SPs and one with 50 unfertilized SPs deployed directly on the mineral soil. At the sites where both types of plots were established, we had enough buffer area between circular and rectangular plots, and therefore, we used both plot types as similar replicates at each site in the subsequent analyses.
Establishment rate and growth of the seedlings were recorded for each of the two regeneration methods in late August and September of 2018 and 2019. Seedlings were recorded as surviving when there was a live and vigorous seedling next to a marker stick. Growth was measured as distance from the ground to the top of the shoot. Scots pine seedlings growing within a radius of 10 cm around marker sticks were counted as originating from our SPs to account for the possible slight movement of seeds due to precipitation. Marker sticks with either dead (characterized as dried out with brown needles) or missing plants within the 10 cm radius were recoded as non-surviving. The establishment rate was estimated at the plot scale, as the number of live seedlings per the total number deployed SeedPADs. In addition, in 2019 five seedlings from each treatment plot within each clearcut were carefully excavated and taken for biomass estimation. These samples were dried for 24 h at 60 °C (to a constant weight) cut at the stem base and then the root and shoot parts weighed separately. Based on the harvested samples, we developed allometric equations for roots and shoots using height as a predict variable for each treatment at each site. In contrast to establishment rate, height was recorded on randomly chosen 20 seedlings across all plots for each treatment in each site. Thus, the fertilization effect within a site was not assessed on height or biomass due to non-plot-replication, therefore we only assessed the effect of fertilization across sites.
To assess the cross-site response of establishment rate to fertilization, we recorded environmental factors for each of the 12 sites; type of vegetation present on the clearcut according to National Inventory of Landscapes in Sweden—NILS (Ståhl et al. 2011) and topsoil composition, using visual assessment. Climate data were downloaded from the open database of the Swedish Meteorological and Hydrological Institute. We used data on daily precipitation, temperature, wind, topsoil type, and dates of the onset and end of the vegetation period (a daily mean air temperature ≥ 5 °C) from the nearest available climate monitoring station for each site (mean distance between sampling site and climate station was 32 km and the maximum distance was 46 km). Because we considered that weather conditions around the time of seed deployment and during the growing season after germination, are important factors, we extracted weather data for the first six weeks after deployment in 2017, and during the growing season in 2018 and 2019.
For all the statistical analyses we used R-Studio software (R Core Team 2019). We analyzed response of establishment rate, separate from height and biomass, because the sampling unit of the former is a replicated plot while the latter has no replication at a plot scale. For establishment rate, we assessed the response using a generalized linear model following binomial distribution with logit-link function. After finding no significant effect of the plot shape (circular or rectangular) or its interaction with site, we set fertilizer-treatment, site and their interaction as independent variables, and establishment rate as response variables. For this analysis we used the 10 sites with both fertilizer-treated plots and control plots at each site. Due to an unbalanced dataset (unequal numbers of observations for each treatment), we employed a type III ANOVA model using the car package in R, investigating the effect of fertilizer addition while considering the interaction with site. After concluding that there was no interaction, we employed type II ANOVA. Next, we examined the effects of site-specific weather variables on seedling survival in control plots and fertilized plots, using a generalized linear model across all 12 sites. In order to examine the causality, stepwise selection procedures were employed, based at a level of α < 0.05, and the new model was checked with Akaike Information Criteria (AIC) against the previous one. When values of AIC differed by less than 2, the model with fewer degrees of freedom was selected. After several steps this results in a final model including only significant variables. For the response of height and biomass, we performed a type II ANOVA model, without interaction term between fertilization and site. The normality of all models’ final residuals were checked visually by plotting them against predictions.
Since precipitation during the first six weeks following deployment of the SPs showed a significant positive effect on seedling establishment, we performed a follow up experiment in a controlled laboratory environment to investigate the relationship between water addition and SP attachment to the ground. The aim was to examine how much water was required for dissolution of the coating and to attach SPs to the ground. We used 135 SPs in total, separated into three different water addition rates combined with three different soil types. The soil used was collected in February 2020 from a forest clearcut site in mid-Sweden and dried in the lab, then sifted to give three different grain sizes. Water addition rates were determined based on local precipitation data averaged for all sites. The grain sizes were set to 1.7, 5.6 and 10.0 mm diameter for fine, medium, and coarse soil, respectively. SPs were placed on a pile of each soil type and room temperature water was dripped onto them using a pipette. At each step, 0.5 ml of water was dripped directly onto the SPs every 1.5, 3 and 5 min for fast, medium, and slow water addition rates, respectively. Dissolution was considered successful when the SP was firmly attached to the ground and could not easily be pushed sideways from the soil without breaking apart.