Site description and experimental design
This research was conducted in situ at the two sites of the B4WarmED experiment in northern Minnesota, USA established in 2008. The sites are located at the Cloquet Forestry Center (CFC, 46°40’46” N, 92°31’12” W, 382 m a.s.l.) near Cloquet, MN and the Hubachek Wilderness Research Center (HWRC, 47°56’42” N, 91°45’29” W, 415 m a.s.l.) near Ely, MN in the ecotone of the boreal-temperate forest. The research sites are characterized by a mean annual precipitation and temperature (1980–2019) of 824 mm and 4.9°C for the CFC and 715 mm and 2.8°C for the HWRC (based on nearby weather stations). At each site 24 research plots 3 m in diameter were established, half in relatively open areas that were recently cleared (open canopy) and half in the understory (closed canopy) of existing stands of ≈70 years old mixed aspen-pine-birch with scattered fir, spruce, and other species; in both sites on coarse-textured upland soil. The study includes an incomplete factorial of sites, canopy types, warming and rainfall manipulation, which we analyzed as two overlapping factorial experiments. One experiment consisted of two sites, two canopy conditions (closed and open), and two temperature treatments (ambient and elevated), replicated in three blocks per canopy condition per site. In addition, rainfall was manipulated but only in open canopy plots; thus the second experiment consisted of two sites, one canopy condition (open), two temperature treatments (ambient and elevated), and two rainfall manipulation (ambient and reduced), replicated in three blocks per site.
An open air (chamberless) warming treatment was implemented simultaneously for the above- and belowground part of the plot via an integrated microprocessor-based feedback control system (Rich et al., 2015), designed to maintain a fixed temperature differential between ambient and warmed plots. Infrared ceramic heaters mounted above each plot in an octagonal pattern were used for the aboveground warming of plant surfaces, while resistance-type warming cables were buried (10 cm deep and spaced 20 cm apart) to achieve belowground warming. For more details about the project warming methodology see Rich et al.,(2015) as well as in Reich et al., (2015), Sendall et al.,(2015), and Reich et al., (2016). Aboveground temperature on each plot was measured at mid-canopy height (i.e., roughly the average for all planted tree seedlings in each plot). For temperature measurements below ground (i.e., soil temperature), we used soil temperature probes randomly inserted on each plot at the depth of 10 cm. During the mid-summer and daytime periods, across all 11 years, average temperature differentials between treatments specifically for the above ground were slightly different than the target of 3.3°C (Tables 1 and 2, Figures S1 and S2) but as they were close for the full period of warming treatments, we call the warming treatment +3.3°C throughout the paper.
The summer rainfall reduction treatment began in 2012 via rainout shelters installed only in the open canopy, on randomly selected plots across warming treatments at both sites. Rainout shelters were used to reduce both total summer rainfall and the number of rain events in each year from June 1st to September 30th(for details on rain shelter design and implementation see Stefanskiet al., (2020)). To minimize shading of tree seedlings, rainout shelters were typically deployed during overcast condition or at night shortly before and closed shortly after (typically 0.5-1h) the rain event. Over the course of seven seasons, rain shelters were deployed for an average total time of ~8% of the entire rainfall reduction period (i.e., June 1st – September 30th). In each growing season, about half of this time occurred during night hours (for more information about treatments see Table 1 and Figures S1 and S2). Across the seven years of the summer rainfall removal, we saw an average reduction of 40.7% summer rainfall as compared to ambient plots (Table 1 and Figures S1 and S2). That translated to a reduced mean summer rainfall of 269.5 mm (±15.5 SE) and 222.9 mm (±11.6 SE) as compared to ambient mean realized summer rainfall of 454.5 mm (±26.4 SE) and 376.8 mm (±19.9 SE) at the CFC and HWRC sites respectively. Consequently, our rainfall treatments were representative of relatively wet summer (~70thpercentile wettest) and rather dry summer (~10th percentile driest) for ambient and reduced rainfall respectively as compared to the broader temporal context of the 100 year of the weather record (1912-2011 availiable for the CFC site). Soil moisture on the research plots was monitored over the course of this research using water reflectometers (Model CS616 from Campbell Scientific). Soil Volumetric Water Content (VWC - cm3 water/cm3 soil) was measured across 0-30 cm soil profile on an hourly basis through all years. The reduction of soil moisture by warming treatment is a result of increased evapotranspirative demand due to elevated temperatures that reduces soil moisture on a continuous manner as we warm 24/7 from early spring until late fall. In contrast to warming, rainfall reduction treatment is implemented on a cyclic demand basis that allows replenishing soil water in between rainfall removal events (see Table 1 and Figure S2 for more details).
Over the course of this experiment between both the open and closed canopy we grew seedlings of 17 native and four invasive tree species (a total of 21) in different combinations among years and canopies (for details see Tables 3 and S2). Seedlings were sourced from local ecotypes, well suited for the research sites typical environmental conditions. We planted one or two year old seedlings produced by MN DNR (Minnesota Department of Natural Resources) nurseries into an existing matrix of native vegetation. The chosen species represent dominant tree species from the boreal-temperate ecotonal region of northern Minnesota. Newly planted cohorts were given one year (≈14 months) to acclimate after transplant before any gas exchange measurements were performed except for the 2012 and 2013 cohorts when plants were measured in the same growing season following spring planting (but see below on requirements of foliage selection for the measurement). The observations reported in this study were made throughout all years of the experimental operation from 2009 to 2019 on different cohorts of seedlings that ranged from two to eight years of age (See Tables 3 and S2 for more details).