Previous studies of Eriophorum vaginatum have detected adaptational lag in response to climate change. We revisit this concept through a short-term reciprocal transplant experiment combined with passive warming via open-top chambers (OTCs). We asked: 1) if population growth rates of different ecotypes responded differently to reciprocal transplant, 2) if home-site advantage existed, and 3) whether there is an interaction of ecotype, transplant garden, and OTC treatment. Three transplant gardens were established, two north of the Brooks Range (Toolik and Sagwon) and one south (Coldfoot), and OTCs were deployed in the northern gardens. Tillers were censused in 2016 and 2017. Yellow Taxi Analysis of Lefkovitch matrices quantified each tiller’s individual contribution to overall population growth rate, which was used as the dependent variable in nested ANOVAs. Of tussocks grown in ambient temperature, mean tiller population growth from different source ecotypes did not respond differently to transplant. Home site advantage was not observed. Mean growth rate of warmed tillers was higher at Toolik than Sagwon. Unlike prior long-term studies, this study failed to detect differences in tiller population growth as a function of garden or ecotype, emphasizing the need for long-term monitoring to capture the impact of warming on Arctic species.

Three reciprocal transplant gardens were established along the Dalton Highway in Alaska, U.S.A. The Coldfoot site (CF; 67.26° N, 150.17° W) was situated south of the Brooks Range, while two additional sites, Toolik Lake (TL; 68.63° N, 149.36° W) and Sagwon (SG; 69.42° N, 148.72° W), were north of the Brooks Range. Haphazardly selected tussocks of E. vaginatum were cut from the soil at each site in August 2014 by using a serrated knife to slice the roots below the moss line. Severed tussocks of each ecotype were reciprocally transplanted to each of the three gardens by placing the harvested tussocks in the same open location in the tundra where the local tussocks had been removed. Controls were established by re-planting severed tussocks into the garden from which they originated. Sixty tussocks per ecotype were planted into each of the three gardens. Tussocks from the same ecotype were clustered in groups of three separated by approximately 0.5 m. At the two most northern sites (SG and TL), the twenty resulting clusters were randomly assigned to either an ambient treatment (n = 10) or to passive warming (n = 10) using open-top chambers (OTC). Open-top chambers were deployed from July 11 – August 28, 2015, June 2 – August 28, 2016, and May 30 – August 26, 2017 in order to raise growing season temperatures experienced by the tussocks. Open-top chambers were not deployed at the most southern CF site due to the site’s close proximity to the Dalton Highway.

Two clusters of tussocks (6 total tussocks) from each ecotype and garden combination and from each warming treatment were randomly selected, and were visited while in peak leaf biomass in July 2016. Eighteen haphazardly selected tillers from each tussock were tagged using numbered plant tags affixed to each tiller using a zip tie. In instances where fewer than 18 tillers were present in a tussock, all living tillers were tagged. During the 2016 census, the total number of green, living leaves per tagged tiller was tallied, and the length of the two longest leaves in each tiller was measured. All the tagged tillers were re-censused in July 2017 using the same technique as in 2016. After being censused in the field, each tussock was carefully removed from its location and dissected to count the number of daughter shoots (easily distinguished from older leaves by the lack of any browned, over-wintered tips) produced vegetatively by each adult tiller. Tillers that died over the transition period were identified by the lack of any living leaves. Sixteen tillers were removed from the dataset due to the loss of tags from tillers during transportation and potential double-tagging. 

Total leaf areas of each tiller for both 2016 and 2017 were estimated using the allometric relationship between leaf area and the tiller size index, which is the product of the number of leaves per tiller and the sum of the length of the two longest leaves. Four size-based stage classes were identified using the 2017 tillers, in addition to a separate stage class (stage class 1; S₁) comprised of 2016 daughter tillers. A stage-based Lefkovitch model was constructed based on the possible transitions among these stage classes.

The transition of daughter tillers (in 2016) to higher stage classes (in 2017) was inferred using previously collected data. This inference was necessary, because while S₁ tillers were tallied in 2017, their 2017 size was unintentionally excluded, and so it was impossible to determine the size transition of those tillers between 2016 and 2017. Using this prior dataset, daughter tillers (from 2010) from the same garden and ecotype combinations used in the current study were identified (N = 84). The 2011 leaf area of these identified daughter tillers was noted and was used to determine the proportion of those daughter tillers that fell into each of the size-based stage classes (2-5) identified in the 2017 study. None of the daughter tillers died over the transition period, 69% of daughter tillers transitioned into the equivalent of the 2017 size class two, 16.7% into class three, 10.7% into class four, and 3.6% into class five. These calculated transition probabilities along with the actual number of daughter tillers identified in the 2017 study were used to estimate the number of daughter tillers (in 2016) that transitioned to each stage within each garden, ecotype, and warming treatment combination in 2017. Each daughter tiller (2016) was assigned a random number, and a cumulative distribution was used to assign 2017 classes to these individuals using the derived proportions.

Two separate 5x5 matrices were constructed using MATLAB version R2019B. The first excluded all tillers exposed to OTC treatments and focused only on the transplanted tillers in all combinations of garden and ecotype. The second excluded the CF garden, as no OTC treatments were deployed at CF. However, all other tillers in the design were analyzed, which included all tillers from the combination of garden (TL and SG), source ecotypes (CF, TL, and SG), and treatment (ambient vs. OTC). Population growth rates (λ) were calculated as the dominant eigenvalue of each matrix. Yellow Taxi Analysis (YTA), a derivative of Tukey’s Jackknife, was performed to provide the replication necessary to run statistical analyses. Yellow Taxi Analysis provides pseudovalues of lambda (f_i) for each observation, which measure each tiller’s contribution to overall population growth rate.

A mixed model two-way ANOVA design was developed and analyzed using restricted maximum likelihood (REML) in order to determine if tussocks from different source ecotypes responded differentially to transplanting along the latitudinal gradient. The pseudovalue of lambda (f_i) was the dependent variable, garden and source ecotype were the main effects, cluster (of three tussocks) was a nested random effect within the garden x ecotype interaction term, and tussock was a nested random effect within the cluster(garden x ecotype) term. Only tussocks subjected to ambient temperature treatments (non-OTC) were used in this analysis. 

A separate mixed model two-way ANOVA was constructed and analyzed by REML using data from all ambient temperature (non-OTC) combinations of garden (CF, TL, and SG) and source ecotype (CF, TL, and SG) to determine if tussocks that were planted back into their original home garden differed in performance from those tussocks that were transplanted to different “away” gardens. The pseudovalue of lambda (f_i) was the dependent variable, garden was the first independent variable, and source ecotype re-coded as “home” or “away” was the second independent variable. Cluster (of three tussocks) was a nested random effect within the garden x home/away interaction term, and tussock was a nested random effect within the cluster(garden x home/away) term. 

Using data only from tillers growing in the two most northern sites (TL and SG), a three-way ANOVA model was constructed and analyzed using REML to determine if the two-way interaction of garden (TL and SG) and source ecotype (TL, SG, and CF) depended on warming treatment (Ambient vs. OTC). Pseudovalue of lambda (f_i) was the dependent variable, while garden, source ecotype, and warming treatment were main effects. Cluster (of three tussocks) was a nested random effect within the garden x ecotype x treatment three-way interaction term, and tussock was a nested random effect within the cluster(garden x ecotype x treatment) term.