- Top of page
- Materials and Methods
- Conflict of Interest
- Supporting Information
Observed and projected climate change, especially pronounced in Arctic regions, suggest future increase of air temperature and precipitation rates, thereby influencing snow depth, density and duration of snow cover (Serreze et al. 2000; ACIA 2005; Serreze and Francis 2006; IPCC 2007; Førland et al. 2011). Together with increasing air temperatures, these changes are already provoking responses from some ecosystems, for instance changes in carbon and nutrient cycling, and “shrubification” in Arctic and alpine ecosystems (Sturm et al. 2005; Parmesan 2006). Climate change also increases the frequency and magnitudes of extreme climatic events (Hansen et al. 2012), which can lead to winter warming events and associated reductions in snow cover during winter (Shabbar and Bonsal 2003; IPCC 2007). These warming events can be asso-ciated with heavy rainfall, as was the case in this study, which can be very effective in removing and compacting snow. Specifically, projected earlier snow melt and increased frequency and magnitude of extreme climatic events, in the form of warming periods and rain events during winters in the high Arctic, could have long-term effects on plant community composition. Frost damage can occur by exposing plants to unexpectedly low air temperatures through removing snow in mid-winter or by exposure to spring frosts due to very early snow melt (Inouye 2000; Bokhorst et al. 2008, 2011; Preece et al. 2012). However, increased solid precipitation during years without winter warming events might increase snow depth and thereby delay the onset of the growing season, thus protecting plants, while naturally occurring inter-annual differences in spring temperatures might delay or advance snow melt.
Many arctic-alpine plant species produce flower buds in the year prior to flowering; these then overwinter in a variety of developmental stages (Bliss 1971). For that reason growing conditions in the year of bud production, such as growing season length and air temperature, are partly responsible for flower abundance and therefore for a given species' reproductive success (Inouye et al. 2002; Inouye and Saavedra 2003; Körner 2003; Høye et al. 2007). However, some species initiate flower primordia during the same year that they flower, and therefore their flower abundance might depend mostly on the current year's growing conditions. Snow cover has been recognized as one of the main drivers for plant growing conditions in the Arctic, and its inter-annual variability is well documented (Hinkler et al. 2008). To test the role of snow cover on flowering as a proxy for reproductive success, studies with multi-year monitoring of response variables during differing natural snow conditions are needed. Although several arctic/alpine snow manipulation experiments exist, only a few of these exceed 3 years duration, and even fewer consider inter-annual flower abundance fluctuations (Wipf and Rixen 2010). The aim of this study is to fill that gap.
Snow depth controls the duration of snow lie and thereby length of the growing season (Walker et al. 1999; Borner et al. 2008; Wipf 2009; Wipf and Rixen 2010; Cooper et al. 2011). Early snowmelt and resulting longer growing seasons may be favorable for flower bud production due to potentially higher energy and photosynthate accumulation throughout the summer. However, snow cover also directly controls soil and canopy temperatures during winter and spring, thereby protecting arctic and alpine plants from damaging sub-zero temperatures. Premature snow melt in spring, as well as shallow snow cover or snow melt during winter caused by extreme warming events, expose above ground tissues to detrimental winter and spring frosts. This negatively affects flower buds and substantially reduces flower abundance in subsequent growing seasons through freezing, desiccation, or deacclimation without sufficient reacclimation (Gates 1912; Firmage and Cole 1988; Larcher 2004; Høye et al. 2007; Bokhorst et al. 2008; Inouye 2008). Many processes control snow depth and melt out timing and lead to large spatial and temporal variations in arctic snow cover (Hinkler et al. 2008). These processes would therefore affect flower abundances, and species-specific responses would be expected due to specific physiological parameters and growth requirements. For instance, a species with greater frost hardiness would lose fewer flower buds in the case of exposure to extremely low temperatures than a species with low frost hardiness.
This study was originally intended to experimentally assess the role of timing of spring snow melt (and thus the length of growing season) on flower abundance for a set of common high-arctic plant species. Our initial hypothesis was that an experimentally delayed spring snow melt will reduce flower abundance. We also expected that the responses would be species specific. However, during 5 years of monitoring flower abundance in the study site, we experienced two extreme warming events during mid-winter which exceeded normal warm periods in the study area, and we opportunistically report these here with the post-hoc hypothesis that deeper snow would prevent plants from being exposed to winter air. Thus, we present a combination of both a manipulation and observation study which was not originally intended to include winter warming events. For some species we observed that deeper snow cover buffered plants from extreme winter warming events and saved the subsequent flower crop in one case, but not in the other case. Here, we present species specific responses of flower abundance to (1) snow melt timing, and (2) extreme winter warming events under contrasting snow depths.
- Top of page
- Materials and Methods
- Conflict of Interest
- Supporting Information
Our hypothesis that delayed spring snow melt would reduce flower abundances held for only two of the six observed species. Stellaria crassipes had fewer flowers in Deep than in all other snow regimes, although that signal is visible only from the 4th year of the experiment. We assume, however, that this species does not produce overwintering flower buds because it (1) produces flowers at the end of new shoots instead of overwintering axillas (own observations), (2) flowers very late in the season (Cooper et al. 2011), and (3) is not affected by winter warming events, which might be a trait of some, though not all, chamaephytes with overwintering flower buds, as we will argue for later on. The lower flower abundance of S. crassipes in Deep can therefore be attributed to insufficient resource allocation to flower production due to a shortened period between onset of growth and flowering, thereby indirectly supporting our hypothesis. That flower abundance was reduced only 4 years after snow manipulation started could be due to possible delayed costs of reproduction or direct fecundity costs, that is growing seasons in Deep were not long enough to replenish energy reserves used for previous years' reproduction in the long run (Obeso 2002). The fact that this response was not provoked in Medium points out that Deep crossed a certain threshold for snow melt date.
Of the remaining observed species, only Cassiope tetragona showed the hypothesized response to increased snow depth; the later the individuals melted out, the fewer flowers they had in the following season. In 2007, the first summer following snow manipulation, indications for no difference between Deep and Normal were found (data could not be included in the analysis here due to incomplete observations). The effect of Deep increased each year until 2009, that is the snow manipulation effect on C. tetragona became more pronounced over the initial years of the study, pointing out an accumulative effect of the previous years' growing conditions; overwintering flower buds of one season might contribute to the pool of flower buds of more than only one following seasons, as shown for Bistorta vivipara by Diggle (1997). Growing seasons shortened by later snow melt contributed fewer C. tetragona flower buds, which might be explained by shorter annual growth increments caused by a shortened (and therefore, in terms of growing degree days colder) growing season, as found by Mallik et al. (2011) and Rumpf et al. (in prep) in the same study site, and by Weijers et al. (2013) in Ny-Ålesund and Endalen, Svalbard and sub-arctic Sweden. In other studies (Rozema et al. 2009; Weijers et al. 2012, 2013), the number of flower buds formed per year seems to be related to annual shoot length growth, and thus to accumulative summer temperatures (Stef Weijers, pers. Comm.); longer and warmer seasons yielded longer shoots with more leaf axillae, the location where actual flower bud formation occurs. This is confirmed by Mallik et al. (2011), who found fewer leaves in Deep than in Normal after the 2007 growing season, the first year of the study with a shortened growing season, and no difference before treatment allocation. However, during our study, C. tetragona's flower abundance response to the snow regimes was overlain by its response to the winter warming events in 2010 and 2012, which will be discussed next.
Winter warming events are common on Svalbard, but are usually not as severe as those observed in early 2010 and in 2012. Accumulated temperature sums and precipitation during January to March throughout 37 years (1976–2012) recorded at Longyearbyen airport show that 2010 and 2012 were among the warmest (fourth warmest and warmest, respectively) and by far the wettest. Temperature sums were 1.4 and 3.1 SD, and precipitation during warming 2.2 and 2.7 SD above the 1976–2012 mean for 2010 and 2012, respectively (data from Norwegian Meteorological Institute, not shown). Following the reasoning of Smith (2011), both warming events reported here could be considered as “climate extremes”, while following a more climatological definition the observed warming periods in 2010 and 2012 might be called “warm” and “extremely warm”, respectively, and both events were “very moist”, not “extremely moist” (nomenclature used in Hansen et al. 2012). However, not enough data was available to compare our observations with an earlier standard reference period (Norwegian Meteorological Institute), and the fact that our data are based on only one measurement station makes comparison difficult.
Snow-poor or mild winters have been shown to freeze, desiccate, or deharden overwintering meristems and flower buds of berry yielding, ericaceous dwarf shrubs (Raatikainen and Vänninen 1988; Ogren 1996; Taulavuori et al. 1997; Bokhorst et al. 2008) and other species (Gates 1912; Firmage and Cole 1988; Høye et al. 2007; Inouye 2008; Mallik et al. 2011) in sub-arctic and temperate regions and alpine habitats, thereby significantly reducing shoot growth, berry and capsule yield, and flower abundances. Similar effects have been observed on flower abundances in this study for two of the four observed chamaephytes, that is species which keep their overwintering meristems above ground; snow melting by warm temperatures together with rainfall might expose overwintering tissues, which are normally protected by the snowpack, to subsequent cold winter air temperatures and winds which may destroy exposed tissue. Of all our studied species, C. tetragona showed the strongest response to winter warming events by significantly reduced flower abundances. In 2010, C. tetragona flower abundances in all snow regimes except Deep, and in all regimes in 2012 were clearly affected. Dryas octopetala responded to these warming events only in the un-manipulated snow regime Shallow and Normal in 2012, although its response was not as strong as that of C. tetragona. The lower the initial snow depth, the higher the proportion of removed snow by warm air temperatures and heavy rain, that is a deep snowpack will last longer than a shallow snowpack. Thus, the severity of flower abundance reduction might have increased with decreasing snow cover in both cases because plants under a deeper snow pack might have still been protected from exposure to detrimental winter temperatures after the warming event by a remaining, sufficiently deep snowpack.
The influence of the observed warming events was stronger on C. tetragona than on D. octopetala, and the reason for this might be twofold; (1) the shoots of Cassiope tetragona are more erect and taller than the procumbent D. octopetala. In addition, C. tetragona produces its flower buds on the shoot tips. Therefore, C. tetragona flower buds might be exposed to colder air temperatures over a longer time period than D. octopetala, which keeps its flower buds close to the ground and might be still protected by a remaining layer of snow and ice after mid-winter snow melt by warm events (personal observation). Additionally, the rosette like structure of D. octopetala shoot tips might serve as protection for flower buds (Inouye 2000). Raatikainen and Vänninen (1988) came to similar conclusions on the difference of proportions of surviving flower buds after a particularly snow-poor and cold winter in Finland: Vaccinium myrtillus has a high canopy and therefore lower proportion of flower bud survival and V. vitis-idea has low canopy and therefore higher proportion of flower bud survival. For the same reason one of the remaining two chamaephytes of this study, that is Saxifraga oppositifolia might not have been affected by the warming events; it is of very low stature. Secondly, (b) Dryas octopetala is adapted to grow in areas with shallow snow, as opposed to C. tetragona which requires a consistent snow cover during winter (Rønning 1996). Therefore the smaller effect of warm periods on D. octopetala might be not only of morphological, but also of a physiological nature, that is D. octopetala might develop stronger frost hardening and withstand cold temperatures better than C. tetragona, as found for snow bed species in alpine New-Zealand by Bannister et al. (2005).
In 2012, C. tetragona individuals in Deep were affected by the warming event, unlike in 2010. The 2012 warming event was more severe than the one in 2010, with higher temperatures and greater precipitation, and two possible scenarios might have been responsible for the flower abundance crash in Deep during that year. (1) Warm temperatures and rain might have been sufficient to remove enough snow in Deep to expose plants to following cold winter air, thereby freezing flower buds to death. This might be possible given the fact that 2012 was a particularly snow-poor year (Fig. 2). However, the winter of 2012 was also relatively warm, and soil temperatures after the warming event never reached abnormally low temperatures, as was the case for Normal during long periods in 2010. Therefore, an alternative explanation is possible where (2) the warm temperatures themselves were long and warm enough to deharden overwintering flower buds, thus rendering them susceptible to the subsequent intermediately cold temperatures. Similar mechanisms might have been responsible for the lower flower abundance in all treatments for the hemicryptophyte Bistorta vivipara in 2012, although we unfortunately cannot compare with the 2010 event since data is not available for that year.
Unfortunately, we cannot disentangle whether the effect of the warming periods was due to temperature sums, accumulated precipitation, or if both had to be high to cross the threshold of inducing a loss of flower buds. In any case, in order to be considered an “extreme climatic event”, the observed response should be extreme enough to impact the ecosystem severely enough to result in temporary or even permanent community structure changes or similar (Smith 2011). This was not the case in our study, where only one species' threshold was clearly exceeded by the warming periods, and its recovery was fast enough to cover the events' effect only one season after. The flower abundance of C. tetragona can therefore be described as very resilient, while the other species' flower abundances are resistant to the climate extremes observed here. However, although this study focuses on flowers, it may be reasonable to assume that other above-ground organs may respond in a similar way to shorter growing seasons or exposure to freezing air temperatures through mid-winter mild events (Inouye 2000). For instance, survival of overwintering vegetative stages of a monocarpic species was drastically reduced by exposure to cold winter temperatures if thermal insulation was not sufficient enough (Simons et al. 2010). Thus, this study may also give a justification for the synchrony of high Arctic herbivore dynamics in relation to wide scale icing events recently reported by Hansen et al. (2013).
The hemicryptophyte and semi-parasite Pedicularis hirsuta is most likely not affected by the warming events due to below ground overwintering and subsequent protection from cold air temperature. However, it had a flower peak in Deep in 2010 and in Normal in 2011, the year after particularly early snow melt caused by a winter warm event in the same snow regime. This elongation of the growing season might have facilitated production of either overwintering rhizomes or viable seeds, leading to larger or more individuals the following year yielding more flowers. Similar, although not as pronounced or statistically significant flower peaks were observed for C. tetragona, D. octopetala, and S. crassipes. These peaks were followed by significant crashes of flower abundances, which might indicate direct fecundity costs caused by excessive flowering events the year before (Obeso 2002), while a combination of this and winter warming events might have been the case for C. tetragona and D. octopetala.
Although not examined in this study, the observed effects of season length and winter warming events could have specific effects on the only known annual species on Svalbard overwintering as seeds (Koenigia groenlandica). Winter warming might only have an effect if it breaks seed dormancy and thereby reduces the seed bank. Short, late starting seasons could potentially restrict seed set by delaying seed ripening processes too late into autumn, while seasons starting too early could expose seedlings to late spring frosts and thereby not only kill reproductive organs but the whole plant. Both scenarios are also valid for the perennials examined in this study (Inouye 2000), however would have stronger implications on annuals, since for those whole individuals and not only vegetative parts of individuals are at risk.
This study fails to estimate what would happen in the case of earlier snowmelt caused by warmer air temperatures during spring, as suggested for the future by climate change models. Repeated trials of snow removal in this study failed because of insufficient marking of sub-plots (marking poles removed by reindeer), or because of wind refilling the removed snow, thereby reducing the number of replicates to a useless level. In any case, artificial snow removal would fall into a period of the year with very low air temperatures and expose protected plants to the cold, thereby confounding the experimental treatment of snow removal with exposure to early season frost. A combination of snow removal and warming is suggested to mimic a natural, earlier snow melt (see also Wipf and Rixen 2010).
Given the evidence presented in our study, we conclude: Season length as dictated by snow melt timing has various plant species-specific effects, independent of life-form. Species with overwintering above-ground flower buds (chamaephytes) are affected by winter warm events in various degrees, depending on the positioning of buds, and on the snow depth during winter. An increase of frequency and amplitude of extreme winter warm events will decrease flower abundance and thus reproductive success of some species (here: Cassiope tetragona) and thereby favor the fitness of others. This underlines the importance of winter conditions and their influence on summer processes. The impact of potential snow cover changes on high-Arctic plant community composition dynamics caused by altered reproductive success is complex and cannot be answered with the current knowledge of the system; more multi-year, multi-season, and multi-species studies incorporating a set of predictor variables are required to fill this gap.