Flowering and fruiting responses to climate change of two Arctic plant species , purple saxifrage ( Saxifraga oppositifolia ) and mountain avens ( Dryas integrifolia )

In temperate regions, there are clear indications that spring flowering plants are flowering earlier due to rising temperatures of contemporary climate change. Temperatures in temperate regions are rising predominantly in spring. However, Arctic regions are seeing unprecedented temperature increases, predominantly towards the end of the growing season. We might, therefore, expect to see earlier flowering of later-season flowering Arctic plants. Parks Canada has been monitoring purple saxifrage (Saxifraga oppositifolia) and mountain avens (Dryas integrifolia) flowering and fruiting times for 20 years at Tanquary Fiord, Quttinirpaaq National Park, Ellesmere Island. Saxifraga oppositifolia flowers in early spring, while D. integrifolia flowers in midsummer. Over the 20-year period, Tanquary Fiord’s annual and late-summer temperatures have risen significantly. During the same timeframe, D. integrifolia showed a trend towards earlier flowering and fruiting, but S. oppositifolia showed no changes in flowering or fruiting time. Flowering time was related to monthly temperatures just prior to flowering. The number of flowers produced was related to the previous autumn’s monthly temperatures. We found no relationship between flowering time and snowmelt date. Our findings suggest that Arctic community level ecological effects from climate change induced phenology changes will differ from those in temperate regions.


Introduction
Flowering times of many temperate spring flowering plants are advancing due to rising temperatures associated with recent climate change (Fitter et al. 1995;Abu-Asab et al. 2001;Primack et al. 2004;Beaubien and Hamann 2011;Panchen et al. 2012).Fruiting responses to climate change in temperate regions are less well studied and the changes less clear (Gordo and Sanz 2010;Gallinat et al. 2015).In temperate regions of Europe, some species are advancing their fruiting time due to rising temperatures of climate change (Menzel et al. 2006;Gordo and Sanz 2010;Van Vliet et al. 2014).However, in a North American study, the fruiting time of earlier-flowering species advanced with warmer temperatures, while the fruiting time of later-flowering species was delayed with warmer temperatures (Sherry et al. 2007).Temperatures in temperate regions have been rising since the 1960s, with increases being most pronounced during spring (Menne et al. 2010).Temperatures in the eastern Canadian Arctic, however, have been rising very rapidly since the 1990s but predominantly at the end of the summer and during winter (Przybylak 2003;McBean 2004;Furgal and Prowse 2007;AMAP 2011) Time of flowering is most often controlled by temperature or day length (photoperiod) or a combination of the two (Bernier and Périlleux 2005).Photoperiod control of flowering time is seen in some Arctic species but temperature is the main driver for Arctic plant flowering time (Porsild and Cody 1980;Thórhallsdóttir 1998;Keller and Körner 2003;Hülber et al. 2010).Vernalisation or a chilling period followed by a cumulative period of warmer temperatures (growing degree-days) controls time of flowering (Bernier and Périlleux 2005).
Unlike temperate regions, where there are many sources of long-term data on flowering and fruiting times of plants (Primack 2003;Primack and Miller-Rushing 2009;Culley 2013), the remoteness of the Arctic introduces challenges in obtaining long-term phenology data, and consequently, there are few data sets on the flowering and fruiting times of Arctic plants over an extended period of time (Høye et al. 2007b).Parks Canada's Nunavut Field Unit established an International Tundra Experiment (ITEX) (Molau and Mølgaard 1996;Henry and Molau 1997) phenology monitoring program in 1994 at Quttinirpaaq National Park to monitor the flowering and fruiting times of purple saxifrage (Saxifraga oppositifolia L. (Saxifragaceae), aupilattunnguat (Inuktitut)) and mountain avens (Dryas integrifolia Vahl (Rosaceae), malikkaat (Inuktitut)) at the Tanquary Fiord warden station.Saxifraga oppositifolia and D. integrifolia are two common circumpolar Arctic plant species with a wide distribution across the Canadian Arctic Archipelago (Aiken et al. 2003;Elven 2003).These two species are often dominant in the Quttinirpaaq National Park landscape (Soper and Powell 1985;Edlund and Alt 1989).Saxifraga oppositifolia is one of the first Arctic plants to flower in spring, while D. integrifolia flowers in midsummer.The 20-year Tanquary Fiord ITEX phenology data set provides a unique opportunity to study how two species of Arctic plants that flower at different times during the growing season are responding to climate change.
With Arctic temperature changes occurring predominantly at the end of the growing season and winter, our hypothesis is that later-season flowering plants are flowering earlier and Arctic plant fruits in general are maturing earlier, which we test herein with the S. oppositifolia and D. integrifolia Tanquary Fiord, Quttinripaaq National Park data set.We determine which monthly temperatures have the greatest influence on the time of flowering and fruiting of S. oppositifolia and D. integrifolia, whether snowmelt influences time of flowering of these two species, and whether temperatures in the previous year influence the quantity of flowers produced by these two species in the following year.We also address which monthly temperatures are rising at Tanquary Fiord.

Site
Quttinirpaaq National Park is situated on northern Ellesmere Island in the Canadian Arctic Archipelago.Tanquary Fiord is the name of the main warden station in the park and is located at the head of Tanquary Fiord at 81.4°N, 76.7°W.The region is a polar desert with mean annual precipitation of less than 50 mm (Edlund and Alt 1989;Przybylak 2003).The ITEX phenology plots are located on the edge of a gravel wash on the MacDonald River delta.The vegetation is discontinuous in the vicinity of the ITEX plots (Fig. 1A).
In 1994, Parks Canada's Nunavut Field Unit established two ITEX phenology monitoring 50 m × 50 m plots within a 15-min walk from the Tanquary Fiord warden station.The park staff randomly tagged 25 S. oppositifolia plants in the first plot and 25 D. integrifolia plants in the adjacent plot.The tagged plants ranged in diameter from 15 to 60 cm and were spaced at least 0.5 m apart (Fig. 1) (Raillard 1999).Every 3 days, park staff also counted the number of mature fruits on each S. oppositifolia plant and the number of twisted fruits on each D. integrifolia plant.A S. oppositifolia fruit was considered mature when its two carpels turned red and were swollen (Fig. 2A).A D. integrifolia fruit was considered to have twisted fruit when its styles had extended and twisted around each other (Fig. 2B).No mature fruit were recorded on the S. oppositifolia in 2009 and 2011.There were also no D. integrifolia twisted fruit counted in 1996.
For all 25 S. oppositifolia plants and separately for all 25 D. integrifolia plants, we summed the number of flowers and summed the number of fruits per visit per year.We determined peak flower and peak fruit day of year for each year for both species.We defined peak flower or fruit day of year as the number of days from 1 st January to the day when the number of flowers or fruit first reached 90% of the maximum number of flowers or mature/ twisted fruit counted that year.This ameliorated the problem in the protocol of requiring flowers to be counted until they dropped their petals.In the case of S. oppositifolia and, to a lesser extent D. integrifolia, petals persisted on plants well beyond anthesis (Fig. 2A), causing flowers to continue to be counted beyond the true open flower stage.

Temperature data
Parks Canada established a weather station at Tanquary Fiord in 1989.Temperature data from the weather station provided daily minimum, maximum, and mean temperatures from 1989 to 2002 and hourly temperatures from 2003 to 2014.We calculated the monthly mean minimum, monthly mean maximum, and monthly mean temperature by averaging the daily minimum, maximum, and mean temperatures for each month from 1989 to 2002.For 2003-2014, we first determined the daily minimum, maximum, and mean temperature from each day's hourly temperatures followed by averaging these daily minimum, maximum, and mean temperatures for each month.There were some months in which the Tanquary Fiord daily or hourly temperature data were incomplete.For months with incomplete temperature data, we reconstructed the Tanquary Fiord monthly temperatures from the Eureka, Ellesmere Island, Nunavut, Environment Canada Weather Station temperature data (Environment Canada 2014) using the method described by Leathers et al. (2008) and Panchen et al. (2012) (Table S1 1 ).The Eureka Weather Station (79.59°N, 85.56°W) is the closest weather station to Tanquary Fiord and experiences a similar climate (Edlund and Alt 1989).We also calculated the annual mean minimum, annual mean maximum, and annual mean temperature from the monthly mean minimum, monthly mean maximum, and monthly mean temperatures.

Statistical analysis
To account for interannual variation, we calculated the 5-year rolling average of peak flower and peak fruit day of year for each species each year from the species' yearly peak flower and peak fruit day of year.To determine the rate of change in flowering or fruiting time over the 20 years, we ran a regression analysis with the species' 5-year rolling average of peak flower or fruit day of year as the response variable and year as the independent variable.We also determined the year-to-year trend towards earlier flowering or fruiting by running a regression with the species' peak flower or fruit day of year as the response variable and year as the independent variable.We used a single standard least squares model per species to determine how much of the species' peak flower day of year was explained by year and April, May, June, and July mean temperatures and a similar model per species to explain the fruit day of year, i.e., peak flower or peak fruit day of year was the response variable and year and April, May, June, and July mean temperatures were the independent variables.To determine the rate of change in flowering or fruiting time with rising temperature for each species, a regression analysis was run with peak flower or fruit day of year as the response variable and June mean temperature as the independent variable for D. integrifolia and May mean temperature as the independent variable for S. oppositifolia.June and May temperatures were used because in the standard least squares models, June was the monthly temperature that had the greatest influence on the timing of flowering and fruiting for D. integrifolia, while May mean temperature had the greatest influence on the timing of flowering for S. oppositifolia (Table 1).To determine if time of flowering of each species is related to the snowmelt date, we ran a regression with peak flower day of year as the response variable and first snow-free date as the independent variable.We used a standard least squares model to determine how much of each species' peak flower count of one year was explained by June, July, August, and September mean minimum temperatures of the previous year, i.e., a single multiple regression for each species with peak flower count as the response variable and June, July, August, and September mean minimum temperatures as the independent variables.That is, does the previous year's summer or autumn temperatures affect the number of flowers that open in the following year?To determine the rate of rising temperatures over the past 25 years at Tanquary Fiord, we ran regressions of Tanquary Fiord annual and monthly mean minimum, mean maximum, and mean temperatures versus year.We also assessed the 1989-2014 average monthly temperature variation 1 Supplementary material is available with the article through the journal Web site at http://nrcresearchpress.com/doi/suppl/10.1139/as-2015-0016.through the year.We used the JMP11 software package (SAS Institute, Cary, North Carolina, USA) for all statistical analyses.
Much of the variation in D. integrifolia peak flower day of year was explained by year and April, May, June, and July mean temperatures (Table 1).Dryas integrifolia flowering time advances by 2.3 days/°C rise in June mean temperature (R 2 = 0.54, P = 0.0001, N = 21).Similarly, much of the variation in D. integrifolia peak fruit day of year was explained by year and April, May, June, and July mean temperatures, albeit with only June mean temperature as a significant term in the model (Table 1).Dryas integrifolia fruiting time advances by 3.4 days/°C rise in June temperature (R 2 = 0.47, P = 0.0008, N = 20).Some of the variation in S. oppositifolia peak flower day of year was explained by year and April, May, and June mean temperatures, albeit with only May mean temperature as the significant term in the model (Table 1).Saxifraga oppositifolia flowering time advances by 1.8 days/°C rise in May mean temperature (R 2 = 0.30, P = 0.0127, N = 20).However, no variation in S. oppositifolia peak fruit day of year was explained by year and monthly mean minimum temperatures (Table 1).There was no relationship between peak flower day of year and snowmelt date for either species (S. oppositifolia: R 2 = 0.12, P = 0.1638, N = 18 and D. integrifolia: R 2 = 0.01, P = 0.6813, N = 18).
Much of the variation in the peak number of D. integrifolia flowers is explained by the previous year's June, July, August, and September mean minimum temperatures (R 2 = 0.47, P = 0.0316, N = 21), with August mean minimum temperature as the only significant term in the model (P = 0.0111).There was a positive relationship between peak number of S. oppositifolia flowers and the previous year's October mean minimum temperatures (R 2 = 0.27, P = 0.0188, N = 20), although the model with the previous year's June, July, August, and September mean minimum temperatures as the independent variables was not significant (R 2 = 0.13, P = 0.7072, N = 20).

Temperature
The annual mean minimum, mean maximum, and mean temperatures at Tanquary Fiord have risen significantly over the past 25 years (Fig. 4A; Table 2).There is a significant trend towards warmer temperatures for January, March, July, August, September, and October (either at α = 0.05 or α = 0.10) (Fig. 4B; Table 2).The annual and monthly mean minimum temperatures are rising more than the annual and monthly mean maximum and mean temperatures (Fig. 4; Table 2).The Tanquary Fiord average 1989-2014 monthly temperature varied little across the months of January, February, and March, hovering at approximately −35 °C (Fig. 5; Table S2).Temperatures rose dramatically in April and May to above zero by the start of June.Temperatures in June, July, and August varied between months by approximately 5 °C.Temperatures dropped dramatically in September and October and less steeply in November and December.

Discussion and conclusions
Flowering and fruiting times of the midsummer flowering D. integrifolia have advanced significantly over the 20 years from 1994 to 2014, but flowering and fruiting times of the early spring flowering S. oppositifolia have not changed significantly.This pattern is in striking contrast to temperate regions, in which it is predominantly spring-flowering plants that are flowering earlier, while summer-flowering plants and fruiting times are showing less of a trend towards earlier flowering and fruiting (Fitter et al. 1995;Abu-Asab et al. 2001;Primack et al. 2004;Menzel et al. 2006;Panchen et al. 2012).The earlier flowering of summer-, but not spring-, flowering species might be expected in the Arctic given that the  2).
temperatures towards the end of the growing season are warming the most and flowering times are highly dependent on temperatures in the 1 or 2 months prior to flowering (Fitter et al. 1995;Thórhallsdóttir 1998;Panchen et al. 2012).Given that fall and winter temperatures are warming in the Arctic, the early spring flowering species S. oppositifolia may not be able to flower earlier because of vernalisation limitations (Cook et al. 2012b).
The advance in flowering time of D. integrifolia at Tanquary Fiord of 3.6 day/decade and 2.3 days/°C is comparable to spring flowering plants that are advancing the most in temperate northeastern North America (Bradley et al. 1999;Abu-Asab et al. 2001;Ledneva et al. 2004;Primack et al. 2004;Wolfe et al. 2005;Lavoie and Lachance 2006;Miller-Rushing and Primack 2008;Panchen et al. 2012), perhaps reflecting the rapid climate change that Table 2. Tanquary Fiord change in annual and monthly mean maximum, mean, and mean minimum temperatures from 1989 to 2014 (β = slope in °C/year, dark grey = significant at α = 0.05, and pale grey = significant at α = 0.10).
the Arctic is experiencing (McBean 2004;Høye et al. 2007b;AMAP 2011) and the greater sensitivity of colder climate plants (Cook et al. 2012a) but also indicating the substantial phenological plasticity of D. integrifolia.The above-noted differences in the pattern of phenological responses to climate change between the Arctic and temperate regions may result in different community-level ecological effects in Arctic versus temperate regions.
Our results also indicate that temperatures in the months just prior to flowering are most important in the timing of flowering.Saxifraga oppositifolia flowering time is primarily induced by temperatures 1 month prior to flowering, whereas D. integrifolia flowering is induced by the cumulative temperatures of the prior 3 months.Dryas integrifolia and S. oppositifolia also showed a difference in their flowering time temperature sensitivity (1.8 days/°C for S. oppositifolia versus 2.3 days/°C for D. integrifolia).Variation in phenological responses across taxa to a warming climate is a common finding across temperate, alpine, and Arctic regions (Fitter et al. 1995;Thórhallsdóttir 1998;Abu-Asab et al. 2001;Panchen et al. 2012;CaraDonna et al. 2014).The variation in flowering time temperature sensitivity plus the different monthly temperatures to which these two species showed flowering time temperature sensitivity may have implications for ecological community assembly, such as mismatches in timing between flowering time and pollinators (Hegland et al. 2009;McKinney et al. 2012;Høye et al. 2013).These differences in phenological responses to climate change of two species that dominate the landscape could also result in change of the dominant species in the ecological community assembly.
Our findings are consistent with a metadata analysis of the pan-Arctic ITEX data set using warmed and control plots and a long-term study in Swedish Lapland, which showed greater sensitivity to warming of later flowering species (Molau et al. 2005;Prevéy et al. 2014).Our findings are also consistent with a long-term study of flowering time in Iceland showing warming just prior to flowering as the main reason for earlier flowering (Thórhallsdóttir 1998).Studies with warmed and control plots and long-term phenology studies indicate a trend towards earlier flowering and fruiting for D. integrifolia and Dryas octopetala L., a congener and close relative of D. integrifolia, but no trend for S. oppositifolia (Wookey et al. 1993(Wookey et al. , 1995;;Stenström et al. 1997;Welker et al. 1997;Molau 2001;Høye et al. 2007a;Cadieux et al. 2008;Ellebjerg et al. 2008).In addition, Keller and Körner (2003) found S. oppositifolia time of flowering not to be sensitive to either temperature or photoperiod, which also supports our findings of no trend towards earlier flowering time for S. oppositifolia.However, our findings differ from an ITEX study at Alexandra Fiord, Ellesmere Island, Nunavut, Canada, where summer-flowering plants trended towards later flowering (Bjorkman et al. 2015).This discrepancy could be a consequence of (1) monitoring start of flowering time, rather than peak flowering time, the former of which typically has more variation (Miller-Rushing et al. 2008), (2) experimental warming studies underprediction of responses to climate change (Wolkovich et al. 2012), or (3) a possible increase in snow accumulation over the study period at Alexandra Fiord (Bjorkman et al. 2015).
Twenty years, however, appears to be too short of a time period to see a consistent yearto-year trend in earlier flowering or fruiting times.In comparison to the previous decade, 2013 was exceptionally cold and snowy in the Canadian Arctic Archipelago (Fig. 4B).The addition of the 2013 flowering and fruiting data to the analysis therefore, not too surprisingly, eliminated the earlier flowering and fruiting trend that had been seen in D. integrifolia up until 2012.However, the rolling average, which accounts for interannual variation, showed D. integrifolia' flowering and fruiting times advancing with the inclusion of the 2013 data.This should come as no surprise insofar as multiyear moving averages smooth through single-year anomalies.It will be interesting to see what the next few years of temperature and phenology data tell us about the trend of D. integrifolia towards earlier flowering and fruiting.
Because Arctic temperature rises are predominantly occurring at the end of the growing season (and winter), we had expected to see earlier fruiting times for S. oppositifolia as well as D. integrifolia.An earlier fruiting trend for S. oppositifolia might have been masked by the fact that ripening fruits continued to be counted as open flowers even when the petals were wilted because the Tanquary Fiord ITEX protocol required counting something as an open flower until the petals dropped.Saxifraga oppositifolia petals persist into the fruiting stage and hence, some mature fruits may have been counted as open flowers when they should have been counted as mature fruits.The protocol has now been changed and the use of the method whereby the peak flowering date is assumed to be the date of 90% of the maximum flower counts will enable the data to be analysed consistently across the pre-and post-protocol change.
In contrast to some alpine and Arctic studies (e.g., Høye et al. 2007a;Anderson et al. 2012), but in agreement with others (e.g., Kudo and Hirao 2006;Kimball et al. 2014), we found no relationship between flowering time and snowmelt date for S. oppositifolia and D. integrifolia.This could perhaps be because snow accumulation at Tanqaury Fiord is low, with less than 50 mm precipitation per year (Edlund and Alt 1989;Thompson 1994;Przybylak 2003).Similarly, a long-term study in the central highlands of Iceland, another polar desert region, also found no relationship between time of flowering and snowmelt date except in two years with high snow accumulation (Thórhallsdóttir 1998).The lack of a relationship between flowering time and snowmelt date could also be an artifact of the arrival time of park staff who, in some years, arrived too late to record an accurate first snowfree day.
At Tanquary Fiord, the previous year's late-summer temperatures were correlated with the number of D. integrifolia flowers opening the following year.This is to be expected because Arctic plants preform their flower buds in the year or years prior to flowering, presumably to maximise reproductive success in the very short Arctic growing season (Sørensen 1941).Given the dependence on the prior year's temperatures to form flower buds for the next year, extreme events and more varied climates could be problematic for reproductive success of Arctic plants (Inouye and McGuire 1991;Inouye 2008).The correlation of the number of S. oppositifolia flowers with the previous year's October mean minimum temperatures may indicate a greater susceptibility of S. oppositifolia flower buds to cold damage because the flower buds are aboveground, while D. integrifolia flower buds are belowground.There is likely little snow in October at Tanquary Fiord to insulate and protect the aboveground buds of S. oppositifolia when the temperatures drop dramatically (Fig. 5).
The sampling protocol used in this monitoring program was to select 25 plants at the start of the monitoring program and monitor these same plants each year.The disadvantage, as was seen at Tanquary Fiord, is that Arctic plants do not flower every year and some of the plants may die, reducing the yearly sample size.Nevertheless, with the exception of one year for each species, the sample size remained reasonably close to the original sample size across the 20-year period.An alternative approach is to select 25 plants each year that have indications that they will flower to ensure a more consistent sample size each year.However, the alternative approach could bias the results towards earlier-flowering plants, is more labour intensive than the approach used at Tanquary Fiord, introduces variance in the time series due to different phenotypes of the individual plants, and could introduce inconsistencies in the plant selection method from year to year.The advantage of the approach used at Tanquary Fiord is that it is less labour intensive than the alternative approach and the selected sample is consistent year after year and eliminates variation in the time series due to individual plant phenotypes.
Unlike temperate regions, but comparable to the Arctic in general, the most significant warming at Tanquary Fiord is towards the end of the growing season and during the winter (Przybylak 2003;McBean 2004;Furgal and Prowse 2007;AMAP 2011).Tanquary Fiord's monthly mean minimum temperatures are rising more than the monthly mean maximum temperatures and monthly mean temperatures.Monthly mean minimum temperatures are also rising more than monthly mean maximum temperatures and monthly mean temperatures in temperate regions (Menne et al. 2010).
Our study indicates a different pattern in the response of Arctic plants to a warming climate compared to temperate regions, with the midsummer flowering species D. integrifolia trending towards earlier flowering and fruiting, but the early spring flowering species S. oppositifolia showing no trend towards earlier flowering and fruiting.Hence, changes in ecological and community-level interactions could play out differently in the Arctic than in temperate regions (Hegland et al. 2009;McKinney et al. 2012;Høye et al. 2013).
. In comparison to temperate regions, the different pattern and timing of climate change in the Arctic might result in different phenological responses of Arctic plants to climate change.By documenting Arctic phenological responses to climate change, we add to the picture of how plants respond to climate change.

Flowering
and fruiting data Every three days during spring and summer of 1994-2014, park staff recorded the number of flowers open on each tagged S. oppositifolia and D. integrifolia plant (Parks Canada 2002).A flower was considered open when the petals were open and the anthers visible.The protocol required that a flower be considered open until all of its petals dropped.

Fig. 1 .
Fig. 1.International Tundra Experiment (ITEX) phenology monitoring area at Tanquary Fiord, Quttinirpaaq National Park, Ellesmere Island, Nunavut, Canada.(A) Dryas integrifolia tagged plants and (B) Saxifraga oppositifolia plant 19.The string is to aid in accurately counting the large number of flowers.

Fig. 4 .
Fig. 4. Change in temperatures from 1989 to 2014 at Tanquary Fiord, Quttinirpaaq National Park, Ellesmere Island, Nunavut, Canada.(A) Annual mean minimum, annual mean maximum, and annual mean temperatures and (B) July, August, September, and October mean minimum temperatures (Table2).

Table 1 .
Standard least squares models showing how much of the Dryas integrifolia and Saxifraga oppositifolia peak flower and fruit day of year is explained by year and April, May, June, and July mean temperatures.