Environmental Research Letters

LETTER bull OPEN ACCESS

Vulnerability and resilience of the carbonexchange of a subarctic peatland to an extremewinter eventTo cite this article Frans-Jan W Parmentier et al 2018 Environ Res Lett 13 065009

View the article online for updates and enhancements

Related contentLow impact of dry conditions on the CO2exchange of a Northern-Norwegianblanket bogMagnus Lund J W Bjerke B G Drake etal

-

Tundra shrub effects on growing seasonenergy and carbon dioxide exchangePeter M Lafleur and Elyn R Humphreys

-

Record-low primary productivity and highplant damage in the Nordic Arctic Regionin 2012 caused by multiple weather eventsand pest outbreaksJarle W Bjerke Stein Rune Karlsen KjellArild Hoslashgda et al

-

This content was downloaded from IP address 13022524108 on 13062018 at 1212

Environ Res Lett 13 (2018) 065009 httpsdoiorg1010881748-9326aabff3

LETTER

Vulnerability and resilience of the carbon exchange of asubarctic peatland to an extreme winter event

Frans-Jan W Parmentier1678 Daniel P Rasse1 Magnus Lund12 Jarle W Bjerke3 Bert G Drake4 SimonWeldon1 Hans Toslashmmervik3 and Georg H Hansen5

1 Norwegian Institute of Bioeconomy Research (Nibio) Department of Soil Quality and Climate Change Arings Norway2 Aarhus University Arctic Research Centre Department of Bioscience Roskilde Denmark3 Norwegian Institute for Nature Research (NINA) FRAMmdashHigh North Centre for Climate and the Environment Tromsoslash Norway4 Smithsonian Environmental Research Center Edgewater MD United States of America5 Norwegian Institute for Air Research FRAMmdashHigh North Centre for Climate and the Environment Tromsoslash Norway6 Department of Geosciences University of Oslo Oslo Norway7 Department of Physical Geography and Ecosystem Science Lund University Lund Sweden8 Author to whom any correspondence should be addressed

OPEN ACCESS

RECEIVED

16 February 2018

REVISED

13 April 2018

ACCEPTED FOR PUBLICATION

25 April 2018

PUBLISHED

1 June 2018

Original content fromthis work may be usedunder the terms of theCreative CommonsAttribution 30 licence

Any further distributionof this work mustmaintain attribution tothe author(s) and thetitle of the work journalcitation and DOI

E-mail frans-janthissideofthearcticorg

Keywords arctic browning CO2 exchange extreme winter event carbon cycle eddy covariance northern ecosystems

Supplementary material for this article is available online

AbstractExtreme winter events that damage vegetation are considered an important climatic cause of arcticbrowningmdasha reversal of the greening trend of the regionmdashand possibly reduce the carbon uptake ofnorthern ecosystems Confirmation of a reduction in CO2 uptake due to winter damage howeverremains elusive due to a lack of flux measurements from affected ecosystems In this study we reporteddy covariance fluxes of CO2 from a peatland in northern Norway and show that vegetation CO2uptake was delayed and reduced in the summer of 2014 following an extreme winter event earlier thatyear Strong frost in the absence of a protective snow covermdashits combined intensity unprecedented inthe local climate recordmdashcaused severe dieback of the dwarf shrub species Calluna vulgaris andEmpetrum nigrum Similar vegetation damage was reported at the time along sim1000 km of coastalNorway showing the widespread impact of this event Our results indicate that gross primaryproduction (GPP) exhibited a delayed response to temperature following snowmelt From snowmeltup to the peak of summer this reduced carbon uptake by 14 (0ndash24) g C mminus2 (sim12 of GPP in thatperiod)mdashsimilar to the effect of interannual variations in summer weather Concurrentlyremotely-sensed NDVI dropped to the lowest level in more than a decade However bulkphotosynthesis was eventually stimulated by the warm and sunny summer raising total GPP Speciesother than the vulnerable shrubs were probably resilient to the extreme winter event The warmsummer also increased ecosystem respiration which limited net carbon uptake This study shows thatdamage from a single extreme winter event can have an ecosystem-wide impact on CO2 uptake andhighlights the importance of including winter-induced shrub damage in terrestrial ecosystem modelsto accurately predict trends in vegetation productivity and carbon sequestration in the Arctic andsub-Arctic

Introduction

The frequency of extreme winter warming events isincreasing in the Arctic (Vikhamar-Schuler et al 2016Graham et al 2017) and these episodes are capable ofcausing widespread and severe plant damage (Bjerkeet al 2017) When warm spells melt away snow in the

middle of winter shrubs and other vegetation are leftvulnerable to a subsequent return to freezing condi-tions (Bokhorst et al 2011) A partial melt and re-freezeof snow is also damaging due to the formation ofthick hermetic ground ice (Bjerke et al 2015 Mil-ner et al 2016) These extreme winter events may bean important driver of arctic and subarctic browning

copy 2018 The Author(s) Published by IOP Publishing Ltd

Environ Res Lett 13 (2018) 065009

(Phoenix and Bjerke 2016)mdashreductions in greennessthat have been observed by satellites (Bhatt et al 2013)

The recent browning of the Arctic appeared assomewhat of a surprise since satellite data had shown agreening of the region until recently (Bhatt et al 2014)Field observations connected these past increases inremotely sensed greennessmdashexpressed as NDVI (nor-malized difference vegetation index)mdashto an expansionof shrubs that responded to increases in summerwarmth (Myers-Smith et al 2011 Elmendorf et al2012) Despite continued warming large parts ofthe Arctic have exhibited the contrasting process ofbrowning in recent years which has been attributedto a multitude of processes that affect vegetationcover including fires outbreaks of pests and fungipermafrost degradation flooding and changes in graz-ing pressure (Cohen et al 2013 Bjerke et al 2014Phoenix and Bjerke 2016 Lara et al 2018) Despitethe broad range of possible causes of arctic browningextreme winter events that affect snow cover and icingare considered the main climatic cause (Bjerke et al2014) and the subsequent impact on the arctic carboncycle may be large The widespread vegetation dam-age indicated by arctic browning implies a reductionin vegetation productivity and possibly a reductionin the net uptake of CO2 by affected ecosystems

However the vulnerability and resilience of theCO2 exchange of ecosystems to extreme winter eventsremains unclear due to a dearth of flux measurementsin damaged areas While numerous eddy covariancetowers have been deployed across the arctic and sub-arctic in recent years almost all of them are placedin areas where extreme winter events have not (yet)occurred ormdashperhapsmdashhave not been detected Theimpact on the CO2 exchange of ecosystems thereforehas so far been assessed through small-scale manip-ulation experiments and flux chambers (Bokhorstet al 2011 Zhao et al 2016 2017) More commonlyresearch focuses on phenology and mortality ratherthan the carbon budget (Bjerke et al 2015 Preece et al2012 Joslashrgensen et al 2010 Milner et al 2016) Dueto these small scales and general lack of flux mea-surements it remains largely unknown whether CO2fluxes are impacted by extreme winter events at thelandscape scale

In this study therefore we present a dataset span-ning five summers from 2010 to 2014 of the CO2exchange of a blanket bog located on the island ofAndoslashya in northern Norway In January 2014 duringthe last year of measurements boreal Norway experi-enced a severe drought combined with a lack of snowand strong frost which led to widespread vegetationdamage along a north-south transect of Norwegiancoast about 1000 km in length (Meisingset et al 2015Timmermann et al 2015 Bjerke et al 2017) The eddycovariance tower on Andoslashya where frost drought alsodamaged shrub vegetation was the only one to capturethis extreme winter event In connection to this eventthis study sets out to answer two questions did the

winter damage to shrub vegetation lead to a substantialreduction in vegetation productivity in the followingsummer and if so how large was this reduction whenput in the context of inter-annual variations in CO2exchange

Materials and methods

Site descriptionThis research focuses on a large blanket bog locatedon the island of Andoslashya in northern Norway near thesmall settlement of Saura (69 08rsquoN16 01rsquoE 17 maslsee figure 1) The Saura bog is located nearly 300 kmNorth of the Arctic Circle but the climate is mild forthis latitude due to the influence of the nearby AtlanticOcean Long-term climate data (1981ndash2010) from aweather station near the town of Andenes (sim17 km tothe North) operated by the Norwegian MeteorologicalInstitute indicate an average temperature of 114 C forJulyndashAugust and minus14 C for JanuaryndashFebruary Aver-age annual precipitation is 1030 mm This classifies theclimate as being on the boundary between the subpolaroceanic and subarctic climate zones (Koppen classifi-cations Cfc and Dfc respectively) The wet climate andrelatively cool summers have been favorable for peatformation on the island and by comparison to a simi-lar bog a few km to the south-west (Vorren et al 2007)it is expected that peat depth at the Saura field site isabout 2ndash3 m

The Saura bog is characterized by relatively dryhummockswithhollows inbetweenTheratiobetweenthe two is about 7030 with an estimated heightdifference of 015 m Vegetation on the hummocksconsists of dwarf shrubs (Calluna vulgaris Empetrumnigrum Vaccinium uliginosum and Rubus chamae-morus) mosses (Dicranum scoparium Hylocomiumsplendens Pleurozium schreberi Racomitrium lanug-inosum Sphagnum fuscum) and lichens (Cladoniaspp) Hollows are dominated by Sphagnum mosses(S warnstorfii S magellanicum S cuspidatum) andsedges (Carex rariflora) In August 2009 a vegetationsurvey showed that the cover of cryptogams (lichensand bryophytesmosses) was almost twice as high asthe cover of vascular plants (76 versus 44 whenaccounting for overlap) and lichens covered 41 ofthe hummocks on average Shrub height was very lowwith an average of 005 m

InstrumentationDuring the summer of 2008 an eddy covariance towerwas placed near the center of the Saura bog A CSAT33D Sonic anemometer (Campbell Sci UK) and a Li-7500 open-path gas-analyzer (Li-Cor NE USA) wereinstalled at a height of 23 m to measure wind speedand concentrations of CO2 and H2O Data from thissetup was collected at 10 Hz on a CR3000 data logger(Campbell Sci UK) Ancillary meteorological data wasmeasured on a separate tower at approximately 10 m

2

Environ Res Lett 13 (2018) 065009

Figure 1 Aerial overview of the Saura peat bog where the lsquoxrsquo denotes the location of the tower The inset on the top left shows the sitersquoslocation in Norway and the continuous line denotes the Arctic Circle Background image source Google maps acquired on May 192013

distance and averaged for each half hour This includedair temperature at canopy height (5 cm Tcanopy) and at2 m (Tair) relative humidity (RH HMP45C VaisalaFinland) photosynthetic photon flux density (PPFDLI-190 Li-Cor NE USA) global solar radiation (RgLI-200 Li-Cor NE USA) net radiation (Rn Qlowast7REBS USA) soil temperature (Tsoil TCAV-L Camp-bell Sci UK) and soil water content (SWC CS616Campbell Sci UK) Due to large gaps in the data from2008 and 2009 that preclude detailed time series anal-ysis this study focuses on the last five summers ofthe dataset from 2010 to 2014 The processing of thedata was previously described in detail by Lund et al(2015) while the partitioning of the fluxes into GPPand Reco followed Lasslop et al (2010) Details of thesemethods are given in the supplementary informationavailable at stacksioporgERL13065009mmedia

Survey of vegetation damageIn April 2015 shortly after snowmelt we analyzed thevegetation at the Saura bog in eight stratified ran-domly selected plots of 40 cmtimes 60 cm along a 125 mlong west-east transect passing 1 m from the towerAt 15 m intervals along the transect plots were ran-domly chosenwithin a radius of 5 m The two evergreendwarf shrubs Calluna vulgaris and Empetrum nigrumshowed signs of damage typically caused by winterdesiccation (Hancock 2008 Bokhorst et al 2011)mdashie intact but brown leaves with strongest damageratio at top shoots and decreasing towards the baseThe leaves were pale brown and flat indicating thatleaves had died the year before Recently dead leavesare inflated and chestnut brown rather than pale

brown while leaves that have been dead for longerthan a year turn grey and shriveled and easily detachwhen being touched (Bokhorst et al 2009 Bjerke et al2017) Hence we estimated the green and pale browncover of both plants to calculate plot-level damageratios

Assessment of vegetation developmentIf large parts of an ecosystem are damaged we expectthat it will take more time following snowmelt and ahigher amount of accumulated degree days for pho-tosynthesis rates to develop compared to other yearsHowever comparisons to other years are complicatedby high variations in daily GPP due to changes inincoming solar radiation Therefore rather than ana-lyzing GPP rates under observed radiation levels weuse the photosynthetic parameters from the partition-ing model of Lasslop et al (2010) to calculate GPP ratesat light saturation (GPPsat ) In the caseofAndoslashyamax-imum light levels in summer are about 700 W mminus2 andGPPsat is calculated as follows

GPPsat =120572120573119877119892

120572119877119892 + 120573(1)

where 120572 (in 120583mol C Jminus1) is the canopy light utiliza-tion efficiency which represents the initial slope of thelight response curve and 120573 (120583mol C mminus2 sminus1) is themaximum CO2 uptake rate when light availability isnon-limiting (119877119892 rarr infin) Rg (W mminus2) is the incom-

ing radiation and in this case fixed to 700 W mminus2 tocalculate GPPsat under typical clear-sky conditions

3

Environ Res Lett 13 (2018) 065009

Figure 1 Aerial overview of the Saura peat bog where the lsquoxrsquo denotes the location of the tower The inset on the top left shows the sitersquoslocation in Norway and the continuous line denotes the Arctic Circle Background image source Google maps acquired on May 192013

distance and averaged for each half hour This includedair temperature at canopy height (5 cm Tcanopy) and at2 m (Tair) relative humidity (RH HMP45C VaisalaFinland) photosynthetic photon flux density (PPFDLI-190 Li-Cor NE USA) global solar radiation (RgLI-200 Li-Cor NE USA) net radiation (Rn Qlowast7REBS USA) soil temperature (Tsoil TCAV-L Camp-bell Sci UK) and soil water content (SWC CS616Campbell Sci UK) Due to large gaps in the data from2008 and 2009 that preclude detailed time series anal-ysis this study focuses on the last five summers ofthe dataset from 2010 to 2014 The processing of thedata was previously described in detail by Lund et al(2015) while the partitioning of the fluxes into GPPand Reco followed Lasslop et al (2010) Details of thesemethods are given in the supplementary informationavailable at stacksioporgERL13065009mmedia

Survey of vegetation damageIn April 2015 shortly after snowmelt we analyzed thevegetation at the Saura bog in eight stratified ran-domly selected plots of 40 cmtimes 60 cm along a 125 mlong west-east transect passing 1 m from the towerAt 15 m intervals along the transect plots were ran-domly chosenwithin a radius of 5 m The two evergreendwarf shrubs Calluna vulgaris and Empetrum nigrumshowed signs of damage typically caused by winterdesiccation (Hancock 2008 Bokhorst et al 2011)mdashie intact but brown leaves with strongest damageratio at top shoots and decreasing towards the baseThe leaves were pale brown and flat indicating thatleaves had died the year before Recently dead leavesare inflated and chestnut brown rather than pale

brown while leaves that have been dead for longerthan a year turn grey and shriveled and easily detachwhen being touched (Bokhorst et al 2009 Bjerke et al2017) Hence we estimated the green and pale browncover of both plants to calculate plot-level damageratios

Assessment of vegetation developmentIf large parts of an ecosystem are damaged we expectthat it will take more time following snowmelt and ahigher amount of accumulated degree days for pho-tosynthesis rates to develop compared to other yearsHowever comparisons to other years are complicatedby high variations in daily GPP due to changes inincoming solar radiation Therefore rather than ana-lyzing GPP rates under observed radiation levels weuse the photosynthetic parameters from the partition-ing model of Lasslop et al (2010) to calculate GPP ratesat light saturation (GPPsat ) In the caseofAndoslashyamax-imum light levels in summer are about 700 W mminus2 andGPPsat is calculated as follows

GPPsat =120572120573119877119892

120572119877119892 + 120573(1)

where 120572 (in 120583mol C Jminus1) is the canopy light utiliza-tion efficiency which represents the initial slope of thelight response curve and 120573 (120583mol C mminus2 sminus1) is themaximum CO2 uptake rate when light availability isnon-limiting (119877119892 rarr infin) Rg (W mminus2) is the incom-

ing radiation and in this case fixed to 700 W mminus2 tocalculate GPPsat under typical clear-sky conditions

3

Environ Res Lett 13 (2018) 065009

Figure 2 Weather conditions during the frost drought event in the winter of 20132014 Hourly measured temperature at canopyheight (5 cm) is indicated with the blue line and the orange line shows soil temperature at 5 cm depth The thick black line denotesmodeled snow cover from seNorge (wwwsenorgeno) Note that the temperature measurement at canopy height may have been insidethe snow pack rather than exposed to the outside air before snowmelt completed

Snow and NDVI datasetsIn addition to the data collected by the eddy covari-ance and meteorological towers information on snowcover and vegetation productivity was obtained fromexternal datasets to compare the 2014 winter to thelong-term record Snow cover was obtained from TheNorwegian Water Resources and Energy Directorate(NVE) which provides maps of snow cover and inter-polated air temperature for the whole of Norway ona daily basis and at a 1 kmtimes 1 km resolution (wwwsenorgeno) This model performs well for Norway(Saloranta 2012) and snow and temperature data forthe location of the tower were retrieved starting in 1963For each year the total amount of freezing degree daysduring snowless periods was calculated as a measureof potential vegetation damage due to frost droughtThese totals were calculated separately for polar night(28November28ndash17 January) and theperiod thereafteruntil the start of the growing season

To ascertain whether the vegetation damage at theSaura bog was visible as a browning event remotelysensed NDVI data were downloaded from the MODISLand Product Subsets project (ORNL DAAC 2017)which provides subset data from both the Terra andAqua satellites at a 250 mtimes 250 m spatial resolutionThe size of a MODIS pixel happens to be very com-parable to the footprint of the tower ie the upwindsurface area that contributes to the measured fluxThe 90 fetch length is typically about 200 m (fig-ure S1) Changes in MODIS NDVI data are thereforeexpected to provide useful information on the ecosys-tem at a similar scale to that of the flux tower OnlyNDVI data with the highest quality flag was keptand maps of NDVI were visually inspected for obvi-ous outliers which were then rejected Few additionalmeasurements had to be rejected during the sum-mers of 2010ndash2014 with one invalid measurementin the summers of 2010 and 2013 and two in 2012

Following this quality check NDVI values were aver-aged over the four pixels closest to the location ofthe tower

Results

The extreme winter of 20132014In January 2014 large parts of coastal Norway werefree of snow following a winter warm spell Once thisevent passed and temperature dropped back below0 C snow cover remained absent and vegetation alonglarge parts of the Norwegian arctic and subarctic coastwere exposed to severe frost leading to wide-spreaddamage to shrub vegetation due to winter desiccation(Bjerke et al 2017) The Saura bog on Andoslashya wasno different in that regard Remote sensing and datamodels from the NVE indicate that snow cover wasabsent during almost all of January and February (fig-ures 2 and S2) The strong drop in soil temperaturealso indicates that snow cover was absent while thetotal amount of precipitation at the nearby meteoro-logical station of Andenes was 10 mm in January 2014From January 9ndashFebruary 2 temperature at canopyheight was well below 0 C approaching minus15 C onseveral occasions and frost events kept occurring reg-ularly throughout February (figure 2) Although notas strong as in the preceding month they coincidedwith clear sky conditions and plenty of incoming sun-light Such conditions can lead to frost desiccationWhile thaw-freeze events may happen occasionally onAndoslashya the total amount of freezing degree daysfor periods without snow was unprecedented in theclimate data going back to 1963 (figure 3) and espe-cially high during the part of the winter where sunlighthad returned

The three weeks of frost combined with intensedrought severely damaged the shrub species Calluna

4

Environ Res Lett 13 (2018) 065009

Figure 3 Total amount of freezing degree days (sum daily average temperatures lt 0 C) in the absence of snow cover for each winterfrom 1964ndash2017 Data is shown in different colors for polar night (28 Novemberndash14 January) and the period thereafter when sunlighthas returned

Figure 4 Percentage of frost-damaged vegetation per species per plot at the Saura bog as surveyed in April 2015 Empetrum nigrumwas present but showed no damage in plot 1ndash4

vulgaris and Empetrum nigrum (heather and crow-berry) as surveyed on April 26 2015 and shown infigure 4 Both shrub species had large amounts ofdamaged vegetation dieback of Calluna vulgaris wasrecorded in all plots ranging from low to high whileEmpetrum nigrum was only affected in four plotsalbeit severely (gt50 of dead vegetation) in two Nodamage to Empetrum was observed in the other fourplots On average 43 of Calluna vulgaris and 27 ofEmpetrum nigrum was damaged or dead

Year-to-year variations in summer weather condi-tions and CO2 budgetsSummer weather conditions (JunendashAugust) differedconsiderably among the years studied (table 1) Thesummers of 2010 and 2012 were cold with an aver-age temperature of 90 and 91 C and temperaturenever exceeded 20 C in both years 2011 was consider-ably warmer at 107 C with a maximum at 241 C

The summers of 2013 and 2014 were the warmestwith average temperatures of 115 and 114 C andmaximumtemperaturesof 249 Cand256 C respec-tively The wettest summers occurred in 2010 and2013 although 2012 was nearly as wet Precipitation in2011 and 2014 was sim30 to sim45 lower The sun-niest summer of these five years occurred in 2014although 2011 was not that dissimilar with 5 lessincoming radiation The other three summers receivedsim20 less radiation than in 2014 Detailed plots oftemperature radiation and vapor pressure deficit areshown in figure S3

In figure 5 the fluxes of GPP Reco and NEE areshown for the years 2010ndash2014 and split up for themonths of June to August June is normally the monthin which green-up occurs and maximum GPP ratesare reached in the first half of July By mid-July daysshorten and light conditions begin to decline whichgradually lowers GPP over the rest of the summer

5

Environ Res Lett 13 (2018) 065009

Table 1 Average air temperature at 2 m (Tair ) maximum recorded air temperature (Tmax) average global radiation (Rg) total precipitation(P) and cumulative CO2 fluxes (NEE GPP and Reco) at the Saura bog from 1 Junendash31 August during 2010ndash2014 All data was recorded atthe site apart from P which was measured sim17 km away at the weather station near the local town of Andenes Standard deviations of Tairand Rg are determined on daily values The ranges given for the carbon fluxes represent random flux uncertainty rather than ordinarystandard deviations Due to model uncertainties the sum of GPPmod and Recomod does not exactly equal NEEobs

Tair(C) Tmax(

C) Rg (Wmminus2) P (mm) NEEobs(g C) GPPmod(g C) Recomod(g C)

2010 90plusmn 23 198 1536plusmn 675 2145 minus920plusmn 15 minus1923plusmn 22 971plusmn 212011 107plusmn 24 241 1795plusmn 886 1423 minus1066plusmn 17 minus2391plusmn 24 1281plusmn 242012 91plusmn 20 198 1581plusmn 791 1869 minus1157plusmn 18 minus2168plusmn 25 993plusmn 262013 115plusmn 24 249 1540plusmn 761 2102 minus1062plusmn 18 minus2390plusmn 25 1270plusmn 252014 114plusmn 37 256 1872plusmn 830 1160 minus999plusmn 17 minus2464plusmn 26 1423plusmn 25

Figure 5 Total amounts of (a) GPP (b) Reco and (c) NEE for the months June July and August from 2010ndash2014 GPP and NEE areplotted here as positive values for a straightforward visual comparison at the same scale

Figure 5 clearly shows that 2010 had the lowest GPPIn that year snowmelt didnrsquot occur until the first weekof Maymdashtwo to four weeks later than in the otheryears (table S1) Moreover that summer was also thecoldest with the least amount of incoming radiation(table 1) limiting vegetation development The follow-ing year was much warmer and sunnier with snowmeltin early April and GPP in June and July was high2012 also had less GPP in June but July and Augustwere similar to the other years Photosynthesis ratesin June 2013 were exceptionally high but August ofthat year had the lowest cumulative flux of all fiveyears Finally 2014 started off slowly but had veryhigh photosynthesis rates in July and August due towarm and sunny weather which provided exceptionalgrowing conditions

The respiration by the ecosystem Reco followed apredictable pattern for all years where the warmestsummers had the highest amounts of respirationand the coldest summers the lowest (figure 5 table1) The summer with the highest NEE (differencebetween GPP and Reco) therefore occurred in 2012when both low temperatures and wet conditions sup-pressed respiration Such behavior is not uncommonfor high latitude ecosystems where changes in Reco andGPP can be more pronounced than changes in NEE(Parmentier et al 2011) A detailed overview of GPPReco and NEE is given in figure S4

Response of GPP to environmental forcingThe observations of vegetation damage (figure 4)appear to be at odds with the large increase in GPP in

2014 (figure 5) Despite the documented frost damageecosystem functioning seems to have been unaffectedHowever the exceptional growing conditions in Julyand August of 2014 when compared to the otheryears obscures any reductions in vegetation produc-tivity due to winter damage To assess the effect ofwinter damage on GPP the interannual variability influxes due to differences in radiation and temperatureshould first be removed

In figure 6(a) the potential photosynthesis rateat 700 W mminus2 (GPPsat) has been plotted against theamount of days following snowmelt up until peaksummer (day of year 200) In this figure it becomesclear that in 2010 and 2013 plant growth started veryquickly following snowmelt and GPPsat increased tomore than 3120583mol mminus2 sminus1 within the first month Inboth years snowmelt was immediately followed by aperiod of warm and sunny weather and vegetationdeveloped promptly In the other years temperaturesfollowing snowmelt stayed low vegetation develop-ment took longer and photosynthesis rates did notincrease beyond 3 120583mol mminus2 sminus1 until sim60 days aftersnowmelt However when we plot GPPsat against theamount of accumulated degree days the differencesbetween years strongly reduce in the period up tosim300 D as shown in figure 6(b)

At values greater than sim300 D however there areclear divergent patterns in 2011 2012 and 2013 GPPsatcontinued its linear response to accumulated degreedays and in all three years GPPsat reached its maxi-mum value after another two or three weeks In 2014this linear response to temperature increases halted

6

Environ Res Lett 13 (2018) 065009

Figure 6 7 day running mean of GPP at saturated light levels (700 Wmminus2) vs d after snow melt and the temperature sum followingsnow melt expressed in degree days (D) Time series shown are from snowmelt until day of year 200 (July 19 in non-leap years)

only to pick up at a later time Vegetation develop-ment took another five weeks up until the second halfof July Of all snow-free seasons only 2010 showed adegree-day response similar to that of 2014 Howevera simple comparison of these two years is problematicsince weather conditions in 2010 were vastly differ-ent from 2014 snowmelt occurred 35 weeks later andincoming radiation and temperature were much lower(table 1 figure S3)

A delayed response in 2014 similar to a cold andcloudy year is the kind of behavior that would beexpected when a high number of shrubs are damagedand their contribution to GPP is lowered (Bokhorst etal 2011) It appears therefore that the capacity of theecosystem to take up carbon was reduced during thesummer of 2014

Toquantify this reductionwe interpolated thepho-tosynthetic parameters 120572 and 120573 of the years 2010ndash2013obtained from the partitioning model (Lasslop et al2010) to specific dates in 2014 by using the temper-ature sum as a lookup tablemdashsimilar to figure 6(b)This interpolation approximates what the photosyn-thetic parameters 120572 and 120573 would have been in 2014if the vegetation had developed with temperature as inthe other years Subsequently GPP was calculated withthe observed radiation in 2014 following equation 1from the day that 300 D was reached (day of year159) up until the peak of summer (day of year 200)The period following the peak of summer is omittedto avoid an influence due to varying onsets of senes-cence (the whole time series is shown in figure S4) Amedian of these estimates showed that the vegetationcould have photosynthesized an additional 14 g C mminus2

in 2014 with an upper estimate of 24 g C mminus2 (whencompared to 2013) and a lower estimate of 0 g C mminus2

(when compared to 2010)mdashif there had been no neteffect from the damaged vegetation Since cumulativeGPP was 116 g C mminus2 during the same period in 2014this flux could have been sim12 higher with a lowerand upper estimate of 0 and 21

Comparison to remote sensing dataIn figure 7 a time series is plotted of the maximum andaverage NDVI value for each summer (day of year 175ndash225) from 2000ndash2017 which shows that 2014 had thelowest value in a decademdashup to that point The averagevalue for the summer of 2010 was nearly as low butwith a higher maximum The peak season was missedin 2013 due to bad coverage (figure S5) and NDVIvalues are probably underestimated for that year sinceGPP was high (figure 5) Average NDVI values in 2014are lower than in the other measurement years but notunprecedented in the long-term satellite record Thisis probably due to the excellent growing conditions inthe summer of 2014 which boosted vegetation growthafter mid-summer (figures 5 and S4)

However the maximum NDVI value reached in2014 was the second-lowest until then (after 2003) andit took much longer than normal to reach the max-imum (table S2 figures 7 and S5) On average peakNDVI values are reached on day of year 207plusmn 11 daysbut the maximum in 2014 was on day of year 222(August 10) The low NDVImdasha browning eventmdashandthe delayed peak were probably due to the large amountof damaged vegetation The only years with a later time-to-peak were 2007 (223) and 2017 (225) althoughconsiderable uncertainty exists on these dates due tocloud cover and their average values are much higher(figures 7 and S5)

Interestingly average NDVI values were at theirall-time lowest in 2015mdashthe year following the extremewinter event The browning event worsened indicatingno recovery of the ecosystem and this was possibly dueto another extreme winter (figure 3) Unfortunatelyflux measurements at the Saura peat bog had ceased by2015 and we do not know how this was reflected inthe ecosystem fluxes The same goes for the upwardsreturn of NDVI levels in 2016 However NDVI showsa reasonable agreement with GPPsat (Figure S5) andit is therefore likely that photosynthesis rates in 2015were lower than in 2014

7

Environ Res Lett 13 (2018) 065009

Figure 7 Maximum and average NDVI values for the summer (day of year 175ndash225) from 2000ndash2017 obtained from MODIS (v6)averaged over the four pixels closest to the position of the eddy covariance tower The years covered in this study are shown in colorThe values for 2013 are probably underestimated due to a coverage gap during peak summer A more detailed NDVI time series isshown in figure S5

Discussion

Impact of the 20132014 winter on summer CO2exchangeThis study shows that the severe frost drought eventof January and February 2014 unprecedented in theclimate record on Andoslashya led to the strong diebackof the shrub species Calluna vulgaris and Empetrumnigrum Cumulative GPP however was higher in 2014than in other summers This contradictory result canbe explained by the fact that 2014 also had the sunniestand warmest summer of the 5 years in this dataset(table 1) This provided ideal conditions for growth ofundamaged plants

However when interannual variability in radiationand temperature is compensated for it is clear thatvegetation productivity showed a delayed response fol-lowing snowmelt when compared to other years (figure6) This indicates a vulnerability of this ecosystem to theextreme winter event While briefly following a sim-ilar development for GPPsat as for the other yearsa clear departure occurred at two months followingsnowmelt at a point when shrub bud break normallywould occur Although a period of colder weather mayhave contributed to this delayed response this pat-tern remained present when GPPsat was comparedto accumulated degree days Vegetation develop-ment was lagging behind other years most likelydue to the large number of winter-damaged shrubs

After the initial anomaly in GPPsat the ecosystemshowed high photosynthesis rates later in the summer(figures 5 and S4) indicating some resilience to theextreme winter event A possible explanation for thismay be that the ecosystem partly recovered its car-bon uptake through compensatory growth (Bokhorstet al 2011) spurred on by the exceptionally warmand sunny weather of July and August 2014 Highertemperatures however also stimulated ecosystem res-piration with record high respiration in July and

August 2014 (figure 5) It is possible that part ofthese high respiration rates was related to decompos-ing dead plant material limiting NEE but a separationof ecosystem respiration into autotrophic and het-erotrophic respiration rates is not possible with thisdataset In future studies of the impact of extreme win-ter events such effects on respiration need to be takeninto account during field campaigns

Possibility of moisture limitationsIn addition to the documented damage to the shrubsother causes of the lower vegetation productivity atthe Saura bog need to be considered Droughts andheatwaves in particular can reduce the carbon uptakeof an ecosystem when plants close their stomata toconserve water (Lund et al 2012 van der Molen et al2011) This behavior is taken into account by the par-titioning method used in this study where GPP isreduced for high vapor pressure deficit values (Lass-lop et al 2010) Minor events did occur in early July2014 as in other years but at a very low level and foronly a few days which cannot explain the observeddifferences (figure S3) Low soil moisture could beanother limiting factor but 2014 was not exception-ally dry (table 1 figure S6) and significant reductionsin soil moisture occurred after the deviation in GPPrates Moreover a previous study of the Saura bog byLund et al (2015) showed that dry conditions had a lowimpact on the ability of this ecosystem to store carbonIt is therefore unlikely that summer drought condi-tions caused the divergent pattern of GPP as shownin figure 6

Extreme winter events and remotely sensed browningAlthough the Saura bog has experienced multiple win-ters with strong frost in the absence of snow as shownin figure 3 many of these did not lead to strongreductions in NDVI It is striking that the strong frostevent that occurred during polar night in the winter of

8

Environ Res Lett 13 (2018) 065009

20152016 did not negatively affect NDVI values Tothe contrary after two years of browning NDVI val-ues jumped back up to a normal value A possibleexplanation for this may lie in the timing of theseevents The extreme winter events in early 2014 and2015 occurred when the sunlight had returned afterone and a half month of darkness Under sunny con-ditions plants may attempt to transpire but cannotaccess the frozen soil water and they desiccate (Bjerkeet al 2017) The absence of sunlight during the frostevent in the winter of 20152016 probably preventedextensive plant damage due to frost drought Thisshows that the damage of an extreme winter eventvaries depending on its timing within the cold sea-son Other factors such as interannual variation in theamount of frost resistance that was built up may alsohave played a role

Resilienceandvulnerabilityof ecosystemfunctioningto wintertime impactsOur results show a considerable delayed response ofthe vegetation to temperature as shown in figure 6(a)but the estimated impact on GPP varies strongly from24 g C mminus2 to no effect at all when compared to 2010This appears to suggest that the impact of the frostevent on CO2 fluxes could have been negligible butthis is unlikely since weather conditions in the sum-mers of 2010 and 2014 were strongly dissimilar In2010 snowmelt occurred almost three and a half weekslater than in 2014 and values of 300 D werenrsquot reacheduntil June 18 compared to June 8 for 2014 (table S1)Besides this difference in the length of the growing sea-son there was also a stark contrast in the amount ofincoming radiation up to the peak of summer 432 MJin 2010 vs 763 MJ in 2014 The highly unfavorablegrowing conditions in 2010 are reflected in the GPPsatvalues which by mid-summer had not reached thesame maximum uptake as in the other years andaverage summer NDVI values were among the low-est recorded The similar vegetation development in2014 and 2010mdashone of the warmest and sunniest yearsvs the shortest coldest and cloudiest growing seasonin this datasetmdashis in fact a strong indication that theextreme winter event reduced GPPmdashcomparable insize to interannual variations in summer weather Fullyaccounting for large differences in weather remainschallenging which is why a large uncertainty remainson our estimate of the impact of the extreme winterevent on ecosystem carbon exchange

Besides these uncertainties the CO2 uptake of theecosystem may have been somewhat resilient to thefrost drought due to a contribution from vegetationtypes other than shrubs About 30 of the surfacearea of the Saura bog consists of hollows where shrubsare absent and Carex spp is abundantmdashwhich couldhave responded to the warm weather Moreover theSaura bog has a large abundance of lichens and mossesThese functional vegetation groups were not stronglyaffected by the frost drought event and the warm

and sunny weather may have boosted their photosyn-thesis rates In other words while the CO2 exchange ofthis bog was vulnerable at the species level (ie shrubs)to a certain degree it was resilient at the ecosystem levelThe 2014 frost drought event may have had a muchlarger impact on the net CO2 exchange at other affectedareas along the Norwegian coast in places where thefraction of shrubs vs mosses and sedges would havebeen highermdasheg in dry heathlands (Bokhorst et al2009 Bjerke et al 2014)

While the further decline in NDVI in 2015 showsthat the peatland did not recover in the following yearpossibly due to an additional extreme winter eventthe return to normal NDVI values in 2016 shows thatthis ecosystem can recover from an extreme winterevent in a relatively short time Such behavior hasbeen reported before for a browning event in north-ern Scandinavia caused by a winter warming event(Bokhorst et al 2012) If however extreme winterevents will increase in frequency eg every other yearsubsequent browning events may constitute a brown-ing trend In that case the species distribution of anecosystem may change with a lasting effect on CO2 andenergy exchange

Conclusions

The extreme winter event in January 2014 severelydamaged shrubs at the Saura bog and reduced bothvegetation CO2 uptake and NDVI in the followingsummer A comparison with the photosyntheticparameters of other years indicates that the ecosystemcould have taken up an additional 14 (0ndash24) g C mminus2

(sim12 of GPP) from day 159 to 200 if it had not beendamaged This means that the reduction in GPP causedby the winter event of 2014 was similar in size to inter-annual differences due to summer weather conditions(table 1)

Vegetation damage from extreme winter eventsshould be included in model simulations Current landsurface models project an increase in arctic vegeta-tion productivity following high latitude warming (Xiaet al 2017 Zhang et al 2014 Sitch et al 2007) despiterecent browning trends showing the opposite (Phoenixand Bjerke 2016) This suggests an overestimation ofGPP in areas prone to winter damage However speciesthat are more resilient to extreme winter events maycompensate the impact of extreme winter events onthe net CO2 exchange of ecosystems Observationsand modeling studies that focus on the impact ofextreme winter events on CO2 exchange thereforeshould not exclusively focus on vulnerable speciessuch as shrubs but determine the resilience of theecosystem as a whole

This study focused on one extreme winter eventin one particular year but when such events increasein frequency and vegetation is damaged more oftenthis may lead to shifts in ecosystem composition

9

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

Table 1 Average air temperature at 2 m (Tair ) maximum recorded air temperature (Tmax) average global radiation (Rg) total precipitation(P) and cumulative CO2 fluxes (NEE GPP and Reco) at the Saura bog from 1 Junendash31 August during 2010ndash2014 All data was recorded atthe site apart from P which was measured sim17 km away at the weather station near the local town of Andenes Standard deviations of Tairand Rg are determined on daily values The ranges given for the carbon fluxes represent random flux uncertainty rather than ordinarystandard deviations Due to model uncertainties the sum of GPPmod and Recomod does not exactly equal NEEobs

Tair(C) Tmax(

C) Rg (Wmminus2) P (mm) NEEobs(g C) GPPmod(g C) Recomod(g C)

2010 90plusmn 23 198 1536plusmn 675 2145 minus920plusmn 15 minus1923plusmn 22 971plusmn 212011 107plusmn 24 241 1795plusmn 886 1423 minus1066plusmn 17 minus2391plusmn 24 1281plusmn 242012 91plusmn 20 198 1581plusmn 791 1869 minus1157plusmn 18 minus2168plusmn 25 993plusmn 262013 115plusmn 24 249 1540plusmn 761 2102 minus1062plusmn 18 minus2390plusmn 25 1270plusmn 252014 114plusmn 37 256 1872plusmn 830 1160 minus999plusmn 17 minus2464plusmn 26 1423plusmn 25

Figure 5 Total amounts of (a) GPP (b) Reco and (c) NEE for the months June July and August from 2010ndash2014 GPP and NEE areplotted here as positive values for a straightforward visual comparison at the same scale

Figure 5 clearly shows that 2010 had the lowest GPPIn that year snowmelt didnrsquot occur until the first weekof Maymdashtwo to four weeks later than in the otheryears (table S1) Moreover that summer was also thecoldest with the least amount of incoming radiation(table 1) limiting vegetation development The follow-ing year was much warmer and sunnier with snowmeltin early April and GPP in June and July was high2012 also had less GPP in June but July and Augustwere similar to the other years Photosynthesis ratesin June 2013 were exceptionally high but August ofthat year had the lowest cumulative flux of all fiveyears Finally 2014 started off slowly but had veryhigh photosynthesis rates in July and August due towarm and sunny weather which provided exceptionalgrowing conditions

The respiration by the ecosystem Reco followed apredictable pattern for all years where the warmestsummers had the highest amounts of respirationand the coldest summers the lowest (figure 5 table1) The summer with the highest NEE (differencebetween GPP and Reco) therefore occurred in 2012when both low temperatures and wet conditions sup-pressed respiration Such behavior is not uncommonfor high latitude ecosystems where changes in Reco andGPP can be more pronounced than changes in NEE(Parmentier et al 2011) A detailed overview of GPPReco and NEE is given in figure S4

Response of GPP to environmental forcingThe observations of vegetation damage (figure 4)appear to be at odds with the large increase in GPP in

2014 (figure 5) Despite the documented frost damageecosystem functioning seems to have been unaffectedHowever the exceptional growing conditions in Julyand August of 2014 when compared to the otheryears obscures any reductions in vegetation produc-tivity due to winter damage To assess the effect ofwinter damage on GPP the interannual variability influxes due to differences in radiation and temperatureshould first be removed

In figure 6(a) the potential photosynthesis rateat 700 W mminus2 (GPPsat) has been plotted against theamount of days following snowmelt up until peaksummer (day of year 200) In this figure it becomesclear that in 2010 and 2013 plant growth started veryquickly following snowmelt and GPPsat increased tomore than 3120583mol mminus2 sminus1 within the first month Inboth years snowmelt was immediately followed by aperiod of warm and sunny weather and vegetationdeveloped promptly In the other years temperaturesfollowing snowmelt stayed low vegetation develop-ment took longer and photosynthesis rates did notincrease beyond 3 120583mol mminus2 sminus1 until sim60 days aftersnowmelt However when we plot GPPsat against theamount of accumulated degree days the differencesbetween years strongly reduce in the period up tosim300 D as shown in figure 6(b)

At values greater than sim300 D however there areclear divergent patterns in 2011 2012 and 2013 GPPsatcontinued its linear response to accumulated degreedays and in all three years GPPsat reached its maxi-mum value after another two or three weeks In 2014this linear response to temperature increases halted

6

Environ Res Lett 13 (2018) 065009

Figure 6 7 day running mean of GPP at saturated light levels (700 Wmminus2) vs d after snow melt and the temperature sum followingsnow melt expressed in degree days (D) Time series shown are from snowmelt until day of year 200 (July 19 in non-leap years)

only to pick up at a later time Vegetation develop-ment took another five weeks up until the second halfof July Of all snow-free seasons only 2010 showed adegree-day response similar to that of 2014 Howevera simple comparison of these two years is problematicsince weather conditions in 2010 were vastly differ-ent from 2014 snowmelt occurred 35 weeks later andincoming radiation and temperature were much lower(table 1 figure S3)

A delayed response in 2014 similar to a cold andcloudy year is the kind of behavior that would beexpected when a high number of shrubs are damagedand their contribution to GPP is lowered (Bokhorst etal 2011) It appears therefore that the capacity of theecosystem to take up carbon was reduced during thesummer of 2014

Toquantify this reductionwe interpolated thepho-tosynthetic parameters 120572 and 120573 of the years 2010ndash2013obtained from the partitioning model (Lasslop et al2010) to specific dates in 2014 by using the temper-ature sum as a lookup tablemdashsimilar to figure 6(b)This interpolation approximates what the photosyn-thetic parameters 120572 and 120573 would have been in 2014if the vegetation had developed with temperature as inthe other years Subsequently GPP was calculated withthe observed radiation in 2014 following equation 1from the day that 300 D was reached (day of year159) up until the peak of summer (day of year 200)The period following the peak of summer is omittedto avoid an influence due to varying onsets of senes-cence (the whole time series is shown in figure S4) Amedian of these estimates showed that the vegetationcould have photosynthesized an additional 14 g C mminus2

in 2014 with an upper estimate of 24 g C mminus2 (whencompared to 2013) and a lower estimate of 0 g C mminus2

(when compared to 2010)mdashif there had been no neteffect from the damaged vegetation Since cumulativeGPP was 116 g C mminus2 during the same period in 2014this flux could have been sim12 higher with a lowerand upper estimate of 0 and 21

Comparison to remote sensing dataIn figure 7 a time series is plotted of the maximum andaverage NDVI value for each summer (day of year 175ndash225) from 2000ndash2017 which shows that 2014 had thelowest value in a decademdashup to that point The averagevalue for the summer of 2010 was nearly as low butwith a higher maximum The peak season was missedin 2013 due to bad coverage (figure S5) and NDVIvalues are probably underestimated for that year sinceGPP was high (figure 5) Average NDVI values in 2014are lower than in the other measurement years but notunprecedented in the long-term satellite record Thisis probably due to the excellent growing conditions inthe summer of 2014 which boosted vegetation growthafter mid-summer (figures 5 and S4)

However the maximum NDVI value reached in2014 was the second-lowest until then (after 2003) andit took much longer than normal to reach the max-imum (table S2 figures 7 and S5) On average peakNDVI values are reached on day of year 207plusmn 11 daysbut the maximum in 2014 was on day of year 222(August 10) The low NDVImdasha browning eventmdashandthe delayed peak were probably due to the large amountof damaged vegetation The only years with a later time-to-peak were 2007 (223) and 2017 (225) althoughconsiderable uncertainty exists on these dates due tocloud cover and their average values are much higher(figures 7 and S5)

Interestingly average NDVI values were at theirall-time lowest in 2015mdashthe year following the extremewinter event The browning event worsened indicatingno recovery of the ecosystem and this was possibly dueto another extreme winter (figure 3) Unfortunatelyflux measurements at the Saura peat bog had ceased by2015 and we do not know how this was reflected inthe ecosystem fluxes The same goes for the upwardsreturn of NDVI levels in 2016 However NDVI showsa reasonable agreement with GPPsat (Figure S5) andit is therefore likely that photosynthesis rates in 2015were lower than in 2014

7

Environ Res Lett 13 (2018) 065009

Figure 7 Maximum and average NDVI values for the summer (day of year 175ndash225) from 2000ndash2017 obtained from MODIS (v6)averaged over the four pixels closest to the position of the eddy covariance tower The years covered in this study are shown in colorThe values for 2013 are probably underestimated due to a coverage gap during peak summer A more detailed NDVI time series isshown in figure S5

Discussion

Impact of the 20132014 winter on summer CO2exchangeThis study shows that the severe frost drought eventof January and February 2014 unprecedented in theclimate record on Andoslashya led to the strong diebackof the shrub species Calluna vulgaris and Empetrumnigrum Cumulative GPP however was higher in 2014than in other summers This contradictory result canbe explained by the fact that 2014 also had the sunniestand warmest summer of the 5 years in this dataset(table 1) This provided ideal conditions for growth ofundamaged plants

However when interannual variability in radiationand temperature is compensated for it is clear thatvegetation productivity showed a delayed response fol-lowing snowmelt when compared to other years (figure6) This indicates a vulnerability of this ecosystem to theextreme winter event While briefly following a sim-ilar development for GPPsat as for the other yearsa clear departure occurred at two months followingsnowmelt at a point when shrub bud break normallywould occur Although a period of colder weather mayhave contributed to this delayed response this pat-tern remained present when GPPsat was comparedto accumulated degree days Vegetation develop-ment was lagging behind other years most likelydue to the large number of winter-damaged shrubs

After the initial anomaly in GPPsat the ecosystemshowed high photosynthesis rates later in the summer(figures 5 and S4) indicating some resilience to theextreme winter event A possible explanation for thismay be that the ecosystem partly recovered its car-bon uptake through compensatory growth (Bokhorstet al 2011) spurred on by the exceptionally warmand sunny weather of July and August 2014 Highertemperatures however also stimulated ecosystem res-piration with record high respiration in July and

August 2014 (figure 5) It is possible that part ofthese high respiration rates was related to decompos-ing dead plant material limiting NEE but a separationof ecosystem respiration into autotrophic and het-erotrophic respiration rates is not possible with thisdataset In future studies of the impact of extreme win-ter events such effects on respiration need to be takeninto account during field campaigns

Possibility of moisture limitationsIn addition to the documented damage to the shrubsother causes of the lower vegetation productivity atthe Saura bog need to be considered Droughts andheatwaves in particular can reduce the carbon uptakeof an ecosystem when plants close their stomata toconserve water (Lund et al 2012 van der Molen et al2011) This behavior is taken into account by the par-titioning method used in this study where GPP isreduced for high vapor pressure deficit values (Lass-lop et al 2010) Minor events did occur in early July2014 as in other years but at a very low level and foronly a few days which cannot explain the observeddifferences (figure S3) Low soil moisture could beanother limiting factor but 2014 was not exception-ally dry (table 1 figure S6) and significant reductionsin soil moisture occurred after the deviation in GPPrates Moreover a previous study of the Saura bog byLund et al (2015) showed that dry conditions had a lowimpact on the ability of this ecosystem to store carbonIt is therefore unlikely that summer drought condi-tions caused the divergent pattern of GPP as shownin figure 6

Extreme winter events and remotely sensed browningAlthough the Saura bog has experienced multiple win-ters with strong frost in the absence of snow as shownin figure 3 many of these did not lead to strongreductions in NDVI It is striking that the strong frostevent that occurred during polar night in the winter of

8

Environ Res Lett 13 (2018) 065009

20152016 did not negatively affect NDVI values Tothe contrary after two years of browning NDVI val-ues jumped back up to a normal value A possibleexplanation for this may lie in the timing of theseevents The extreme winter events in early 2014 and2015 occurred when the sunlight had returned afterone and a half month of darkness Under sunny con-ditions plants may attempt to transpire but cannotaccess the frozen soil water and they desiccate (Bjerkeet al 2017) The absence of sunlight during the frostevent in the winter of 20152016 probably preventedextensive plant damage due to frost drought Thisshows that the damage of an extreme winter eventvaries depending on its timing within the cold sea-son Other factors such as interannual variation in theamount of frost resistance that was built up may alsohave played a role

Resilienceandvulnerabilityof ecosystemfunctioningto wintertime impactsOur results show a considerable delayed response ofthe vegetation to temperature as shown in figure 6(a)but the estimated impact on GPP varies strongly from24 g C mminus2 to no effect at all when compared to 2010This appears to suggest that the impact of the frostevent on CO2 fluxes could have been negligible butthis is unlikely since weather conditions in the sum-mers of 2010 and 2014 were strongly dissimilar In2010 snowmelt occurred almost three and a half weekslater than in 2014 and values of 300 D werenrsquot reacheduntil June 18 compared to June 8 for 2014 (table S1)Besides this difference in the length of the growing sea-son there was also a stark contrast in the amount ofincoming radiation up to the peak of summer 432 MJin 2010 vs 763 MJ in 2014 The highly unfavorablegrowing conditions in 2010 are reflected in the GPPsatvalues which by mid-summer had not reached thesame maximum uptake as in the other years andaverage summer NDVI values were among the low-est recorded The similar vegetation development in2014 and 2010mdashone of the warmest and sunniest yearsvs the shortest coldest and cloudiest growing seasonin this datasetmdashis in fact a strong indication that theextreme winter event reduced GPPmdashcomparable insize to interannual variations in summer weather Fullyaccounting for large differences in weather remainschallenging which is why a large uncertainty remainson our estimate of the impact of the extreme winterevent on ecosystem carbon exchange

Besides these uncertainties the CO2 uptake of theecosystem may have been somewhat resilient to thefrost drought due to a contribution from vegetationtypes other than shrubs About 30 of the surfacearea of the Saura bog consists of hollows where shrubsare absent and Carex spp is abundantmdashwhich couldhave responded to the warm weather Moreover theSaura bog has a large abundance of lichens and mossesThese functional vegetation groups were not stronglyaffected by the frost drought event and the warm

and sunny weather may have boosted their photosyn-thesis rates In other words while the CO2 exchange ofthis bog was vulnerable at the species level (ie shrubs)to a certain degree it was resilient at the ecosystem levelThe 2014 frost drought event may have had a muchlarger impact on the net CO2 exchange at other affectedareas along the Norwegian coast in places where thefraction of shrubs vs mosses and sedges would havebeen highermdasheg in dry heathlands (Bokhorst et al2009 Bjerke et al 2014)

While the further decline in NDVI in 2015 showsthat the peatland did not recover in the following yearpossibly due to an additional extreme winter eventthe return to normal NDVI values in 2016 shows thatthis ecosystem can recover from an extreme winterevent in a relatively short time Such behavior hasbeen reported before for a browning event in north-ern Scandinavia caused by a winter warming event(Bokhorst et al 2012) If however extreme winterevents will increase in frequency eg every other yearsubsequent browning events may constitute a brown-ing trend In that case the species distribution of anecosystem may change with a lasting effect on CO2 andenergy exchange

Conclusions

The extreme winter event in January 2014 severelydamaged shrubs at the Saura bog and reduced bothvegetation CO2 uptake and NDVI in the followingsummer A comparison with the photosyntheticparameters of other years indicates that the ecosystemcould have taken up an additional 14 (0ndash24) g C mminus2

(sim12 of GPP) from day 159 to 200 if it had not beendamaged This means that the reduction in GPP causedby the winter event of 2014 was similar in size to inter-annual differences due to summer weather conditions(table 1)

Vegetation damage from extreme winter eventsshould be included in model simulations Current landsurface models project an increase in arctic vegeta-tion productivity following high latitude warming (Xiaet al 2017 Zhang et al 2014 Sitch et al 2007) despiterecent browning trends showing the opposite (Phoenixand Bjerke 2016) This suggests an overestimation ofGPP in areas prone to winter damage However speciesthat are more resilient to extreme winter events maycompensate the impact of extreme winter events onthe net CO2 exchange of ecosystems Observationsand modeling studies that focus on the impact ofextreme winter events on CO2 exchange thereforeshould not exclusively focus on vulnerable speciessuch as shrubs but determine the resilience of theecosystem as a whole

This study focused on one extreme winter eventin one particular year but when such events increasein frequency and vegetation is damaged more oftenthis may lead to shifts in ecosystem composition

9

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

Figure 6 7 day running mean of GPP at saturated light levels (700 Wmminus2) vs d after snow melt and the temperature sum followingsnow melt expressed in degree days (D) Time series shown are from snowmelt until day of year 200 (July 19 in non-leap years)

only to pick up at a later time Vegetation develop-ment took another five weeks up until the second halfof July Of all snow-free seasons only 2010 showed adegree-day response similar to that of 2014 Howevera simple comparison of these two years is problematicsince weather conditions in 2010 were vastly differ-ent from 2014 snowmelt occurred 35 weeks later andincoming radiation and temperature were much lower(table 1 figure S3)

A delayed response in 2014 similar to a cold andcloudy year is the kind of behavior that would beexpected when a high number of shrubs are damagedand their contribution to GPP is lowered (Bokhorst etal 2011) It appears therefore that the capacity of theecosystem to take up carbon was reduced during thesummer of 2014

Toquantify this reductionwe interpolated thepho-tosynthetic parameters 120572 and 120573 of the years 2010ndash2013obtained from the partitioning model (Lasslop et al2010) to specific dates in 2014 by using the temper-ature sum as a lookup tablemdashsimilar to figure 6(b)This interpolation approximates what the photosyn-thetic parameters 120572 and 120573 would have been in 2014if the vegetation had developed with temperature as inthe other years Subsequently GPP was calculated withthe observed radiation in 2014 following equation 1from the day that 300 D was reached (day of year159) up until the peak of summer (day of year 200)The period following the peak of summer is omittedto avoid an influence due to varying onsets of senes-cence (the whole time series is shown in figure S4) Amedian of these estimates showed that the vegetationcould have photosynthesized an additional 14 g C mminus2

in 2014 with an upper estimate of 24 g C mminus2 (whencompared to 2013) and a lower estimate of 0 g C mminus2

(when compared to 2010)mdashif there had been no neteffect from the damaged vegetation Since cumulativeGPP was 116 g C mminus2 during the same period in 2014this flux could have been sim12 higher with a lowerand upper estimate of 0 and 21

Comparison to remote sensing dataIn figure 7 a time series is plotted of the maximum andaverage NDVI value for each summer (day of year 175ndash225) from 2000ndash2017 which shows that 2014 had thelowest value in a decademdashup to that point The averagevalue for the summer of 2010 was nearly as low butwith a higher maximum The peak season was missedin 2013 due to bad coverage (figure S5) and NDVIvalues are probably underestimated for that year sinceGPP was high (figure 5) Average NDVI values in 2014are lower than in the other measurement years but notunprecedented in the long-term satellite record Thisis probably due to the excellent growing conditions inthe summer of 2014 which boosted vegetation growthafter mid-summer (figures 5 and S4)

However the maximum NDVI value reached in2014 was the second-lowest until then (after 2003) andit took much longer than normal to reach the max-imum (table S2 figures 7 and S5) On average peakNDVI values are reached on day of year 207plusmn 11 daysbut the maximum in 2014 was on day of year 222(August 10) The low NDVImdasha browning eventmdashandthe delayed peak were probably due to the large amountof damaged vegetation The only years with a later time-to-peak were 2007 (223) and 2017 (225) althoughconsiderable uncertainty exists on these dates due tocloud cover and their average values are much higher(figures 7 and S5)

Interestingly average NDVI values were at theirall-time lowest in 2015mdashthe year following the extremewinter event The browning event worsened indicatingno recovery of the ecosystem and this was possibly dueto another extreme winter (figure 3) Unfortunatelyflux measurements at the Saura peat bog had ceased by2015 and we do not know how this was reflected inthe ecosystem fluxes The same goes for the upwardsreturn of NDVI levels in 2016 However NDVI showsa reasonable agreement with GPPsat (Figure S5) andit is therefore likely that photosynthesis rates in 2015were lower than in 2014

7

Environ Res Lett 13 (2018) 065009

Figure 7 Maximum and average NDVI values for the summer (day of year 175ndash225) from 2000ndash2017 obtained from MODIS (v6)averaged over the four pixels closest to the position of the eddy covariance tower The years covered in this study are shown in colorThe values for 2013 are probably underestimated due to a coverage gap during peak summer A more detailed NDVI time series isshown in figure S5

Discussion

Impact of the 20132014 winter on summer CO2exchangeThis study shows that the severe frost drought eventof January and February 2014 unprecedented in theclimate record on Andoslashya led to the strong diebackof the shrub species Calluna vulgaris and Empetrumnigrum Cumulative GPP however was higher in 2014than in other summers This contradictory result canbe explained by the fact that 2014 also had the sunniestand warmest summer of the 5 years in this dataset(table 1) This provided ideal conditions for growth ofundamaged plants

However when interannual variability in radiationand temperature is compensated for it is clear thatvegetation productivity showed a delayed response fol-lowing snowmelt when compared to other years (figure6) This indicates a vulnerability of this ecosystem to theextreme winter event While briefly following a sim-ilar development for GPPsat as for the other yearsa clear departure occurred at two months followingsnowmelt at a point when shrub bud break normallywould occur Although a period of colder weather mayhave contributed to this delayed response this pat-tern remained present when GPPsat was comparedto accumulated degree days Vegetation develop-ment was lagging behind other years most likelydue to the large number of winter-damaged shrubs

After the initial anomaly in GPPsat the ecosystemshowed high photosynthesis rates later in the summer(figures 5 and S4) indicating some resilience to theextreme winter event A possible explanation for thismay be that the ecosystem partly recovered its car-bon uptake through compensatory growth (Bokhorstet al 2011) spurred on by the exceptionally warmand sunny weather of July and August 2014 Highertemperatures however also stimulated ecosystem res-piration with record high respiration in July and

August 2014 (figure 5) It is possible that part ofthese high respiration rates was related to decompos-ing dead plant material limiting NEE but a separationof ecosystem respiration into autotrophic and het-erotrophic respiration rates is not possible with thisdataset In future studies of the impact of extreme win-ter events such effects on respiration need to be takeninto account during field campaigns

Possibility of moisture limitationsIn addition to the documented damage to the shrubsother causes of the lower vegetation productivity atthe Saura bog need to be considered Droughts andheatwaves in particular can reduce the carbon uptakeof an ecosystem when plants close their stomata toconserve water (Lund et al 2012 van der Molen et al2011) This behavior is taken into account by the par-titioning method used in this study where GPP isreduced for high vapor pressure deficit values (Lass-lop et al 2010) Minor events did occur in early July2014 as in other years but at a very low level and foronly a few days which cannot explain the observeddifferences (figure S3) Low soil moisture could beanother limiting factor but 2014 was not exception-ally dry (table 1 figure S6) and significant reductionsin soil moisture occurred after the deviation in GPPrates Moreover a previous study of the Saura bog byLund et al (2015) showed that dry conditions had a lowimpact on the ability of this ecosystem to store carbonIt is therefore unlikely that summer drought condi-tions caused the divergent pattern of GPP as shownin figure 6

Extreme winter events and remotely sensed browningAlthough the Saura bog has experienced multiple win-ters with strong frost in the absence of snow as shownin figure 3 many of these did not lead to strongreductions in NDVI It is striking that the strong frostevent that occurred during polar night in the winter of

8

Environ Res Lett 13 (2018) 065009

20152016 did not negatively affect NDVI values Tothe contrary after two years of browning NDVI val-ues jumped back up to a normal value A possibleexplanation for this may lie in the timing of theseevents The extreme winter events in early 2014 and2015 occurred when the sunlight had returned afterone and a half month of darkness Under sunny con-ditions plants may attempt to transpire but cannotaccess the frozen soil water and they desiccate (Bjerkeet al 2017) The absence of sunlight during the frostevent in the winter of 20152016 probably preventedextensive plant damage due to frost drought Thisshows that the damage of an extreme winter eventvaries depending on its timing within the cold sea-son Other factors such as interannual variation in theamount of frost resistance that was built up may alsohave played a role

Resilienceandvulnerabilityof ecosystemfunctioningto wintertime impactsOur results show a considerable delayed response ofthe vegetation to temperature as shown in figure 6(a)but the estimated impact on GPP varies strongly from24 g C mminus2 to no effect at all when compared to 2010This appears to suggest that the impact of the frostevent on CO2 fluxes could have been negligible butthis is unlikely since weather conditions in the sum-mers of 2010 and 2014 were strongly dissimilar In2010 snowmelt occurred almost three and a half weekslater than in 2014 and values of 300 D werenrsquot reacheduntil June 18 compared to June 8 for 2014 (table S1)Besides this difference in the length of the growing sea-son there was also a stark contrast in the amount ofincoming radiation up to the peak of summer 432 MJin 2010 vs 763 MJ in 2014 The highly unfavorablegrowing conditions in 2010 are reflected in the GPPsatvalues which by mid-summer had not reached thesame maximum uptake as in the other years andaverage summer NDVI values were among the low-est recorded The similar vegetation development in2014 and 2010mdashone of the warmest and sunniest yearsvs the shortest coldest and cloudiest growing seasonin this datasetmdashis in fact a strong indication that theextreme winter event reduced GPPmdashcomparable insize to interannual variations in summer weather Fullyaccounting for large differences in weather remainschallenging which is why a large uncertainty remainson our estimate of the impact of the extreme winterevent on ecosystem carbon exchange

Besides these uncertainties the CO2 uptake of theecosystem may have been somewhat resilient to thefrost drought due to a contribution from vegetationtypes other than shrubs About 30 of the surfacearea of the Saura bog consists of hollows where shrubsare absent and Carex spp is abundantmdashwhich couldhave responded to the warm weather Moreover theSaura bog has a large abundance of lichens and mossesThese functional vegetation groups were not stronglyaffected by the frost drought event and the warm

and sunny weather may have boosted their photosyn-thesis rates In other words while the CO2 exchange ofthis bog was vulnerable at the species level (ie shrubs)to a certain degree it was resilient at the ecosystem levelThe 2014 frost drought event may have had a muchlarger impact on the net CO2 exchange at other affectedareas along the Norwegian coast in places where thefraction of shrubs vs mosses and sedges would havebeen highermdasheg in dry heathlands (Bokhorst et al2009 Bjerke et al 2014)

While the further decline in NDVI in 2015 showsthat the peatland did not recover in the following yearpossibly due to an additional extreme winter eventthe return to normal NDVI values in 2016 shows thatthis ecosystem can recover from an extreme winterevent in a relatively short time Such behavior hasbeen reported before for a browning event in north-ern Scandinavia caused by a winter warming event(Bokhorst et al 2012) If however extreme winterevents will increase in frequency eg every other yearsubsequent browning events may constitute a brown-ing trend In that case the species distribution of anecosystem may change with a lasting effect on CO2 andenergy exchange

Conclusions

The extreme winter event in January 2014 severelydamaged shrubs at the Saura bog and reduced bothvegetation CO2 uptake and NDVI in the followingsummer A comparison with the photosyntheticparameters of other years indicates that the ecosystemcould have taken up an additional 14 (0ndash24) g C mminus2

(sim12 of GPP) from day 159 to 200 if it had not beendamaged This means that the reduction in GPP causedby the winter event of 2014 was similar in size to inter-annual differences due to summer weather conditions(table 1)

Vegetation damage from extreme winter eventsshould be included in model simulations Current landsurface models project an increase in arctic vegeta-tion productivity following high latitude warming (Xiaet al 2017 Zhang et al 2014 Sitch et al 2007) despiterecent browning trends showing the opposite (Phoenixand Bjerke 2016) This suggests an overestimation ofGPP in areas prone to winter damage However speciesthat are more resilient to extreme winter events maycompensate the impact of extreme winter events onthe net CO2 exchange of ecosystems Observationsand modeling studies that focus on the impact ofextreme winter events on CO2 exchange thereforeshould not exclusively focus on vulnerable speciessuch as shrubs but determine the resilience of theecosystem as a whole

This study focused on one extreme winter eventin one particular year but when such events increasein frequency and vegetation is damaged more oftenthis may lead to shifts in ecosystem composition

9

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

Figure 7 Maximum and average NDVI values for the summer (day of year 175ndash225) from 2000ndash2017 obtained from MODIS (v6)averaged over the four pixels closest to the position of the eddy covariance tower The years covered in this study are shown in colorThe values for 2013 are probably underestimated due to a coverage gap during peak summer A more detailed NDVI time series isshown in figure S5

Discussion

Impact of the 20132014 winter on summer CO2exchangeThis study shows that the severe frost drought eventof January and February 2014 unprecedented in theclimate record on Andoslashya led to the strong diebackof the shrub species Calluna vulgaris and Empetrumnigrum Cumulative GPP however was higher in 2014than in other summers This contradictory result canbe explained by the fact that 2014 also had the sunniestand warmest summer of the 5 years in this dataset(table 1) This provided ideal conditions for growth ofundamaged plants

However when interannual variability in radiationand temperature is compensated for it is clear thatvegetation productivity showed a delayed response fol-lowing snowmelt when compared to other years (figure6) This indicates a vulnerability of this ecosystem to theextreme winter event While briefly following a sim-ilar development for GPPsat as for the other yearsa clear departure occurred at two months followingsnowmelt at a point when shrub bud break normallywould occur Although a period of colder weather mayhave contributed to this delayed response this pat-tern remained present when GPPsat was comparedto accumulated degree days Vegetation develop-ment was lagging behind other years most likelydue to the large number of winter-damaged shrubs

After the initial anomaly in GPPsat the ecosystemshowed high photosynthesis rates later in the summer(figures 5 and S4) indicating some resilience to theextreme winter event A possible explanation for thismay be that the ecosystem partly recovered its car-bon uptake through compensatory growth (Bokhorstet al 2011) spurred on by the exceptionally warmand sunny weather of July and August 2014 Highertemperatures however also stimulated ecosystem res-piration with record high respiration in July and

August 2014 (figure 5) It is possible that part ofthese high respiration rates was related to decompos-ing dead plant material limiting NEE but a separationof ecosystem respiration into autotrophic and het-erotrophic respiration rates is not possible with thisdataset In future studies of the impact of extreme win-ter events such effects on respiration need to be takeninto account during field campaigns

Possibility of moisture limitationsIn addition to the documented damage to the shrubsother causes of the lower vegetation productivity atthe Saura bog need to be considered Droughts andheatwaves in particular can reduce the carbon uptakeof an ecosystem when plants close their stomata toconserve water (Lund et al 2012 van der Molen et al2011) This behavior is taken into account by the par-titioning method used in this study where GPP isreduced for high vapor pressure deficit values (Lass-lop et al 2010) Minor events did occur in early July2014 as in other years but at a very low level and foronly a few days which cannot explain the observeddifferences (figure S3) Low soil moisture could beanother limiting factor but 2014 was not exception-ally dry (table 1 figure S6) and significant reductionsin soil moisture occurred after the deviation in GPPrates Moreover a previous study of the Saura bog byLund et al (2015) showed that dry conditions had a lowimpact on the ability of this ecosystem to store carbonIt is therefore unlikely that summer drought condi-tions caused the divergent pattern of GPP as shownin figure 6

Extreme winter events and remotely sensed browningAlthough the Saura bog has experienced multiple win-ters with strong frost in the absence of snow as shownin figure 3 many of these did not lead to strongreductions in NDVI It is striking that the strong frostevent that occurred during polar night in the winter of

8

Environ Res Lett 13 (2018) 065009

20152016 did not negatively affect NDVI values Tothe contrary after two years of browning NDVI val-ues jumped back up to a normal value A possibleexplanation for this may lie in the timing of theseevents The extreme winter events in early 2014 and2015 occurred when the sunlight had returned afterone and a half month of darkness Under sunny con-ditions plants may attempt to transpire but cannotaccess the frozen soil water and they desiccate (Bjerkeet al 2017) The absence of sunlight during the frostevent in the winter of 20152016 probably preventedextensive plant damage due to frost drought Thisshows that the damage of an extreme winter eventvaries depending on its timing within the cold sea-son Other factors such as interannual variation in theamount of frost resistance that was built up may alsohave played a role

Resilienceandvulnerabilityof ecosystemfunctioningto wintertime impactsOur results show a considerable delayed response ofthe vegetation to temperature as shown in figure 6(a)but the estimated impact on GPP varies strongly from24 g C mminus2 to no effect at all when compared to 2010This appears to suggest that the impact of the frostevent on CO2 fluxes could have been negligible butthis is unlikely since weather conditions in the sum-mers of 2010 and 2014 were strongly dissimilar In2010 snowmelt occurred almost three and a half weekslater than in 2014 and values of 300 D werenrsquot reacheduntil June 18 compared to June 8 for 2014 (table S1)Besides this difference in the length of the growing sea-son there was also a stark contrast in the amount ofincoming radiation up to the peak of summer 432 MJin 2010 vs 763 MJ in 2014 The highly unfavorablegrowing conditions in 2010 are reflected in the GPPsatvalues which by mid-summer had not reached thesame maximum uptake as in the other years andaverage summer NDVI values were among the low-est recorded The similar vegetation development in2014 and 2010mdashone of the warmest and sunniest yearsvs the shortest coldest and cloudiest growing seasonin this datasetmdashis in fact a strong indication that theextreme winter event reduced GPPmdashcomparable insize to interannual variations in summer weather Fullyaccounting for large differences in weather remainschallenging which is why a large uncertainty remainson our estimate of the impact of the extreme winterevent on ecosystem carbon exchange

Besides these uncertainties the CO2 uptake of theecosystem may have been somewhat resilient to thefrost drought due to a contribution from vegetationtypes other than shrubs About 30 of the surfacearea of the Saura bog consists of hollows where shrubsare absent and Carex spp is abundantmdashwhich couldhave responded to the warm weather Moreover theSaura bog has a large abundance of lichens and mossesThese functional vegetation groups were not stronglyaffected by the frost drought event and the warm

and sunny weather may have boosted their photosyn-thesis rates In other words while the CO2 exchange ofthis bog was vulnerable at the species level (ie shrubs)to a certain degree it was resilient at the ecosystem levelThe 2014 frost drought event may have had a muchlarger impact on the net CO2 exchange at other affectedareas along the Norwegian coast in places where thefraction of shrubs vs mosses and sedges would havebeen highermdasheg in dry heathlands (Bokhorst et al2009 Bjerke et al 2014)

While the further decline in NDVI in 2015 showsthat the peatland did not recover in the following yearpossibly due to an additional extreme winter eventthe return to normal NDVI values in 2016 shows thatthis ecosystem can recover from an extreme winterevent in a relatively short time Such behavior hasbeen reported before for a browning event in north-ern Scandinavia caused by a winter warming event(Bokhorst et al 2012) If however extreme winterevents will increase in frequency eg every other yearsubsequent browning events may constitute a brown-ing trend In that case the species distribution of anecosystem may change with a lasting effect on CO2 andenergy exchange

Conclusions

The extreme winter event in January 2014 severelydamaged shrubs at the Saura bog and reduced bothvegetation CO2 uptake and NDVI in the followingsummer A comparison with the photosyntheticparameters of other years indicates that the ecosystemcould have taken up an additional 14 (0ndash24) g C mminus2

(sim12 of GPP) from day 159 to 200 if it had not beendamaged This means that the reduction in GPP causedby the winter event of 2014 was similar in size to inter-annual differences due to summer weather conditions(table 1)

Vegetation damage from extreme winter eventsshould be included in model simulations Current landsurface models project an increase in arctic vegeta-tion productivity following high latitude warming (Xiaet al 2017 Zhang et al 2014 Sitch et al 2007) despiterecent browning trends showing the opposite (Phoenixand Bjerke 2016) This suggests an overestimation ofGPP in areas prone to winter damage However speciesthat are more resilient to extreme winter events maycompensate the impact of extreme winter events onthe net CO2 exchange of ecosystems Observationsand modeling studies that focus on the impact ofextreme winter events on CO2 exchange thereforeshould not exclusively focus on vulnerable speciessuch as shrubs but determine the resilience of theecosystem as a whole

This study focused on one extreme winter eventin one particular year but when such events increasein frequency and vegetation is damaged more oftenthis may lead to shifts in ecosystem composition

9

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

20152016 did not negatively affect NDVI values Tothe contrary after two years of browning NDVI val-ues jumped back up to a normal value A possibleexplanation for this may lie in the timing of theseevents The extreme winter events in early 2014 and2015 occurred when the sunlight had returned afterone and a half month of darkness Under sunny con-ditions plants may attempt to transpire but cannotaccess the frozen soil water and they desiccate (Bjerkeet al 2017) The absence of sunlight during the frostevent in the winter of 20152016 probably preventedextensive plant damage due to frost drought Thisshows that the damage of an extreme winter eventvaries depending on its timing within the cold sea-son Other factors such as interannual variation in theamount of frost resistance that was built up may alsohave played a role

Resilienceandvulnerabilityof ecosystemfunctioningto wintertime impactsOur results show a considerable delayed response ofthe vegetation to temperature as shown in figure 6(a)but the estimated impact on GPP varies strongly from24 g C mminus2 to no effect at all when compared to 2010This appears to suggest that the impact of the frostevent on CO2 fluxes could have been negligible butthis is unlikely since weather conditions in the sum-mers of 2010 and 2014 were strongly dissimilar In2010 snowmelt occurred almost three and a half weekslater than in 2014 and values of 300 D werenrsquot reacheduntil June 18 compared to June 8 for 2014 (table S1)Besides this difference in the length of the growing sea-son there was also a stark contrast in the amount ofincoming radiation up to the peak of summer 432 MJin 2010 vs 763 MJ in 2014 The highly unfavorablegrowing conditions in 2010 are reflected in the GPPsatvalues which by mid-summer had not reached thesame maximum uptake as in the other years andaverage summer NDVI values were among the low-est recorded The similar vegetation development in2014 and 2010mdashone of the warmest and sunniest yearsvs the shortest coldest and cloudiest growing seasonin this datasetmdashis in fact a strong indication that theextreme winter event reduced GPPmdashcomparable insize to interannual variations in summer weather Fullyaccounting for large differences in weather remainschallenging which is why a large uncertainty remainson our estimate of the impact of the extreme winterevent on ecosystem carbon exchange

Besides these uncertainties the CO2 uptake of theecosystem may have been somewhat resilient to thefrost drought due to a contribution from vegetationtypes other than shrubs About 30 of the surfacearea of the Saura bog consists of hollows where shrubsare absent and Carex spp is abundantmdashwhich couldhave responded to the warm weather Moreover theSaura bog has a large abundance of lichens and mossesThese functional vegetation groups were not stronglyaffected by the frost drought event and the warm

and sunny weather may have boosted their photosyn-thesis rates In other words while the CO2 exchange ofthis bog was vulnerable at the species level (ie shrubs)to a certain degree it was resilient at the ecosystem levelThe 2014 frost drought event may have had a muchlarger impact on the net CO2 exchange at other affectedareas along the Norwegian coast in places where thefraction of shrubs vs mosses and sedges would havebeen highermdasheg in dry heathlands (Bokhorst et al2009 Bjerke et al 2014)

While the further decline in NDVI in 2015 showsthat the peatland did not recover in the following yearpossibly due to an additional extreme winter eventthe return to normal NDVI values in 2016 shows thatthis ecosystem can recover from an extreme winterevent in a relatively short time Such behavior hasbeen reported before for a browning event in north-ern Scandinavia caused by a winter warming event(Bokhorst et al 2012) If however extreme winterevents will increase in frequency eg every other yearsubsequent browning events may constitute a brown-ing trend In that case the species distribution of anecosystem may change with a lasting effect on CO2 andenergy exchange

Conclusions

The extreme winter event in January 2014 severelydamaged shrubs at the Saura bog and reduced bothvegetation CO2 uptake and NDVI in the followingsummer A comparison with the photosyntheticparameters of other years indicates that the ecosystemcould have taken up an additional 14 (0ndash24) g C mminus2

(sim12 of GPP) from day 159 to 200 if it had not beendamaged This means that the reduction in GPP causedby the winter event of 2014 was similar in size to inter-annual differences due to summer weather conditions(table 1)

Vegetation damage from extreme winter eventsshould be included in model simulations Current landsurface models project an increase in arctic vegeta-tion productivity following high latitude warming (Xiaet al 2017 Zhang et al 2014 Sitch et al 2007) despiterecent browning trends showing the opposite (Phoenixand Bjerke 2016) This suggests an overestimation ofGPP in areas prone to winter damage However speciesthat are more resilient to extreme winter events maycompensate the impact of extreme winter events onthe net CO2 exchange of ecosystems Observationsand modeling studies that focus on the impact ofextreme winter events on CO2 exchange thereforeshould not exclusively focus on vulnerable speciessuch as shrubs but determine the resilience of theecosystem as a whole

This study focused on one extreme winter eventin one particular year but when such events increasein frequency and vegetation is damaged more oftenthis may lead to shifts in ecosystem composition

9

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

Vulnerable species such as shrubs may decline infavor of more resilient mosses and sedges which altersthe net carbon uptake and albedo The likelihood ofwhich remains unknown Continued monitoring of theCO2 exchange of ecosystems subject to extreme winterevents and the improved modellingof their response tothese instances is essential to project how the carbonexchange of high latitude ecosystems and associatedclimate-feedbacks will respond to further arctic winterwarming

Acknowledgments

This research has been made possible through fundingfrom Stiftelsen Fondet for Jord- og Myrundersoslashkelser(Foundation Fund for Soil- and Peat Research) theResearchCouncil ofNorwayNIBIOStrategicResearchFunds Norwegian Institute for Nature Research Nor-wegian Institute for Air Research J W B and HT received funding from the Polish-Norwegian Pro-gramme of the EEA Norway Grants (project 198571)and by FRAMndashHigh North Research Centre for Cli-mate and the Environment through its terrestrialflagship program (project 362222) Instrumentation atthe site as well as installation support was suppliedby the Smithsonian Environmental Research CenterLogistic and technical support from the Andoslashya RocketRange is gratefully acknowledged

ORCID iDs

Frans-Jan W Parmentier httpsorcidorg0000-0003-2952-7706Magnus Lund httpsorcidorg0000-0003-1622-2305

References

Bhatt U S Walker D Raynolds M Bieniek P Epstein H Comiso JPinzon J Tucker C and Polyakov I 2013 Recent declines inwarming and vegetation greening trends over Pan-ArcticTundra Remote Sens 5 4229ndash54

Bhatt U S et al 2014 Implications of Arctic Sea Ice Decline for theEarth System Annu Rev Env Resour 39 57ndash89

Bjerke J W Karlsen S R Hoslashgda K A Malnes E Jepsen J ULovibond S Vikhamar-Schuler D and Toslashmmervik H 2014Record-low primary productivity and high plant damage inthe Nordic Arctic Region in 2012 caused by multiple weatherevents and pest outbreaks Environ Res Lett 9 084006

Bjerke J W Treharne R Vikhamar-Schuler D Karlsen S RRavolainen V Bokhorst S Phoenix G K Bochenek Z andToslashmmervik H 2017 Understanding the drivers of extensiveplant damage in boreal and Arctic ecosystems Insights fromfield surveys in the aftermath of damage Sci Tot Environ 5991965ndash76

Bjerke J W Toslashmmervik H Zielke M and Joslashrgensen M 2015Impacts of snow season on ground-ice accumulation soil frostand primary productivity in a grassland of sub-Arctic NorwayEnviron Res Lett 10 095007

Bokhorst S F Bjerke J W Toslashmmervik H Callaghan T V andPhoenix G K 2009 Winter warming events damage sub-Arctic

vegetation consistent evidence from an experimentalmanipulation and a natural event J Ecol 97 1408ndash15

Bokhorst S Bjerke J W Street L E Callaghan T V and Phoenix G K2011 Impacts of multiple extreme winter warming events onsub-Arctic heathland phenology reproduction growth andCO2 flux responses Glob Change Biol 17 2817ndash30

Bokhorst S Toslashmmervik H Callaghan T V Phoenix G K and BjerkeJ W 2012 Vegetation recovery following extreme winterwarming events in the Sub-Arctic estimated using NDVI fromremote sensing and handheld passive proximal sensorsEnviron Exp Bot 81 18ndash25

Cohen J Pulliainen J Menard C B Johansen B Oksanen L LuojusK and Ikonen J 2013 Effect of reindeer grazing on snowmeltalbedo and energy balance based on satellite data analysesRemote Sens Environ 135 107ndash17

Elmendorf S C et al 2012 Plot-scale evidence of tundra vegetationchange and links to recent summer warming Nat ClimChange 2 453ndash7

Graham R M Cohen L Petty A A Boisvert L N Rinke A Hudson SR Nicolaus M and Granskog M A 2017 Increasing frequencyand duration of Arctic winter warming events Geophys ResLett 48 225

Hancock M H 2008 An exceptional Calluna vulgaris winterdie-back event Abernethy Forest Scottish Highlands PlantEcol Diver 1 89ndash103

Joslashrgensen M Oslashstrem L and Hoglind M 2010 De-hardening incontrasting cultivars of timothy and perennial ryegrass duringwinter and spring Grass Forage Sci 65 38ndash48

Lara M J Nitze I Grosse G Martin P and McGuire A D 2018Reduced arctic tundra productivity linked with landform andclimate change interactions Sci Report 8 2345

Lasslop G Reichstein M Papale D Richardson A D Arneth A BarrA G Stoy P and Wohlfahrt G 2010 Separation of netecosystem exchange into assimilation and respiration using alight response curve approach critical issues and globalevaluation Glob Change Biol 16 187ndash208

Lund M et al 2015 Low impact of dry conditions on the CO2exchange of a Northern-Norwegian blanket bog Environ ResLett 10 025004

Lund M Christensen T R Lindroth A and Schubert P 2012 Effectsof drought conditions on the carbon dioxide dynamics in atemperate peatland Environ Res Lett 7 045704

Meisingset E L Austrheim G Solberg E Brekkum Oslash and Lande U S2015 Effekter av klimastress pa hjortens vinterbeiter Utviklingav blabaeligrlyngen etter toslashrkevinteren 2014 Nibio Rapport 1 28

Milner J M Varpe Oslash van der Wal R and Hansen B B 2016Experimental icing affects growth mortality and flowering ina high Arctic dwarf shrub Ecol Evol 6 2139ndash48

Myers-Smith I H et al 2011 Shrub expansion in tundra ecosystemsdynamics impacts and research priorities Environ Res Lett 6045509

ORNL DAAC 2017 MODIS Collection 6 Land Products GlobalSubsetting and Visualization Tool (Oak Ridge TN ORNLDAAC)

Parmentier F J W van der Molen M K van Huissteden J KarsanaevS A Kononov A V Suzdalov D A Maximov T C and DolmanA J 2011 Longer growing seasons do not increase net carbonuptake in the northeastern Siberian tundra J Geophys ResBiogeosci 116 G04013

Phoenix G K and Bjerke J W 2016 Arctic browning extreme eventsand trends reversing arctic greening Glob Change Biol 222960ndash2

Preece C Callaghan T V and Phoenix G K 2012 Impacts of wintericing events on the growth phenology and physiology ofsub-arctic dwarf shrubs Physiol Plantarum 146 460ndash72

Saloranta T M 2012 Simulating snow maps for Norwaydescription and statistical evaluation of the seNorge snowmodel Cryosphere 6 1323ndash37

Sitch S McGuire A D Kimball J S Gedney N Gamon J EngstromR Wolf A Zhuang Q Clein J and Mcdonald K C 2007Assessing the carbon balance of circumpolar Arctic tundrausing remote sensing and process modeling Ecol Appl 17213ndash34

10

Environ Res Lett 13 (2018) 065009

Timmermann V Andreassen K Clarke N Solheim H and Aas W2015 Skogens helsetilstand I Norge Resultater fraskogskadeovervakingen I 2014 Nibio Rapport 1 56

van der Molen M K et al 2011 Drought and ecosystem carboncycling Agric Forest Meteorol 151 765ndash73

Vikhamar-Schuler D Isaksen K Haugen J E Toslashmmervik H Luks BSchuler T V and Bjerke J W 2016 Changes in winter warmingevents in the nordic Arctic Region J Clim 29 6223ndash44

Vorren K-D Blaauw M Wastegard S van der Plicht J and Jensen C2007 High-resolution stratigraphy of the northernmostconcentric raised bog in Europe Sellevollmyra Andoslashyanorthern Norway Boreas 36 253ndash77

Xia J et al 2017 Terrestrial ecosystem model performance insimulating productivity and its vulnerability to climate change

in the northern permafrost region J Geophys Res Biogeosci122 430ndash46

Zhang W Jansson C Miller P A Smith B and Samuelsson P 2014Biogeophysical feedbacks enhance the Arctic terrestrial carbonsink in regional Earth system dynamics Biogeosciences 115503ndash19

Zhao J Peichl M and Nilsson M B 2016 Enhanced winter soil frostreduces methane emission during the subsequent growingseason in a boreal peatland Glob Change Biol 22750ndash62

Zhao J Peichl M and Nilsson M B 2017 Long-term enhancedwinter soil frost alters growing season CO2 fluxes through itsimpact on vegetation development in a boreal peatland GlobChange Biol 23 3139ndash53

11