Interactive comment on “ Satellites reveal an increase in gross primary production in a greenlandic high arctic fen 1992 – 2008 ” by T

GENERAL COMMENTS: The main problem is that in my opinion the presented approach and data is not sufficient to derive a robust relationship between the satellite information and the ground-based GPP. The authors do not write very clearly about the experimental design, i.e. the number of closed chamber measurements of NEE, Reco and GPP over the day performed on the 11 measurement days. However, according to the df in Table 2, it appears that every plot was measured only once per day. I doubt that it is feasible to calculate a robust mean by averaging only 11 measurements per-


Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion

Abstract
Arctic wetlands play a key role in the terrestrial carbon cycle.Recent studies have shown a greening trend and indicated an increase in CO 2 uptake in boreal and sub-to low-arctic areas.Our aim was to combine satellite-based normalized difference vegetation index (NDVI) with ground-based flux measurements of CO 2 to investigate a possible greening trend and potential changes in gross primary production (GPP) between 1992 and 2008 in a high arctic fen area.The study took place in Rylekaerene in the Zackenberg Research Area (74 • 28 N 20 • 34 W), located in the National park of North Eastern Greenland.We estimated the light use efficiency (ε) for the dominant vegetation types from field measured fractions of photosynthetic active radiation (FA-PAR) and ground-based flux measurements of GPP.Measured FAPAR were correlated to satellite-based NDVI.The FAPAR-NDVI relationship in combination with ε was applied to satellite data to model GPP 1992GPP -2008.The model was evaluated against field measured GPP.The model was a useful tool for up-scaling GPP and all basic requirements for the model were well met, e.g., FAPAR was well correlated to NDVI and modeled GPP was well correlated to field measurements.The studied high arctic fen area has experienced a strong increase in GPP between 1992 and 2008.The area has during this period also experienced a substantial increase in local air temperature.Consequently, the observed greening trend is most likely due to ongoing climatic change possibly in combination with CO 2 fertilization, due to increasing atmospheric concentrations of CO 2 .

Introduction
High latitude ecosystems are identified to be susceptible to climatic change through several processes tightly connected to both the regional and global climate systems (Bonan et al., 1995).Arctic wetlands play a key role in controlling the terrestrial carbon cycle, since the prevailing waterlogged, anoxic and cool conditions effectively reduce Figures

Back Close Full Screen / Esc
Printer-friendly Version Interactive Discussion decomposition rates, which favors the formation of peat.Although arctic tundra covers only about 5 percent of the global land area, it holds approximately 12-14 percent of the worlds' total pool of soil organic carbon, and concern have been expressed that the expected changes in climate could cause a decrease in today's sinks and even turn sinks into sources (Post et al., 1982;Oechel et al., 1993).Peat accumulation is primarily governed by the balance between uptake by gross primary production (GPP) and release through decomposition.Changes in the sink strength for high arctic ecosystems are therefore highly affected by responses of these processes to climate variations.
Several studies investigating satellites and the normalized difference vegetation index (NDVI) have shown that there is a greening trend in northern ecosystems, indicating an increase in CO 2 uptake (Slayback et al., 2003;Jia et al., 2003;Zhou et al., 2001;Stow et al., 2007;Thompson et al., 2006;Verbyla, 2008;Myneni et al., 1997a).These studies have mainly focused on boreal, sub-arctic and arctic areas and very few studies exist from the high arctic.In the high arctic, temperatures are colder and the growing season shorter than in lower arctic regions.Consequently high arctic ecosystems normally experience higher temperature constraints which presumably make them more sensitive to rising temperatures (Nadelhofer et al., 1997).
A widely applied concept within remote sensing is the approach to estimate productivity by a light use efficiency (LUE) model (Monteith, 1972(Monteith, , 1977) ) allowing GPP to be estimated from absorbed photosynthetic active radiation according to Eq. (1): where ε is the light use efficiency of the vegetation, PAR is incoming photosynthetic active radiation, and FAPAR is the fraction of absorbed photosynthetic active radiation.The ε value was initially considered to be relatively constant, but substantial differences have been found between vegetation types, ages, species composition, and stress level (Goetz and Prince, 1996;Gower et al., 1999) and it is therefore an important parameter to estimate for various vegetation types when evaluating GPP in an area.FAPAR (Asrar et al., 1984;Goward and Huemmrich, 1992;Myneni and Williams, 1994;Myneni et al., 1997b;Fensholt et al., 2004;Olofsson and Eklundh, 2007).However, these studies have focused on crop stands, prairies, semi-arid areas and deciduous and coniferous forests, and an issue of concern is the lack of data when it comes to arctic regions.Yet, as far as we know, it is a relationship that has never been studied for high arctic ecosystems.
In this paper, our aim was to investigate if there as a result of ongoing climate change has been a change in GPP from 1992 to 2008 in a high arctic fen area in North Eastern Greenland.To estimate changes in GPP, we parameterized the light use efficiency model for the peak of the growing season for the dominant high arctic tundra vegetation types.We also correlated FAPAR from the peak of the growing season measured in situ to satellite-based NDVI.The FAPAR-NDVI relationship was combined with the estimated light use efficiency coefficient and applied to satellite data to model GPP for the area, 1992-2008.Finally, the model was evaluated against site-specific measurements of GPP.

Site description
The study took place in Rylekaerene in the Zackenberg Research Area (74 • 28 N 20 • 34 W), located in the National park of North Eastern Greenland.The area is located in the high arctic zone with a local climate in the Zackenberg valley that deviates slightly from the definition of the high arctic climate.Average temperature of the warmest month is 5.8 • C, and mean annual temperature is −9 • C (Hansen et al., 2008).The Zackenberg valley is underlain by continuous permafrost and the active layer depth ranges between 0.5-1.0 m.The dominant wind direction is N to NNW, except during summer when the prevailing winds are S to SE.Average wind speeds during summer are less than 4 ms −1 (Hansen et al., 2008).1.

Ground-based estimates of FAPAR and its relationship to NDVI
The ASTER image was further used to select two sites within Rylekaerene that enabled FAPAR measurements in a wide range of NDVI values.NDVI was estimated according to Eq. (2); Where ρ NIR is the near infrared (NIR) band and ρ red is the red band.( of ground cover and leaf area index, still FAPAR will be unique (Myneni and Williams, 1994).This validates the use of NDVI in heterogeneous areas with different vegetation types and its use with different satellite sensors data.A total of 13 sampling points were selected in the study sites covering the vegetation types (three continuous fen, two hummocky fen, four grassland, and one each for Cassiope heath, Dryas heath, Vaccinium heath and Salix snowbed).To estimate FAPAR, two measured variables are needed; the fraction of intercepted photosynthetic active radiation (FIPAR), and the fraction of ground covered by photosynthetic active vegetation (Hall et al., 1992).FIPAR was estimated by measuring incoming and reflected PAR every thirtieth second and averaged over ten minutes between 23 June and 5 August 2007 with hemispherical JYP-1000 sensors connected to Minicube loggers.The sensors were placed pointing downwards to measure reflected PAR as close to the corresponding centre of the satellite pixel as possible.The sensors were placed at a height of between 27 and 49 cm above the ground, depending on soil conditions.Less compact soil forced the stand to be pushed further into the ground to prevent it from falling.Incoming PAR was measured by sensors pointing upwards at two locations, and an average value between these was used.To calculate FIPAR, reflected PAR subtracted from incoming PAR was divided by incoming PAR.Data acquired between 18:00 and 9:00 were removed to decrease the impact of low illumination angles.
To estimate the fraction of ground covered by photosynthetic active vegetation, photos were taken of the vegetation underneath the sensors on the 2 and 4 August.The fraction of ground covered by photosynthetic active vegetation on these images was estimated visually using a 0.25 m 2 square placed on the ground.This fraction was multiplied by average FIPAR estimates from 25 July to 5 August to estimate FAPAR for this period of the growing season, from now on referred to as FAPAR max , since this represents the peak of the growing season.Muskoxen had occasionally overturned sensors and data from these periods were removed.NDVI was calculated according to Eq. ( 2) with a Landsat Enhanced Thematic Mapper (ETM+) image from 29 July 2007.All satellite imagery was converted to reflectance Introduction

Conclusions References
Tables Figures

Back Close
Full and atmospheric and terrain corrections were performed with ATCOR 6.1.It has been shown that FAPAR is linearly correlated to NDVI (Asrar et al., 1984;Goward and Huemmrich, 1992;Myneni and Williams, 1994), and a linear regression between NDVI and the ground-based estimates of FAPAR max was fitted.The normality assumption for residuals was checked using a Kolmogorov-Smirnov test.All statistical analyses throughout the analysis were carried out with SPSS 17.0 for Windows.Results of the statistics were regarded as significant if p-values were lower than 0.05.

Ecosystem CO 2 flux measurements
A study site in the centre of the fen areas was chosen to ensure that the main vegetation types were represented within a reasonably small area (Fig. 1).In total, 55 measurement plots were randomly chosen within the different vegetation types; 15 plots for continuous fen, 10 plots each for hummocky fen and grassland, and 5 plots each for Cassiope heath, Dryas heath, Vaccinium heath and Salix snowbed.In these plots CO 2 fluxes were measured using the closed chamber technique with an infrared gas analyser (EGM-4, PP-systems, Hitchin, Hertfordshire UK).Measurements were carried out 11 times between 30 June and 4 August 2007.Measurements were distributed over the day so that each individual plot was measured at different times of day (between 10 a.m. and 6 p.m.) over the measurement period.The chamber was 101 dm 3 and transparent and the intake and outlet of gas was located at one of the sides 0.15 m and 0.25 m above ground, respectively.To ensure proper mixing of the air and representative sampling from the entire chamber two small fans were located at opposite sides in the upper part of the chamber.A 10 cm collar permanently installed at the base of the chamber was used to minimize air flow under the edges of the chamber.
Immediately before the start of each measurement, the chamber was carefully placed on the ground to avoid disturbance.The change in concentration of CO 2 was recorded continuously for 3 min.Net ecosystem exchange (NEE) was measured with the transparent chamber, and ecosystem dark respiration (ER) was thereafter measured by covering the chamber with a lightproof hood.Soil temperature at 10 cm 1107 Introduction

Conclusions References
Tables Figures

Back Close
Full depth, water table depth (for fen vegetation types) and active layer depth were measured in close proximity to each plot during the chamber measurements.Incoming and reflected PAR was measured every 1.3 s within the chamber with hemispherical JYP-1000 sensors and average values were calculated for the same periods as the CO 2 flux measurements.
The CO 2 fluxes were calculated by fitting a line to the concentration change in the chamber over the measurement period.The slope of the line was then used in Eq. ( 3) for calculating the actual flux.
where F c is the CO 2 flux, dC/dt is slope of the concentration line, V ch is volume of the chamber, P atm is air pressure, M is molar weight of CO 2 , R is the universal gas constant, T air is air temperature, and A ch is the area of ground covered by the chamber.GPP was calculated by subtracting ER from NEE, giving negative values to GPP in this first calculation step.However, since this study focus on GPP and changes in production over time we preferred the view where positive values denote a gain of CO 2 to the ecosystem and effluxes were therefore converted to define GPP as positive and ER as negative.Outliers caused by disturbances were identified using boxplot graphs and removed from the rest of the analysis.

Estimation of light use efficiency
Three of the 11 ground-based chamber measurements of CO 2 , conducted closest in time to the Landsat image acquired 29 July 2007, were chosen for detailed comparisons between GPP and FAPAR max .Incoming and reflected PAR measured inside the chamber were used for estimating FIPAR max , which was multiplied with the area covered by photosynthetic active vegetation within the chamber to calculate chamber-based FAPAR max, from now on referred to as chamber-based to separate it from FIPAR max and FAPAR max measured at the 13 plots with different NDVI (see Introduction

Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion above).There were occasional problems with the sensor calibration, and chamberbased FIPAR max values below 0.8 were considered unrealistic and removed from the analysis.The light use efficiency, ε, was estimated by dividing GPP with incoming PAR multiplied by chamber-based FAPAR max .Average ε max was estimated for all vegetation types.It should be noted that both FAPAR and ε varies not only between vegetation types but also over the growing season; chamber-based FAPAR max and ε max was therefore used in the following analyses since they represent the peak of the growing season.A one-way ANOVA assuming Tukey's equal variances was performed to test for significant differences between the chamber-based FAPAR max and ε max of the different vegetation types.

The spatial and temporal extrapolation of GPP
In a previous study in Zackenberg, it was shown that carbon exchange of arctic ecosystems are highly governed by timing of the snow melt (Groendahl et al., 2007).To make sure that the analysis focused on the same relative period of the growing season, satellite images were chosen, that were taken approximately the same number of days (± 7)  To be able to compare images between dates and years, the raw satellite imagery were corrected for atmosphere and terrain using ATCOR 6.1.Additionally, 11 points of non-vegetated flat rock surfaces assumed not to vary in reflectance between years were used as reference to compare the reflectance data between years.Lines forced through zero with reference reflectance of all satellite images against reference reflectance 2007 were fitted and used to recalculate reflectance.To avoid edge effects when comparing the satellite imagery with different sized pixels all images were resampled to 1 m using nearest neighbour, and average reflectance for the corresponding 15 m vegetation cover pixels was subsequently calculated for each image.
Equation ( 2) was used to estimate NDVI, and the linear regression between FAPAR max and NDVI was used to estimate NDVI-based FAPAR max for all images.PAR has been measured hourly between 2002 and 2008 and shortwave irradiance (SW in ) has been measured 1996-2008at C1 (Climatebasis, 2009)).The relationship between PAR and shortwave irradiance is site specific and depending on the time of year (Hansen et al., 2008).A linear regression was fitted using data from 25 July to 5 August 2002-2008 to convert SW in to PAR, (PAR=2.05×SW in +7.41, R 2 = 0.991, df =2015, p = 0.000).PAR estimates at noon for the day of the satellite images 1998 to 2008 were used as incoming PAR.No radiation data existed before 1998 so for years prior to 1998 the average values measured 1998-2008 on the dates of the satellite images were used instead.We considered this method to be acceptable, because the difference between using average instead of minimum or maximum PAR would change LUE-based GPP at maximum 7.7 percent.Furthermore, the same irradiance was assumed over the entire study area, since it is a small area and the satellite images were recorded on days with clear skies.Estimated ε max for the different vegetation types in

BGD Introduction Conclusions References
Tables Figures

Back Close
Full Screen / Esc Printer-friendly Version Interactive Discussion

Evaluation of the LUE model
For evaluation of the parameters in the LUE model we used ground-based measurements of GPP from 2007 and a data set with ground-based measurements of GPP from 1998, 1999 and 2000, measured at 6 wet continuous fen plots using the closed chamber technique (Joabsson and Christensen, 2001).For 1999, no satellite images existed in the time window of (± 7) days after the snowmelt, and this year was omitted from the analysis.GPP measurements for 1998 and 2000 were performed between 10 a.m. and 2 p.m. and an average PAR value for this time period was used as incoming PAR.For 2007, incoming PAR measured at the same time as the GPP measurements were used.Subsequently, the LUE model (Eq. 1) was used to estimate GPP for the two measurements closest in time to the satellite image.In the model, estimated ε max for each vegetation type was used with NDVI-based FAPAR max and measured incoming PAR.Average LUE-based GPP for each satellite pixel was evaluated against average measured GPP with a linear correlation analysis using Pearson's correlation.The normality assumption for residuals was checked using a Kolmogorov-Smirnov test.

The relationship between FAPAR max and satellite-based NDVI
As can be seen in Fig. 2, there was a clear significant linear relationship between NDVI and FAPAR max (FAPAR max =1.173×NDVI-0.072,R 2 = 0.603, F = 16.76,p = 0.002, df =12), indicating that plots with lower FAPAR max had lower NDVI.Further, NDVI from the image 29 July 2007 for the plots where FAPAR max was measured at was on average 0.56 ± 0.07.The measurements of FAPAR max in the respective plots were on average 0.60 ± 0.10 (df =12).where continuous fen had the highest uptake and Cassiope heath had the lowest (Table 2).The vegetation types were generally distributed along a small elevation gradient (within a few m) which in combination with a continuous water flow through the fen areas resulted in a strong moisture gradient ranging from continuous fen, hummocky fen and grassland to heath ecosystems (Salix snowbed, Vaccinium heath, Cassiope heath, Dryas heath).Although, moisture was not specifically measured for all vegetation types it seemed to affect CO 2 fluxes which gradually decreased with decreased moisture level or increased elevation (Table 2).Water table depths were on average 3.4 cm below the ground surface for continuous fen and 8.2 cm below the surface for hummocky fen (not measured in grassland, Salix snowbed, Cassiope, Dryas, and Vaccinium heath since these vegetation types were to dry to find a water table ).A similar trend as for moisture/elevation could be seen in the active layer depth in that the elevated/drier ecosystem types had a deeper average active layer than the wetter ecosystems, range 41 cm to 65 cm.Grassland was in-between the wetter and drier ecosystems.Despite the difference in active layer there were no differences between vegetation types in soil temperature (Table 2).

FAPAR max and light use efficiency for high arctic vegetation types
There were significant differences in chamber-based FAPAR max (on average 0.60) for the vegetation types except for between the continuous fen and the hummocky fen and Introduction

Conclusions References
Tables Figures

Back Close
Full between the different heath vegetation types (Cassiope, Dryas, Vaccinatium, and Salix snowbed).Chamber-based FAPAR max was higher for wet vegetation types whereas it was lower for the heath vegetation types (Table 3).
There was a tendency towards differences in light use efficiency, ε max , between the different vegetation types (Table 3).ε max was on average 1.83 g CO 2 MJ −1 and it differed with a maximum of 2.29 g CO 2 MJ −1 for the continuous fen to a minimum of 1.11 g CO 2 MJ −1 for the Dryas heath.Due to large spatial and temporal variation the differences were however not significant (F = 1.48, p = 0.193, df =109).

Evaluation of the LUE model
LUE-based modeled GPP was highly correlated to ground-based measurements of GPP (Pearson correlation=0.947,p = 0.00001) (Fig. 3).It can be seen in Fig. 3 that for the individual years the LUE-based GPP was somewhat larger than field-measured GPP 1998 and 2007 whereas for 2000 the model fitted well.Over the entire evaluation, the LUE-based GPP was larger (928.2mg CO 2 m −2 h −1 ) than field-measured GPP (720.5 mg CO 2 m −2 h −1 ), indicating that the model on average slightly overestimated GPP.

Investigation of temporal trends in GPP between 1992 and 2008
There was a strong increase in NDVI from 1992 to 2008 for all vegetation types, indicating that there is a greening trend in the Rylekaerene area.There was also a substantial increase in LUE-based GPP 1992-2008 (Fig. 4).The increase was especially dominant up until 2002 whereafter it started to level out.The trend was similar for all vegetation types, even though they differ in GPP.The LUE-based GPP followed the same moisture gradient as the field measurements in that it was gradually decreasing from continuous fen, hummocky fen, grassland, Salix snowbed, Vaccinium heath, Cassiope heath, and Dryas heath.A strong decrease in LUE-based GPP could be seen for 2005 (Fig. 4).Average annual air temperature also increased in Zackenberg between Introduction

Conclusions References
Tables Figures

Back Close
Full 1992 and 2008 (Fig. 4).Average annual air temperature is based on measurements at C1 for 1996C1 for to 2008C1 for (Climatebasis, 2009)).A line between measurements at C1 and measurements at Danmarkshavn (Cappelen, 2007) between 1996 and 2006 was fitted and used for estimating average annual air temperature for the years 1992 to 1995.A fitted line indicates that the average increase in temperature between 1992 and 2008 in Zackenberg was 0.15 • C year −1 (R 2 0.708).

Increased GPP for high arctic fen areas 1992 to 2008
Our result showed a strong overall increase in GPP 1992GPP -2008 (ranging between 100 to 400 mg CO 2 m −2 h −1 ), which is in agreement with expectations for high latitude ecosystems as the climate is warming (Oechel et al., 2000;Shaver et al., 2000).Correspondingly, average annual air temperature in or in close proximity to the Zackenberg area also increased 1992-2008 (Fig. 4).At low temperatures photosynthesis, as most chemical reactions, is temperature limited, and it increases with temperature (Chapin et al., 2002).Additionally, an increase in temperature prolongs the growing season.
The globally averaged CO 2 concentration at the sea surface has increased on average 1.8 ppm per year since 1992 (Tans, 2009) and a frequent explanation to an increased GPP is an elevated atmospheric CO 2 concentration (Loustau et al., 2001).An increase in the CO 2 concentration has been shown both to stimulate photosynthesis and to restrain photorespiration (Loustau et al., 2001) and it also decreases stomatal conductance, resulting in a decreased water loss due to transpiration (Chapin et al., 2002).Climate change may also results in increased plant growth when plants respond to an increased soil nutrient supply due to warming induced increases in weathering, nitrogen fixation (Sorensen et al., 2006) and decomposition (Robinson et al., 1997).Most natural plant communities today are nutrient limited, giving scarce plant cover.
Consequently, an increase in soil nutrients could result in an increased plant cover, as

Conclusions References
Tables Figures

Back Close
Full indicated by the increased NDVI values seen in this study.Furthermore, the increase in NDVI may point towards a continuous change in vegetation type covering the area supported by the fact that such changes have been reported from other sub-arctic and arctic areas (Oechel et al., 1993;Malmer et al., 2005).
It is however debated how climate change will affect the plants on a long term basis, as it has been shown that the effect of increased CO 2 can be reduced due to changes in moisture and nutrient regimes, and species composition (Loustau et al., 2001;Shaver et al., 2000).As can be seen in Fig. 4, there was a clear increase in LUE-based GPP up until 2000 whereafter it levels out.Ellebjerg et al. (2008) showed a decrease in NDVI 1999-2006 for several high arctic tundra vegetation types in Zackenberg.This was explained both by an earlier timing of snow melt, which reduced water availability during the peak of the growing season, and by an elevation in temperature, which increases evapotranspiration (Ellebjerg et al., 2008).Increased evapotranspiration can result in water loss at high temperatures, which results in drier soils, strongly limiting GPP so that the potential temperature driven increase is missed out.Our results also point to the importance of soil water relations since we found that net carbon gain was largest in the wet fen ecosystem and smallest in the dry heath ecosystems (Table 2).
The year 2005 stands out in comparison to the other years (Fig. 4).It was also the year with the earliest snowmelt (DOY 158, Table 1), and with highest average annual temperature (Fig. 4).The temperature was especially high during wintertime, and there were low snow-cover and several thaw events both during winter and spring (Climatebasis, 2009;Ellebjerg et al., 2008).This indicates the importance of all-year climate for the vegetation.

CO 2 flux, NDVI, FAPAR max and the light use efficiency of high arctic vegetation
The measured CO 2 fluxes (NEE, ER and GPP in Table 2) were quite large in comparison to other arctic studies but still in the same range (Lafleur and Humphreys, 2008;Nobrega and Grogan, 2008;Soegaard and Nordstroem, 1999;La Puma et al., 2007;

BGD Introduction
Full Screen / Esc Printer-friendly Version Interactive Discussion Arndal et al., 2009).However, some of these studies were on diurnal basis, lowering both GPP and NEE due to low temperatures and low solar angles in comparison to daytime CO 2 fluxes, which were measured in this study.The linear relationship between NDVI and FAPAR max were shown to be correlated for this high arctic fen area as in accordance with several previous studies for other global land use classes (Myneni et al., 1997b;Olofsson and Eklundh, 2007;Asrar et al., 1984).In Rylekaerene, chamber-based FAPAR max was on average 0.60.We have not found any other studies estimating FAPAR max for high arctic tundra vegetation types, but the fen areas and the grassland had approximately the same values as agricultural and grassland areas studied in the US (Privette et al., 1996;Verstraete et al., 2008) and the heath ecosystems were similar to semi-arid grass savanna (Fensholt et al., 2004;Gobron et al., 2008).Maximum light use efficiency (ε max ) in Rylekaerene was low (1.83 g CO 2 MJ −1 ) in a worldwide comparison; if it is assumed that 0.5 of GPP goes to NPP (Schlesinger, 1997), average ε max would be 0.247 g C NPP MJ −1 , whereas a modeled worldwide average is 0.427 g C NPP MJ −1 (Ruimy et al., 1999).
However, arctic ecosystems are constrained in their productivity due to extreme temperatures, short growing seasons, low water and nutrient availability and low quantum yield and therefore low in a worldwide comparison.The observed light use efficiency of 0.247 g C NPP MJ −1 was similar to other grassland ecosystems but quite large in comparison to an ε max value for arctic tundra as assumed by global models (Gower et al., 1999).LUE-based GPP was larger than field measured GPP.An explanation to this could be that PAR used in the LUE-based GPP was measured outside a chamber, whereas field-measured GPP measured inside a chamber was constrained due to lower transparency.Another explanation could be that incoming and reflected PAR was measured inside the chamber and may have been influenced by the chamber edges and the collars of the box.

Conclusions
By applying the light use efficiency model to a satellite data set ranging from 1992 to 2008, we show a substantial increase in GPP in a high arctic fen area during this period.
During the same period there has also been a strong global increase in CO 2 concentration and a local increase in air temperature indicating a global change induced increase in GPP.We also parameterized the light use efficiency model for the dominant arctic vegetation types of North Eastern Greenland.FAPAR max was well correlated to NDVI and our investigation shows that NDVI can be used for spatial and temporal extrapolation of FAPAR max in high arctic areas.These findings support the view that NDVI is a useful vegetation index for FAPAR investigations.The light use efficiency model is a simple approach relating ecosystem photosynthesis to absorbed PAR.For this high arctic fen area, the light use efficiency was on average 1.8 g CO 2 MJ −1 , which is reasonable for high arctic ecosystems.The LUE-based GPP was well correlated to the field measurements, consequently extrapolating FAPAR spatially and temporally and combining it with the LUE model is a strong tool in the study of global carbon budgets.
The light use efficiency model could be used for up-scaling GPP also for larger areas and it revealed a large GPP change in this high arctic fen area 1992-2008.Introduction

Conclusions References
Tables Figures

Back Close
Full after the snow melt as the Landsat image from 29 July 2007.Snow depth has been measured continuously since 1998 at the climate station (C1) in the centre of the Zackenberg valley (Fig.1).The day of year (DOY) when snow depth was lower than 10 cm was used as the time of snowmelt.Modeled data from1989-2004(Buus-Hinkler et al., 2006) were used to estimate snow cover before 1998.A linear regression was fitted between the DOY of measured 10 cm snow depth and modeled DOY with 18 percent snow cover of the Zackenberg valley for the years 1998-2004 (R 2 = 0.9766, df =5, p = 0.00003), 18 percent was used as a proxy for 10 cm snow depth since the major snow period is considered to end when the snow cover percentage drops below 18 percent according toBuus-Hinkler et al. (2006).The regression line was then used to estimate DOY of 10 cm snow depth for 1992-1997.All satellite images chosen are given in

Fig. 1
was used together with PAR and the NDVI-based FAPAR max estimates to model GPP from 1992 to 2008.Water and gravel areas were given an ε max of zero.Coverage of the vegetation types is set to be static since we only had estimates from 2007; incorporating a dynamic vegetation change in the LUE model could be a way to improve the analysis in the future.

Fig. 1 .
Fig. 1.The Zackenberg valley and the investigation area surrounding the Zackenberg research station (ZERO).The map of the dominant vegetation types estimated in the area surrounding Rylekaerene is superimposed over the area.Climate station (C1) and the Zackenberg research station is marked with crosses.Areas for the FAPAR max measurements are marked with black circles and the area of the chamber measurements are marked with a black rectangle.The red dot on the Greenland map is the location of Zackenberg.
The dominant vegetation Introduction

Measured daytime CO 2 fluxes and environmental variables
Average GPP during the measurement period ranged from 654.1 to 188.2 mg CO 2 m 2 h −1 for the different vegetation types, where continuous fen had the highest GPP and Dryas heath had the lowest.Average ER ranged from