4.1. Multiple Levels of Nitrogen Addition
 Our results show that prostrate dwarf-shrub herb tundra in Northwest Greenland is very sensitive to small N additions, as CO2 exchange and NDVI were significantly altered in N0.5 plots. Although 0.5 g m−2 y−1 of N addition is small compared to other fertilization studies, it represents a doubling of the annual net N mineralization of 0.2 g N m−2 y−1 observed in a polar semi-desert on Svalbard [Robinson et al., 1995]. Rates of GEP (greater plant uptake of CO2) and ER (greater CO2 efflux) showed high sensitivity to N0.5 addition as both increased by a nearly a factor of two compared to CTL plots. NEE was significantly reduced (greater CO2 source) in N0.5 plots resulting in a 70–80% increase in net CO2 efflux to the atmosphere as determined during midday measurements. Similarities in the functional form of response to N additions in measurements of ecosystem CO2 exchange, NDVI and plant cover, coupled with strong correlations between GEP, ER and NDVI suggest increases in vascular plant cover and biomass largely explain increases in rates of GEP and ER. To our knowledge, no other study has assessed the impact of low levels of N addition on CO2 exchange in arctic ecosystems, and increasing industrialization in the Arctic will likely lead to rates of atmospheric N deposition that exceed those of 0.1 g m−2 y−1 observed on Svalbard [Hodson et al., 2005].
 Foliar N responded linearly to N additions with the highest foliar N concentrations in N5.0 plots for C. rupestris, D. integrifolia, and S. arctica, indicating that plants allocated some of the additional N to leaf material. The observed higher leaf N concentrations may have been stored for future use [van Wijk et al., 2003b], or used immediately to support photosynthesis [Baddeley et al., 1994]. Measurements of C. rupestris foliar δ13C revealed a marginally significant increase in foliar δ13C in N1.0 plots and a significant increase in foliar δ13C in N5.0 plots. Increases in foliar δ13C generally reflect a decrease in leaf intercellular CO2 concentrations, which may arise through either a reduction in stomatal conductance or an increase in photosynthesis [Farquhar et al., 1989]. It seems likely that increases in foliar δ13C in N1.0 and N5.0 plots were caused by an increase in photosynthesis of C. rupestris, given that increases in foliar N were observed, while there was no evidence of a change in soil water contents with N addition and studies of plant water relations near Pitufik have found limited signs of water stress [Sullivan and Welker, 2007]. Increased photosynthetic rates of C. rupestris may have contributed to the increase in GEP observed in N1.0 and N5.0 plots.
 The functional form of the response to increasing levels of N addition was nearly identical in CO2 exchange and NDVI measurements and followed a nonlinear pattern such that the greatest response occurred with 1.0 g m−2 y−1 of N addition with a saturating or decreasing response at higher levels of N addition. A trend toward decreasing GEP, ER and NDVI in N5.0 plots suggested that ecosystem, or vegetation, N saturation began to occur between 1.0 and 5.0 g m−2 y−1 of N addition [Aber et al., 1989]. N saturation is defined as the point at which ammonium and nitrate availability exceed plant and microbial demands for N [Aber et al., 1989]. Prior to N saturation, tree growth, leaf N and net primary production typically increase with additional N [McNulty et al., 1996], but as N saturation occurs, it can cause reductions in net ecosystem photosynthesis and net primary production [Aber et al., 1989]. While the suite of measurements collected prevent quantitatively detecting ecosystem N saturation, the addition of multiple levels of N to our site gave rise to several responses similar to those observed in N saturated temperate forests. Foliar N concentration progressively increased with increasing rates of N addition in C. rupestris, D. integrifolia and F. brachyphylla with a trend toward increasing foliar N in S. arctica [Arens, 2007]. A concurrent increase in GEP and NDVI of N0.5 and N1.0 plots occurred in both years, but, as predicted by the final stage of N saturation, GEP declined and NDVI tended to be lower in N5.0 plots compared to N1.0 plots despite very high foliar N in N5.0 plots. Strong positive responses of CO2 exchange and NDVI to small additions of N demonstrate that prostrate dwarf-shrub, herb tundra was primarily limited by N, but a decrease or saturation of response in those variables with high levels of N addition suggest ecosystem, or vegetation N saturation.
4.2. Factorial Nitrogen and Phosphorus Additions
 Factorial additions of N and P shed light on a secondary limitation to ecosystem productivity. Combined N and P addition radically altered GEP, ER, NEE, NDVI, and plant community composition, releasing the system from apparent N saturation, suggesting that a strong colimitation with P arose between N additions of 1 and 5 g m−2 yr−1. Ecosystem CO2 exchange was highly sensitive to N, but showed the greatest magnitude of response to combined addition of N and P. N and P interacted very strongly to increase GEP five to sixfold and enhanced ER by 250% over CTL plots. Addition of N stimulated both GEP and ER more than P. Unlike our results, factorial addition of N, P and K to wet sedge tundra in Alaska, showed that P stimulated CO2 flux more than N and there was only a small interaction between N and P [Shaver et al., 1998]. The positive synergistic effect of N and P addition on CO2 exchange stimulated GEP slightly more than ER resulting in significant NEE increases during 2005. Higher rates of NEE in 2005 compared to 2006 may be the result of either a warmer 2005 growing season or the diminished response of CO2 exchange to fertilization by 2006.
 Changes in CO2 flux due to multiple levels of N and N + P addition can largely be explained by changes in plot-level NDVI and plant community composition. N and P interacted significantly to increase NDVI from 0.3 in CTL plots to 0.6 in N + P plots. NDVI in N + P plots in Pituffik was greater than that observed in unaltered Alaskan wet sedge tundra [Boelman et al., 2003] and similar to that measured in a highly productive fen near Pituffik [Sullivan et al., 2008a]. NDVI increases in N + P plots were due to large decreases in bare ground, a shift in species composition and development of a multilayered canopy. Over two years of measurement, NDVI was significantly correlated to CO2 exchange, explaining 81 to 88% of variation in GEP and 80% of variation in ER. Correlations between NDVI, GEP and ER were also observed in wet sedge tundra [Boelman et al., 2003], but were weaker than those observed at our site.
 Measurements of belowground respiration in 2004 showed that neither N nor combined N and P addition directly limited microbial respiration, similar to results from a study in eastern Greenland [Illeris et al., 2003]. Belowground respiration in N + P plots increased progressively during each year of this study. Such a response could be caused by stimulation of microbial respiration through greater inputs of higher quality litter (including root exudates) or through large increases in plant root respiration. Although litter input was not quantified, NDVI measurements indicate large increases in aboveground biomass by 2005. Increases in foliar N with N + P addition were observed in nearly all measured plant species. Increases in litter quantity and quality may have stimulated microbial respiration by relieving limitations associated with C quality [Hobbie and Chapin, 1996; Hobbie et al., 2000] as substrate quality is one of the primary controls of decomposition [Heal et al., 1981].
 High N addition resulted in a small, nonsignificant decrease in bare ground, but N + P addition decreased bare ground from over 60% to less than 20%. Superimposed on changes to plant cover in N + P plots was an inherent difference in plant cover between blocks. We hypothesize that differences in plant cover were largely responsible for the significant effect of block on CO2 exchange, belowground respiration and NDVI. Establishment of graminoids, principally F. brachyphylla, was the primary reason for the reduction of bare ground. Colonization of bare ground by vascular plants subjected to N and P additions has also been observed in polar semi-desert [Robinson et al., 1998] and alpine tundra [Heer and Körner, 2002]. F. brachyphylla constituted only 1% of cover in CTL plots, but, after two years of N + P addition, it became the dominant vascular plant species, comprising over 40% of the total cover in N + P plots. Significant interactions between block, N and P were present only in NEE measurements. Graminoids were the most responsive functional group in our study and have shown a strong positive response to combined N and P fertilization in many tundra ecosystems, including alpine tundra [Heer and Körner, 2002], low arctic tundra [Chapin and Shaver, 1985; Gough et al., 2002; Gough and Hobbie, 2003], and high arctic tundra [Henry et al., 1986; Robinson et al., 1998]. The strength of the graminoid response at our site appears to be unique within the High Arctic, as other nutrient additions to similar landscapes showed a comparatively weaker graminoid response with forbs and deciduous shrubs responding most strongly and establishing more often on bare ground areas [Henry et al., 1986; Baddeley et al., 1994; Robinson et al., 1998].
 Increases in species richness and diversity accompanied dramatic increases in graminoid cover. Forbs, bryophytes and the deciduous shrub S. arctica all increased cover in response to N + P fertilization. Total forb cover in N + P plots increased nearly twenty-fold compared to CTL plots, with most of that increase from vascular plants not present in CTL plots. The effects of fertilization on species richness and diversity in the Arctic vary from complete dominance by a single species, reducing species richness and diversity [Chapin et al., 1995], to situations in which richness and diversity have increased [Robinson et al., 1998] to others where richness and diversity have remained unchanged [Gough and Hobbie, 2003]. The only fertilization studies to observe increases in species richness and diversity were in the High Arctic where open canopy ecosystems dominate [Robinson et al., 1998]. In a Svalbard polar semi-desert, after three years NPK addition, previously unobserved forbs characteristic of bird cliff communities, established in bare ground areas [Robinson et al., 1998]. The mechanism and type of change to species composition in Svalbard was very similar to changes observed at our site, but the speed and magnitude of changes in Pituffik were substantially greater.
 In our study, prostrate dwarf shrub, herb tundra responded strongly to small N inputs and showed a magnitude and speed of response to factorial N and P additions not previously observed. Significant changes to ecosystem CO2 exchange and NDVI in response to low levels of N addition suggest that future atmospheric N deposition, even at relatively low rates, may alter both vegetation characteristics and net rates of CO2 efflux as N0.5 addition caused a 70% decrease in NEE (greater net CO2 efflux). After N + P addition, ecosystem and vegetation characteristics more closely resembled characteristics of ecosystems one subzone further south [CAVM Team, 2003], or those fertilized for centuries by nesting sea birds [Robinson et al., 1998]. Also unique to our study is the rapidity of ecosystem and community level responses to N and P additions. Changes in plant species composition began to occur after only one year of nutrient addition. Our observations contribute to a growing body of evidence that suggests a strongly divergent response to nutrient additions in ecosystems of the High and Low Arctic. Nutrient additions to low arctic tundra have almost always led to reductions in species diversity and richness, while elevating codominant species to dominance [Chapin et al., 1995]. Nutrient additions in the High Arctic have generally led to increases in species diversity and richness through a strong stimulation of rare subdominant plant species [Henry et al., 1986; Robinson et al., 1998]. Increases in plant available N, due to atmospheric N deposition or warming-induced increases in net N mineralization, will likely alter carbon cycling and vegetation characteristics of prostrate dwarf-shrub, herb tundra, but the magnitude of the response will be constrained by P availability.