4.1. IR Radiation Additions and Air Warming
 IR radiation additions in our study (30 and 60 W/m2) were intermediate among those in previous experiments (22 W/m2 [Harte et al., 1995]; 80 W/m2 [Shaw et al., 2002]). Energy additions increased temperatures in the soil and plant canopy, but not in the atmosphere, above the plant canopy. This observation is consistent with results in a montane meadow [Saleska et al., 1999], but contrasts with reports from high arctic graminoid tundra [Nijs et al., 2000] and tallgrass prairie [Wan et al., 2002]. Open air warming is not expected to occur because there are too few molecules of radiatively active gases between an IR radiator and the plant canopy. Different reports of IR radiator effects on air temperatures may reflect differences in canopy structure and measurement height. In both cases where air warming was reported, air temperature measurements were taken within the canopy of graminoid-dominated ecosystems, which tend to be decoupled from conditions in the overlying atmosphere. Our measurements were made above the canopy, while those in the montane meadow were made in a comparatively rough canopy that is probably well coupled with the atmosphere.
4.2. Plant Canopy Temperatures
 Plant canopy temperatures varied by as much as 8.8°C in ambient plots, reflecting differences between the north- and south-facing sides of D. integrifolia clones. Canopy temperatures were as much as 13.6°C warmer than air temperatures in ambient plots when wind speeds were light. The magnitude of this temperature difference is consistent with reports for other prostrate arctic plant species [Mølgaard, 1982; Hart and Svoboda, 1994; Gold, 1998]. Wind speed was the most important determinant of midday canopy temperatures, alone explaining 55% of the variation observed over six sampling campaigns. Our results suggest that during an average midday period in the high arctic growing season, when air temperatures are between 6 and 9°C, wind speeds greater than ∼4 m/s could be the difference between leaf temperatures that are near optimal for photosynthesis (∼15°C [Larcher, 2003]) and those that are well below optimal. Supplemental radiation additions were identified as a significant term in a multiple regression model used to predict canopy temperatures. Canopy temperatures were always warmer in T1 and T2 plots than ambient plots when measured, but the warming effect varied from near zero to ∼+4°C and ∼+6°C, respectively, with the greatest increases in canopy temperatures generally observed when wind speeds were light.
4.3. Soil Warming
 The soil warming effect at 2 cm depth varied seasonally and diurnally with wind speed. Dependence of the warming effect on wind speed may reflect convection of heat away from the plant canopy and or from the heating element [Kimball, 2005]. Most experiments using IR radiators have held input electrical energy constant and assumed that doing so would hold the IR flux constant [Harte et al., 1995; Wan et al., 2002; Shaw et al., 2002]. However, recent work has shown that the efficiency of IR radiators declines rapidly with increasing wind speed [Kimball, 2005]. Several investigators have suggested modifying traditional IR radiators to achieve a constant canopy warming effect by varying the electrical energy input [Nijs et al., 2000; Kimball, 2005]. Our observations suggest that such an approach should be considered for experiments at windy sites.
 The influence of wind speed and the magnitude of soil warming effect declined with depth. Soils at 10 cm depth were, however, warmer in T1 and T2 plots than ambient plots, suggesting that the warming effect probably penetrated to depths greater than 10 cm. The magnitudes of near surface soil warming in our study are within the range of values reported in other experiments that have used IR radiators [Harte et al., 1995; Bridgham et al., 1999; Nijs et al., 2000; Wan et al., 2002]. There was no evidence that soil temperatures in plots that received supplemental water were significantly different than ambient plots, suggesting that outdoor storage and evening application were sufficient to eliminate temperature artifacts during irrigation.
4.4. Soil Water Contents
 There was no evidence that supplemental radiation reduced soil water contents, nor was there evidence that supplemental water interacted with supplemental radiation to affect soil temperatures. These observations contrast with the results of previous IR warming experiments, where soil drying and strong interactions between energy supplements and soil water contents have been observed [Harte et al., 1995; Wan et al., 2002]. The lack of an interaction between energy additions and soil water at our site was surprising, given that soil drying was observed in similarly dry, well-drained soils [Harte et al., 1995]. The near surface soils of the dry upper zone in the experiment of Harte et al.  have very low bulk density (0.46 g cm−3 [Saleska et al., 2002]) relative to soils at our site (1.70 g cm−3). Comparison of our results with those of Harte et al.  suggests that differences in soil aeration may be an important control on the IR radiator soil drying effect.
 Water additions did not lead to a detectable, sustained increase in soil water contents, in contrast with our expectations. The absence of detectable, sustained increases in soil water can be attributed to the well drained soils (∼70% sand, ∼25% silt, ∼5% clay), the resolution (0.1%) and precision (0.05%) of our soil water probes and the strongly seasonal environmental and biological controls on soil water contents at our site. The soil water regime at our site can be divided into three distinct time periods for the sake of discussion: early season, midseason and late season.
 During the early season period, soil water contents rapidly decline from more than 40% by volume in the postsnowmelt period to become relatively dry (20–30%). Soil water contents during this period are largely governed by soil thaw and are relatively impervious to precipitation. The period begins with thawing of near surface soils in early June and extends until soil water contents reach their seasonal minima in early July. Net radiation tends to be positive in this early season period and evaporative water loss is probably important to the draw-down of soil water contents. Leaf area increases rapidly from late June to early July. Vascular plant transpiration may, therefore, play an important role in the dry-down of soil water contents near the end of this early season period.
 During the midseason period, soil water contents vary between 20 and 30% by volume. During this period, soil water contents show a small response to large precipitation events. The midseason period extends from the first week of July until the first week of August. Air temperatures, soil temperatures and leaf area are highest during this period. Both evaporation and transpiration likely play important roles in regulating soil water contents during this period, but the proportional importance of transpiration is probably greater than during the postsnowmelt period. Experimental water additions during the midseason period were detectable as subtle and ephemeral increases in soil water content when additions were temporally isolated from natural precipitation events. Water supplements were probably lost quickly through drainage, evaporation and transpiration during this period.
 During the late season period, soil water contents rise and are most responsive to precipitation, showing relatively large responses to small events. The late season period extends from the first week of August until soils freezeup, typically in mid-September. Leaf area and evaporation both decline steeply during August and probably become only weak regulators of soil water contents during this late season period.
 Late season water additions were not detectable during 2004, as they were superimposed upon a period of frequent natural precipitation events and rapidly rising soil water contents. In 2005, late season water additions were detectable as subtle increases in soil water content when additions were temporally isolated from natural precipitation events (e.g., 11 August 2005). In general, soil water contents tended to increase more in response to natural precipitation events than to our water supplements when events were of similar size, suggesting that evaporative water loss during or soon after irrigation may have reduced the effect of water supplements on soil water contents.
 Natural precipitation events in August had greater consequences for soil water contents than precipitation events during July. For instance, on 16 and 17 July 2005, the site received 12.0 mm of rain and soil water contents increased by 2.2%. In contrast, on 4 August 2005 the site received 8.7 mm of rain and soil water contents increased by 5.2%. We suggest the greater response of soil water contents to precipitation events in August reflects the seasonal decline in evapotranspiration, which is an important regulator of soil water contents during the midsummer period when soil water contents are relatively impervious to natural precipitation and experimental water additions. A similar hypothesis was presented to explain late season increases in soil water contents of a Mediterranean annual grassland [Zavaleta et al., 2003].
 Prostrate dwarf-shrub, herb tundra generally showed relatively simple microclimate responses to energy and water additions, in contrast with similar experiments in other ecosystems, where microclimate responses have involved strong interactions between radiative forcing and other experimental treatments, soil water contents and or vegetation responses [Harte et al., 1995; Wan et al., 2002; Zavaleta et al., 2003; Klein et al., 2005]. Our experiment provides a framework to investigate the sensitivity of a cold, dry ecosystem to changes in climate and provides further evidence that similar manipulations may lead to contrasting responses in different ecosystems [Harte et al., 1995; Klein et al., 2005].