Some of the papers in this special section have examined processes in the Western Arctic, while others have examined processes in other regions of the pan-Arctic. While all of the studies focused on the Western Arctic were conducted as part of ATLAS, some of the studies that are focused on other regions [e.g., Boike et al., 2003; Nolan et al., 2002] represent international collaborations with ATLAS. To better compare and contrast insights relevant to the Western Arctic versus other regions, we have separated discussion of the papers of this special section with respect to whether they focused on studying processes in the Western Arctic or in other regions of the circumpolar Arctic. As we discuss each paper, we provide the background from prior research relevant to the results of each study, with an emphasis on prior results from LAII research in the Flux Study and the ATLAS Project. The background we provide specifically focuses on the contributions of each study in providing insight concerning controls over spatial and temporal variability in the exchanges of water/energy and trace gases between the land and the atmosphere, and freshwater delivery to the Arctic Ocean.
3.1. Insight From Studies Focused on the Western Arctic
 Both Bonan et al.  and Foley et al.  have conducted studies with general circulation models that indicate that the position of northern tree line has a substantial influence on global climate. Regional modeling studies focused on Alaska have shown that the expansion of shrub tundra at the expense of sedge tundra may result in substantially warmer summers over tundra, with warming effects that extend into the boreal forest of Alaska [Lynch et al., 1999; Chapin et al., 2000b]. While these modeling studies clearly highlight that vegetation change has the potential to influence climate, the magnitude and extent of impacts on climate will depend on the temporal and spatial patterns of land cover change in the circumpolar Arctic. Two ATLAS studies have documented that tundra ecosystems in Alaska are becoming more shrubby on the North Slope [Sturm et al., 2001a] and on the Seward Peninsula [Silapaswan et al., 2001] over the last several decades. The study by Lloyd et al.  complements these two studies by documenting the response of the tree line ecotone on the Seward Peninsula in Alaska to 20th century warming. Through the use of tree rings to reconstruct the response of tree line to warming, the study by Lloyd et al.  indicates that spruce trees located in upland tundra have established progressively farther from the forest limit since the 1880s. This has led to a conversion of shrub tundra into low-density forest–tundra within a band extending approximately 10 km from the forest limit. Modeling experiments conducted by Lloyd et al.  suggest that fire may play a role in the expansion of tree line, and that large and nearly instantaneous responses to warming are likely at the tree line ecotone. Together, the studies by Lloyd et al. , Sturm et al. [2001a], and Silapaswan et al.  provide important information on the temporal and spatial dynamics of vegetation change in arctic Alaska over the last century. A key question raised by these studies is whether similar changes are occurring in other terrestrial regions of the pan-Arctic.
 Based on the results of the FLUX Study, several studies have focused on questions related to spatial and temporal extrapolation of carbon dynamics over the Kuparuk River Basin in Alaska [Hobbie et al., 1998; Clein et al., 2000; McGuire et al., 2000b; Oechel et al., 2000b; Williams and Rastetter, 1999; Williams et al., 2000, 2001]. The study by Le Dizes et al.  builds on these previous studies in several ways. First the uptake of carbon by the vegetation is now simulated by the aggregated canopy model (ACM), which has been developed and tested in the context of eddy covariance data available for a N-S transect across the basin [see Williams and Rastetter, 1999; Williams et al., 2000, 2001]. Second, the dynamics of the new version of the model have been calibrated and verified (1) in the context of decadal-scale experiments that have manipulated temperature, nutrients, and light and (2) in the context of an experiment that has manipulated atmospheric carbon dioxide [see also Hobbie et al., 1998; Clein et al., 2000]. Third, the study has broken new ground by demonstrating how it is possible to use remotely sensed data to verify model dynamics in a retrospective fashion, and then use the model to simulate the dynamics for projected variations in climate. Fourth, the study has conducted a time series analysis and has identified that while the immediate response to year-to-year variation in temperature is the release of carbon in a warmer year, the response a year later is to increase carbon storage. This result has implications for longer-term trends in warming and is consistent with the study of Oechel et al. [2000a], which has documented an initial release of carbon followed by the storage of carbon among studies that have examined summer carbon dynamics of tundra ecosystems on the North Slope of Alaska over the last several decades. The result is also interesting in the context of the study by Braswell et al. , which shows a 9-month lag in carbon storage to increasing temperature at the global scale and has evaluated lags in NDVI response with temperature for various biomes globally. Finally, Le Dizes et al.  has evaluated nitrogen cycle issues responsible for long-term changes in carbon storage by partitioning the responses of carbon storage. This analysis has identified that the increase in vegetation carbon/nitrogen ratio (i.e., more wood) and the change in redistribution of nitrogen from the soil to plants are key factors responsible for increases in carbon storage, and has shown that the relative strength of these factors in the future depends on moisture conditions. These results are particularly important in that they are consistent with information from other ATLAS studies that tundra in Alaska is taking up more carbon in summer [Oechel et al., 2000a] and is becoming more shrubby, i.e., more woody [Sturm et al., 2001a; Silapaswan et al., 2001]. The results from the study by Le Dizes et al.  demonstrate the power and value of integrating field and experimental studies of processes with process-based modeling studies in ATLAS.
 Processes that occur in the nongrowing season are important to understand because the nongrowing season lasts 9 or more months of the year. A number of studies that have measured the exchange of carbon dioxide between tundra ecosystems and the atmosphere during the nongrowing season have documented that substantial losses of carbon dioxide from tundra soils may occur during fall, winter, and spring months [Kelley et al., 1968; Coyne and Kelley, 1971, 1974; Zimov et al., 1993, 1996; Oechel et al., 1997; Fahnestock et al., 1998, 1999; Grogan and Chapin, 1999; Jones et al., 1999]. Modeling studies have also indicated that processes that control the release of carbon dioxide from soils during the nongrowing season are relevant in the context of the global carbon cycle [McGuire et al., 2000a]. While it has been documented that carbon dioxide loss from soils of tundra ecosystems is a major part of the annual carbon budget, relatively little is known about controls and how they operate in the nongrowing season. This is particularly important to understand as high-latitude ecosystems contain approximately 40% of the soil carbon stored globally in terrestrial ecosystems [McGuire et al., 1995], and projections of future warming indicate that high-latitude ecosystems will experience the greatest warming in the nongrowing season [IPCC WGI, 2001]. The study by Michaelson and Ping  is specifically focused on elucidating temperature controls over decomposition in the nongrowing season and on understanding how temperature responses interact with respect to how easily organic matter is decomposed by microbes, i.e., with respect to substrate quality. In laboratory incubations at −2°C, they found that carbon dioxide loss was correlated with water-soluble organic carbon (wsOC) levels, which is generally considered more readily decomposed by microbes in comparison to organic carbon this is not water soluble. Soils collected from permafrost had higher levels of wsOC in comparison with soils from the active layer, and levels of wsOC were not correlated with total organic carbon levels. Thus, the study by Michaelson and Ping  highlights the importance of understanding how substrate quality interacts with soil thermal dynamics to influence the decomposition of soil organic matter during the nongrowing season. The results of the study are also relevant to the issue of understanding how the warming and melting of permafrost will influence the release of carbon dioxide from tundra soils.
 The active layer, i.e., the layer above permafrost that experiences seasonal thawing during the summer and freezing during the winter, is an important area of hydrological and biological activity in tundra ecosystems [Kane et al., 1991]. Thus, an understanding of the controls over spatial and temporal variation in active layer thickness is important to understanding controls over spatial and temporal variability of hydrological and biological activity in tundra ecosystems. The Circumpolar Active Layer Monitoring (CALM) program was established to study the impacts of climate change in permafrost environments, and currently consists of more than 85 sites in 11 countries in the Northern Hemisphere. The study by Hinkel and Nelson  analyzes 6 years of variability in summer thaw depth for three CALM 1 km2 grids located on the arctic coastal plain in Alaska and for four CALM grids located in the northern foothills of the Brooks Range. For each of the grids, interannual variability in the end of season thaw depth is strongly correlated to the local growing season surface air temperature. On the coastal plain, thaw depth is greatest in thaw lake basins. Within each of the grids, spatial variation of thaw depth appears to depend on complex interactions among the local influences of vegetation, substrate properties, snow cover dynamics, and terrain.
 Previous research from the Flux Study and ATLAS has identified that spatial variation in the function and structure of tundra ecosystems is influenced by climate and soil parent material [Hobbie et al., 1998; Epstein et al., 2000, 2001; Walker et al., 1998; Walker, 2000; McGuire et al., 2000b]. While this research has documented that water, energy, and carbon dioxide exchange between tundra ecosystems and the atmosphere vary spatially, this spatial variation is not completely understood and is not well represented in large-scale models of climate and ecosystem dynamics. To better understand controls over the variability in vegetation structure and function, the study by Walker et al.  examined aboveground phytomass, leaf area index (LAI), and the normalized difference vegetation index (NDVI) across a climate gradient in northern Alaska on acidic and nonacidic soil parent material along two transects (Barrow to Ivotuk and Prudhoe Bay to Toolik Lake). Along the summer temperature gradient spanned by the study, phytomass increased over 200% on acidic soils and approximately 50% on nonacidic soils with increasing temperature. There was a 700% increase in shrub phytomass on acidic substrates, but only a 70% increase on nonacidic substrates. While there was a doubling of LAI on acidic substrates over the summer temperature gradient, there was no response of LAI on nonacidic substrates over the gradient. However, NDVI increased on both substrates along the summer temperature gradient. The patterns elucidated by Walker et al.  provide relationships that should be useful in specifying spatial variation in biophysical and biogeochemical parameters of tundra vegetation in climate and ecosystems models applied to the North Slope of Alaska.
3.2. Insight From Studies Focused on Other Regions of the Circumpolar Arctic
 The timing of the transition from the snow-covered period to the snow-free period in tundra ecosystems of the Arctic is critical to understanding energy balance, as albedo decreases substantially during this transition and energy input is quite high. There is a great deal of uncertainty in the representation of this transition in land-surface models, and the snowmelt transition can be biased to occur a month early to a month late depending on the particular land-surface model [Lynch et al., 1998]. Much of this uncertainty is associated with an incomplete understanding of winter and spring ablation of snow. Processes related to winter ablation of snow have been studied in the Flux Study and in the ATLAS Project [Holmgren et al., 1998; Sturm et al. 1997, 2001b, 2001c; Sturm and Holmgren, 1998; Liston et al., 2002]. Processes related to spring ablation have also received attention [Liston, 1995; McNamara et al., 1999]. The study by Boike et al.  at a continuous permafrost site on Spitsbergen is complementary to studies of snow processes that have been conducted by the Flux Study and the ATLAS Project. In the Spitsbergen study, an energy balance model was applied to estimate atmospheric, ground heat and snow heat fluxes for snow covered periods from autumn 1998 to winter 2000. The analysis identified that controls over snow ablation could be attributed to different processes in winter in comparison to spring, with sensible heat and rain primarily responsible for winter ablation of snow, while net radiation was primarily responsible for ablation during the spring. The analysis also suggests that the ground heat flux may be an important energy sink during winter. The importance of winter rain in the study by Boike et al.  represents a key contrast with processes that are responsible for winter and spring ablation of snow in tundra ecosystems of the North American Arctic, but could have increased importance under some future climate scenarios.
 The duration of snowmelt is a crucial period in the annual water budget of arctic terrestrial ecosystems as it can account for up to 80% of annual runoff. In spring 2000, ATLAS conducted a regional-scale exercise to obtain detailed ground-based observations of snow conditions and meteorology during the snowmelt period in eleven sites across Alaska and northern Canada (Atqasuk, Barrow, Caribou-Poker Creeks, Council, Imnaviat Creek, Ivotuk, Prudhoe Bay, Sagwon Bluffs, Franklin Bluffs, Quartz Creek, and Resolute in Canada) (Hinzman et al., unpublished data). The study by Nolan et al.  complements the ATLAS snowmelt intersite comparison by combining the use of SAR and Landsat imagery to analyze the hydrological dynamics from 1998 through 2000 of Lake El'gygytgyn, a lake in Siberia with no outlet that was formed by the impact of a meteor several million years ago. The study uses the remote sensing analyses to validate a lake-ice computer model that will then be used to extend understanding of hydrological dynamics for time periods without remote sensing data. A sediment core containing a 300,000-year record has been obtained from Lake El'gygytgyn, and the results of the lake-ice model will be used to interpret information contained in the sediment core to help reconstruct climate over the history of the lake.
 Freshwater runoff into the Arctic Ocean can influence its salinity and sea-ice dynamics [McDonald et al., 1999; Steele and Boyd, 1998], which have the potential to affect the global thermohaline circulation [Forman et al., 2000]. As climate warms, it is not clear how the dynamics of freshwater inputs into the Arctic Ocean will be affected. The study by Serreze et al.  analyzes climatic control over spatial variability in runoff of the four largest rivers (Ob, Yenisey, Lena, and Mackenzie) draining into the Arctic Ocean from 1960 onward. Cold season runoff has increased through time in both the Yenisey and Lena. This pattern is most pronounced in the Yenisey, where runoff has increased sharply in the spring, decreased in the summer, but has increased for the year as a whole. While the mechanisms responsible for this pattern are not completely clear, the patterns are linked to higher air temperatures, increased winter precipitation, and strong summer drying. It is possible that the changes in runoff patterns for the Yenisey and Lena are associated with changes in active layer thickness and the thawing of permafrost.
 Approximately 40% of tundra in the Arctic occurs in the Canadian Arctic, but the structure and function of this area of the Arctic has been poorly represented in large-scale climate and ecosystem models. The study by Gould et al.  has developed spatial data sets of dominant vegetation types, plant functional types, horizontal vegetation cover, aboveground plant biomass, and above and below ground annual net primary production for Canada north of the northern limit of trees. The study indicates that nearly 90% of the biomass and net primary production is concentrated in the Low Arctic, which is approximately 50% of the tundra area in the Canadian Arctic. In a similar analysis, the study of Walker et al.  applied their relationships between phytomass and summer temperature to the circumpolar Arctic, with the result that 60% of the above ground phytomass is concentrated in the Low Arctic. The spatial data sets developed by Gould et al.  should be useful for specifying the land surface and for evaluating simulations of ecosystem properties of the Canadian Arctic by large-scale climate and ecosystem models.