A CH4 emission estimate for the Kuparuk River basin, Alaska

. Integrated annual methane fluxes measured from 1994 to 1996 at sites representing specific tundra vegetation and land cover types were weighted areally using a vegetation map [Auerbach et al., 199 7] for the Kuparuk River basin and subareas. Wetland and open water CHn emissions dominate the Kuparuk River basin emission estimate. Areal weighting of site fluxes resulted in a regional CHn emission estimate of 2.09 x 10 (cid:127)ø g CH(cid:127) yr '(cid:127) for the Kupamk River basin. The global CH4 emission obtained by extending areally weighted annual fluxes from this study to global tundra area (7.34 x 10 (cid:127)2 m 2) is 5.83 Tg CH4 yr '(cid:127). This is about 15% of the Fung et al. [1991] atmospheric tracer model estimate and indicates that the vegetation distribution of the Kuparuk River Basin is not typical of the entire Arctic. Reconciling results from atmospheric tracer model estimates and areally weighted field flux measurements will require accurate high-resolution circumpolar estimates of wetland and open water areas and fluxes.


Introduction
The atmospheric concentration of methane (CH4), a wellknown radiatively important trace gas, is increasing at a rate of about 1% per year [Dlugokencky, 1994]. Efforts are under way to understand the global CH4 budget as well as the direct and indirect effects of the increase on climate. Methane emission from high-latitude wetlands is an important term in the global budget, and considerable effort has been directed over the last decade toward obtaining CH4 flux measurements in a range of locations ; Bartlett and Harris& 1993;Reeburgh et al., 1994;Vourlitis and Oechel, 1997]. Scaling site level flux measurements up to regional and global scales is an essential part of constructing global and regional material budgets that can be used as a basis for evaluating the effects of climate change. Because of spatial and temporal variability in flux measurements, as well as spatial variability in site types, scaling is difficult [Matson et al., 1989] and is the subject of active investigation using several approaches.
One approach to obtaining emission estimates is to estimate areal coverage of various site types (extensive variable) and multiply by measured fluxes (intensive variable). Vegetation is a measurements were froln a single Swedish study conducted over a decade earlier [Svensson, 1973].  used a 4 year time series of CH4 fluxes from permanent sites, but their areal weighting by vegetation cover types was based on literature values, not actual measurements. The flux transect study of Whalen and Reeburgh [1990] involved seasonal CH4 flux measurements at fixed intervals along the Trans-Alaska Pipeline haul road but required assumptions about the duration of the emission season. Many of the available CH4 flux measurements, which are reviewed by Bartlett and Harriss [1993], have resulted from short-term campaigns that frequently span only a portion of the growing season. Winter flux measurements are rare [Whalen and Reeburgh, 1988;Dise, 1992]. There is a strong North American bias in CI-I4 flux data sets, and additional transect measurements and long-term observations similar to those by Chrtstensen et al. [ 1995] and Panikov et al. [ 1993] are needed.
Methane flux measurements can be made at scales larger than chambers with aircraft boundary layer measurements (100 •km) or micrometeorological measurements using towers (100 m). 1994]. This improved agreement during NOWES/ABLE 3B can be attributed to three factors: 1) the dominant CH4 source was pools on the peatland, which were well resolved by rexnote sensing, 2) overall fluxes were low and site emissions were similar, and 3) peat temperature was the dominant control on emission (N. T. Roulet, personal communication, 1998). These studies show that although time consuming and tedious, static chamber measurements of CH4 flUX are equivalent and comparable to those obtained by micrometeorological and aircraft boundary layer measurements.
Methane emission is the difference between CH4 production in anoxic soil zones and oxidation, which occurs in floodwaters, adjacent to the water table, and in the rhizosphere Whalen et al., , 1995. Factors known to be important in CH4 emission include temperature, moisture content or water table level, and substrate availability. Vascular transport of subsurface CU4 by plants is believed to largely bypass this oxidizing zone, so wetlands populated by vascular plants have higher fluxes [King et al., this issue]. While a great deal of information is available on the different factors which influence CH4 flux from natural sites, no single factor can explain all of the variability. Relationships between soil temperature (or any single variable) and CH4 emission are site specific and are of little value as general predictors. Parameters that integrate conditions influencing flux appear to be the best predictors over the emission period . Process-based models have been introduced recently as a means of overcoming the problems of temporal and spatial variability and limited flux data. Development of general process-based models has ranged from models exploiting the relationship between primary production and CH4 flux [Aselmann and Crutzen, 1989;Whiting and Chanton, 1993] to application of ecosystem models with heterotrophic respiration terms modified to include CH4 emission [Christensen et al., 1996]. Recent process-based models for wetlands, which have successfully modeled seasonal cycles of CH4 emission, are the soil climate model of Frolking and Crill [1994], the primary production/soil organic matter decomposition model of Cao et al. [1996], and the water

CH4 emission sites were thoroughly waterlogged and inundated in many cases throughout the flux season. Observations at a water table manipulation experiment site at Toolik Lake showed no relationship between rainfall and CI-I4 flux for these sites. On the basis of the small contribution of winter emissions reported by Whalen and Reeburgh [1988] andDise [1992]
, we assme that winter CI• emissions from the study area are zero. End points (zero CH4 flux) of the CH4 emission season were estimated using soil temperatures. The beginning of the CH4 emission season was taken as the date at which the soil temperature at 10 cm depth rose above 0øC. Similarly, the end of the emission season was taken as the first day the soil temperattire at 10 cm was less than 0øC. The observed fluxes were linearly extrapolated to zero at these dates and were integrated using the trapezoidal role to obtain annual CH4 emission.
Methane fluxes from lakes were calculated using the stagnant film method [Kling et al., 1992], which involves estimating the surface film thickness from wind speed and the flux from the film thickness and the water-air concentration difference. Methane fluxes from streams were estimated using measured concentrations and evasion coefficients determined by addition of dissolved SF6 and a conservative tracer (Rhodamine or NaBr) [Kling et al., 1995]. Integrated annual CH4 fluxes from similar site types were averaged and are presented in Table 2. Table 2 Table 2. The CH4 emission from each land cover category was summed to obtain an estimate of emission from the total map area. The land cover class areas and calculated annual CH4 emission are presented in Table 3. Figure 1 presents nested pie charts showing land cover distribution (inner pie) and annual CI-h emission (outer pie) for each of the above map subareas and the two flight paths.

Fluxes
Integrated CI-h flux data for each site for the 1994, 1995, and 1996 field seasons are presented in Table 1

Uncertainties
It is difficult to evaluate errors in a map like Plate 1. User and producer errors associated with land cover classes were evaluated by Muller et al. [ 1998] and suggest a map accuracy of about 85%. Areas like barrens and water have distinctive spectral signatures, so their areas can be evaluated quite accurately.
However, it is difficult to evaluate the water status of areas dominated by sedges (acidic, nonacidic, and wet tundra), and some moist tundra could be wetlands. Results of the land cover map validation [Muller et al., 1998] suggest that wetlands were from site measurements. The study, which was conducted in parallel with the present work and involves many of the same CH4 flux data, focused on three images of the Toolik Lake area, one derived from a SPOT (Syst/•me pour l'Observation de la Terre) multispectral image, an ERS-1 (European Remote Sensing satellite) SAR (synthetic aperture radar) image, and a digital elevation model based on aerial photographs. The first two images were resampled to have 60 m pixels. A slope image with 60 m grid cells was calculated from the digital elevation model. Methane fluxes were measured in combination with soil moisture, temperature, pH, aboveground biomass, slope, and inundation by water. Using a regression tree approach, CI-h emission was explicitly linked to environmental conditions rather than vegetation types. However, the environmental conditions that defined the terminal nodes in the regression analysis were also related to vegetation type. The weighted CH4 emission rate for the Toolik Lake area was 750 mg CH4 m '2 yr •, which compares reasonably with the 645 mg CH4 m '2 yr • calculated for the Toolik Lake area in this study ( Table 3)