Extrapolating biosphere-atmosphere CO2 flux observations to larger scales in space, part of the so-called "upscaling" problem, is a central challenge for surface-atmosphere exchange research. Upscaling CO2 flux in tundra is complicated by the pronounced spatial variability of vegetation cover. We demonstrate that a simple model based on chamber observations with a pan-Arctic parameterization accurately describes up to 75% of the observed temporal variability of eddy covariance?measured net ecosystem exchange (NEE) during the growing season in an Abisko, Sweden, subarctic tundra ecosystem, and differed from NEE observations by less than 4% for the month of June. These results contrast with previous studies that found a 60% discrepancy between upscaled chamber and eddy covariance NEE sums. Sampling an aircraft-measured normalized difference vegetation index (NDVI) map for leaf area index (L) estimates using a dynamic flux footprint model explained less of the variability of NEE across the late June to mid-September period, but resulted in a lower root mean squared error and better replicated large flux events. Findings suggest that ecosystem structure via L is a critical input for modeling CO2 flux in tundra during the growing season. Future research should focus on quantifying microclimate, namely photosynthetically active radiation and air temperature, as well as ecosystem structure via L, to accurately model growing season tundra CO2 flux at chamber and plot scales.