Carbon flux measurements using the eddy covariance method rely on several assumptions, including reasonably flat terrain and homogeneous vegetation cover. An increasing number of flux sites are located over partially or completely inhomogeneous areas, but the implication of such inhomogeneities on carbon budgets, and particularly the influence of year-to-year variations in wind patterns on annual budgets, remains unclear. Moreover, directional homogeneity of climatic drivers of carbon fluxes is often assumed, although climatic variables vary with wind direction at many locations. In this study, we examined the directional flux characteristics, incorporating the combined effects of variable surface characteristics and climatic drivers on the annual carbon budgets of an evergreen forest. Our study area was characterized by moderate variation in surface characteristics (leaf area index: 1.5–2; topographic wetness index: 4–16), and significant variation in the key drivers of carbon fluxes with wind direction (such as temperature, VPD and turbulence). Interactions among surface characteristics and climatic variables resulted in carbon uptake ‘hotspots’. These localized hotspots influenced mean CO2 fluxes from several wind directions, and were most distinctive during the summer months. Hotspot contributions to yearly budgets varied from year to year, depending on prevailing weather conditions. Consequently, directional variations in flux characteristics affected quarterly estimates of carbon budgets by up to 22%, and annual budgets by up to 25%. We present a procedure to quantify and adjust for the effects of year-to-year variations in directional flux characteristics on interannual comparisons of carbon budgets. Any remaining differences in budgets (after the adjustment) can then be linked more accurately to variations in ecophysiological drivers. Our study clearly highlights that directional variations in flux characteristics can have a significant influence on annual carbon budgets, and that these should be accounted for in interannual comparisons.