To quantify the impact of surface albedo feedback (SAF) on unforced climate variations, a coupled ocean-atmosphere model (R15 resolution) is run in two configurations for 1000 years: One where sea ice and snow variability affect surface albedo (control), and one where sea ice and snow vary, but have no effect on surface albedo, which is prescribed to seasonally-varying climatology (fixed albedo).

In the northern hemisphere (NH), SAF has surprisingly little impact on the magnitude of local surface air temperature (SAT) variability during fall, winter, and summer. In the spring, SAF amplifies local SAT variability over continents by 20-40%. The reason for this strong seasonal dependence of the strength of SAF lies in the fact that surface albedo variations in the NH stem mainly from variations in continental snow cover. In the summertime, snow covers little area, so surface albedo fluctuations are small and SAF is correspondingly weak. In the fall and winter, the local snow budget is dominated by accumulation, which occurs when storms are randomly generated by the turbulent cold-season jet stream. The result is that local surface albedo fluctuations throughout the NH are largely uncorrelated with one another. Though these fluctuations may be large and have correspondingly large effects on the local top-of-the-atmosphere radiative balance, they tend to cancel one another out through enormous exchanges of heat from region to region, virtually eliminating any systematic SAF to local temperature variability. In the springtime, the snow budget is dominated almost everywhere by melting, a process closely tied to temperature in a way that facilitates positive SAF. If the springtime temperatures averaged over the NH are warmer than normal, this will lead to more snowmelt everywhere, decreasing surface albedo, increasing sunshine and raising temperatures further. In the southern hemisphere (SH) mid-latitudes, SAF behavior during the entire year is similar to its behavior during NH springtime: It amplifies local SAT variability by 20-60%. SH surface albedo variations stem mainly from variations in sea ice extent and thickness, rather than continental snow cover. Like snowmelt, both sea ice growth and decay are often linked to temperature anomalies in a way that facilitates positive SAF. Thus a positive SAF can occur whether the sea ice budget is dominated by growth (fall and winter) or decay (spring and summer).

To evaluate whether the model needs SAF to generate realistic SAT variability, the NH springtime variability is compared with the variability in the 50 year NCEP reanalysis record. If the raw NCEP data is used, the fixed albedo model has about half as much springtime variability averaged over the NH poleward of 30N as the observed record, while the control model has ¾ as much variability as observed. When the systematic trend in the observations is removed, the fixed albedo model has 70% of the observed levels of variability, while the control model agrees with the observations to within 5%. This suggests the model needs a positive SAF to simulate observed magnitude of springtime variability, and that the trend in the observations is not internally-generated.