Evaluation of pan-Arctic seasonal melt-freeze transitions retrieved from satellite radar, CMIP5 climate models, reanalyses, and observations

The seasonal melt-freeze transitions are fundamental features of the pan-Arctic climate system. Like sea ice, snow has a much higher albedo than the underlying surface and typically attenuates heat fluxes by 1-2 orders of magnitude, thereby having a major impact on the surface energy budget. Due to the vast expanses of snow and ice that undergo seasonal transitions – about 55 million km2 on the Northern Hemisphere – these transitions have a considerable impact on the region and the global climate. For example, an early sea-ice melt enables an additional absorption of insolation by open water during summer. The additional heat may foster a later freezeup and thinner ice in the fall, and has the potential to affect the atmospheric circulation considerably. On land, the transitions modify, e.g., timing and amount of river runoff, the surface albedo, and the growing season of vegetation and the resulting switch from a sink to a source of atmospheric carbon. It is therefore of interest to accurately and continuously estimate the timing of the seasonal transitions. However, this is severely hampered by the limited amount of surface observations. Both models and microwave instruments provide wide spatiotemporal coverage of the region, but still need to be evaluated against observations in order to ensure reliable results.

We map the timing of the seasonal melt-freeze transitions across the pan-Arctic region – land and sea ice north of 60°N – from microwave instrument QuikSCAT (1999-2009), eleven CMIP5 GCMs (1850-2100), and two reanalyses (1979-2011, 1871-2010). We evaluate these results against surface observations (1979-2009). In order to retrieve transitions from microwave measurements, we exploit the abrupt changes in the time series associated with major melt-freeze events using a sophisticated algorithm. A running median and a threshold estimates the major melt-freeze events in data from GCMs, reanalyses, and observations.

From the evaluation we conclude that the edge-detector algorithm applied to QuikSCAT measurements successfully captures the seasonal melt-freeze transitions; the bias from observations is typically a few days. ERA-Interim is the best performing dataset over the full region, closely followed by several GCMs. In general, the seasonal transitions are better represented over the landmass and old ice, than in the marginal seas where seasonal ice is abundant. According to the RCP8.5 projection during the 21st century, melt and freeze onset will be displaced by ~10 days over land, while over sea ice, the freezeup will be delayed by ~90 days as old ice retreats. This is a consequence of freezeup occurring 1-2 months later over seasonal ice than old ice, indicated by results from all datasets.

Seasonal melt-freeze transitions have been studied in the past, but the studies typically focus on one transition, domain, or data source. The variety of datasets we use provides spatiotemporal evolution serving as a foundation for hydrological applications: e.g., for improving the monitoring of melt-water runoff and river discharge in spring; to assess or complement operational snow and sea-ice cover products; and to study mechanisms that onset major melt and freeze events. Moreover, the results can be used for model development in this sensitive and important region. We currently work on prolonging our transition map from microwave measurements, employing the ASCAT instrument presently in orbit. Preliminary results are ready and look promising; we expect to have more developed results shortly. Additionally, we aim to further study the mechanisms in play, on a large scale.