全球变化知识资源中心

The Arctic has experienced the most rapid warming in the world in recent decades. Complex topography combines with low solar elevation to create distinct microclimates in Arctic regions, and for many applications such as ecological response and cryospheric change it is critical to obtain reliable temperature trends at the local scale. Due to lack of weather stations, satellite land surface temperature (LST) is increasingly important as a proxy for air temperature (T-air), but how accurately it can represent microclimates is unknown. For the first time, we compare 10 years (2007-2017) of T-air recorded over a dense network of 65 sites (25km(2)) around Kevo Subarctic Research Station in Finland with equivalent moderate resolution imaging spectroradiometer (MODIS) LST at 1 km resolution from MOD11A2/MYD11A2 8-day products. We assess whether LST can pick up the extreme local gradients in air temperature (>20 degrees C/km) caused by cold air drainage. Although there is a high correspondence between LST and T-air anomalies on a synoptic timescale, small-scale patterns in T-air (lapse rates, aspect contrasts) are not picked up by LST. Temperature gradients in T-air become positive (temperature inversions) in winter, and at night, but LST gradients show almost the reverse. Aspect contrasts in T-air peak in spring and autumn during the day, but LST shows biggest differences in the evening. Land cover has a large influence on LST, tundra heating up/cooling down more than birch or pine forest. The conflation between land cover and elevation means that differential land-cover response dominates the elevational LST signal. Contrasts between T-air and LST cannot be explained by the number of stations measuring T-air in a pixel, elevation error, timing differences or the frequency of cloud cover within the 8-day composite. Important features of the Arctic climate such as microscale cold air drainage are thus potentially obscured by land-cover effects.