Systematic bias of average winter-time land surface temperatures inferred from MODIS at a site on Svalbard, Norway

• Published in: Remote Sensing of Environment
• Main Authors: Westermann, Sebastian, Langer, Moritz, Boike, Julia
• Format: Article in Journal/Newspaper
• Language: unknown
• Published: ELSEVIER SCIENCE INC 2011
• Subjects: Arctic; Svalbard; Norway; Climate change
• Online Access: https://epic.awi.de/id/eprint/25719/; https://epic.awi.de/id/eprint/25719/1/Westermann%2B11d.pdf; http://www.sciencedirect.com/science/article/pii/S0034425711003828; https://hdl.handle.net/10013/epic.38719; https://hdl.handle.net/10013/epic.38719.d001

Thermal remote sensing can quantify climate change in the Arctic, where ground-based measurements continue to be rare. The land surface temperature (LST) is accessible on the pan-arctic scale through a number of remote sensing platforms, such as the “Moderate Resolution Imaging Spectrometer” (MODIS). This study compares remotely sensed LST from MODIS to ground-based point measurements of the snow surface temperature on Svalbard for seven consecutive winters, thus covering more than half of the winter seasons in the operation period of MODIS Terra and Aqua. We find a systematic negative bias of the average winter surface temperature computed from single LST measurements between 1.5 and 6 K, with a mean bias of 3 K. The bias consistently occurs both for the MODIS L2 and for the daily and eight-day MODIS L3 products, which is explained by two reasons: i) During winter on Svalbard, cold surface temperatures are associated with clear-sky conditions, while warm surface temperatures typically occur during overcast periods. This leads to an overrepresentation of cold temperature in averages computed from remotely sensed LST measurements. ii) The MODIS cloud detection scheme fails to recognize some cloud-covered or partially cloud-covered situations, thus leading to admixing of colder cloud top temperatures. Both effects contribute equally to the total average bias accumulated over the winter season, with effect (i) dominating in some winters, while the observed bias can be fully explained by (ii) in other winters.