Data fusion of Radarsat and MODIS images for Arctic sea ice mapping

There are more and more demands on the Arctic sea ice mapping, tracking and forecasting. This is due to increasing activities in marine transportation and natural resources exploitation in the Arctic area, and also due to pressing needs of understanding climate change impact on the Arctic environment. For either real-time or long-term monitoring of vast area, satellite data play a unique and effective role. With advanced development in instrument technology and data communication, multiple sources and large volume of data are becoming available now and increasingly in the future. There are two relevant issues: one is how to take advantages of data from different sensors; secondly is how to fully automate data processing and to efficiently produce useful products. In this abstract, an optimal data fusion algorithm will be developed and it will be operationally implemented with Radarsat-2 and MODIS images at this time. These two sources of data have their specific features and limitations in terms of sea ice observations. The SAR image has higher spatial resolution and has no influence from weather or atmospheric conditions and works day and night. Normally, there is a sufficient brightness contrast between sea ice and open water. However, with high sea states and ice melt in springtime, this contrast can be washed out. MODIS data includes multiple visible and infrared bands that can easily distinguish between sea ice and open water because the latter is always darker than the former. MODIS also has much wider scan swath and higher revisit frequency. However, its spatial resolution is relatively coarse and the presence of cloud or haze can block the surface radiance and impedes MODIS from sea ice imaging. The purpose of fusion of these two data is to efficiently take advantage of the benefits of each dataset. In order to obtain large area results, a multi-scene Radarsat mosaic image and multi-day MODIS composite image for identical coverage of the Canadian Arctic are first generated. Various fusion schemes will then be used to be able to optimally present important ice features such as ice concentration, ice extent and ice type. The results will be evaluated with the Canadian Ice Service ice analysis data. Finally, a data fusion product will be generated on a regular basis to help improving the quality of public sea ice service.