Mapping of Lake and River Ice in Northern Europe using Envisat ASAR and Radarsat-2 Data
Hindberg, Heidi; Malnes, Eirik

Synthetic aperture radar (SAR) data is useful for detection of lake ice at high latitudes, in particular during the polar night period and during cloud cover when optical sensors are useless. The lake ice observations can be assimilated in numerical weather prediction models, regional climate models or hydrological modeling as parameterizations and will increase the accuracy of such models. We have developed processing chains for lake ice retrieval based on co-polarized Envisat ASAR wide-swath data. SAR data has good temporal coverage, but due to overlapping backscatter signatures between ice and waves on the lake surface there is a higher probability for misclassification. Also, since the backscatter from water varies with the state of the lake surface, determining which classes represent water and which classes represent ice in a segmented product is not trivial. Up until now, MODIS snow cover fraction products have been used as an auxiliary data source to label the classes. However, this method has problems related to poor coverage from MODIS (mostly caused by cloud cover) and the fact that snow free ice often is misclassified as open water in the MODIS product. A validation effort comparing results over Lake Päijänne with in situ data showed that the classification method requires a manual post-processing selection step to remove obvious erroneous results, which is both time-consuming and subjective. In upcoming SAR sensors such as Sentinel-1, many of the problems faced using single polarization data will be solved with the use of dual-polarization data. It is well known that cross-polarization SAR data is less sensitive to waves on the water, resulting in ambiguous response from ice and water. By utilizing cross-polarization data we increase the seperability between the backscatter from water and ice. Since the backscatter from water always is lower that the backscatter from ice in the cross-polarization channel, the labeling of the classes will no longer require auxiliary data. In this study we use Radarsat-2 dual-polarization data over lakes in Norway and Finland, to investigate how dual-polarization data can improve the lake ice classification. Radarsat-2 data also has increased resolution compared to Envisat ASAR, allowing us to apply the method to smaller lakes and even large rivers. Application of dual-pol Radarsat-2 data for lake and river ice classification is comparable to Sentinel-1 IWS data, and the results of this study hence provide insight into operational use of Sentinel-1 for lake and river ice classification.