| Summary: | The NASA Scatterometer (NSCAT) is designed to measure wind vectors over oceans; but there are land and ice applications as well. This paper presents recent work to develop sea ice classification algorithms based on neural network technology. Multi-Layer Perceptron (MLP) neural networks are trained using multi-azimuth, dual-linear polarized normalized radar cross section measurements from Ku-band NSCAT. Algorithms are developed to classify the first-year sea ice edge in both the Arctic and Antarctic. For the Arctic region, after classifying the ice boundary, both first-year and multi-year classifications are made and expressed as multi-year fraction. NSCAT results are compared with corresponding ice products from the passive microwave Special Sensor Microwave Imager. Results show the utility of satellite scatterometers and neural network techniques for classifying sea ice in near-real time and independently of other sensors.
|
|---|