The EisKlass2 project divides into 8 different topical work packages:
WP1: SAR image ice classification
- Contributors: DLR
- Goals:
-Improvement of existing SAR classification algorithms
-Compensation for incident angle dependencies
-Smooth mosaicing of multiple SAR scenes
-Adaption of SAR classification for a fusion with optical data - Delivery: June 2022
WP2: Optical image ice classification
- Contributors: König & Partner
- Goals:
-Improvement of optical/thermal image ice classification
-Improvement of the cloud mask
-Creation of a large classified Sentinel-3 data set - Delivery: June 2022
WP3: Fusioned image classification SAR/optical
- Contributors: DLR, König & Partner
- Goals:
-Common geometry of Sentinel-1 and Sentinel-3 images
-Creation of training data sets
-Training of the combined AI - Delivery: December 2022
WP4: Operationalization
- Contributors: DLR
- Goals:
-Reduction of computing time
-Set-up of a cloud based processing framework
-Implementation of an interface to acess operational data
-Running of a test service - Delivery: June 2023
WP5: Validation
- Contributors: DLR, König & Partner, OAsys
- Goals:
-Comparison with official Ice Charts
-Comparison with in-situ observations
-Comparison with Ensemble of sea ice-ocean models
-Error estimation for the automated ice classification - Delivery: June 2022
WP6: Model-Based Product Evaluation
- Contributors: OAsys, The Inversion Lab
- Goals:
-Quantification of the impact of improved classification on sea-ice forecasts - Delivery: June 2023
WP7: User Interaction
- Contributors: Drift+Noise Polar Services
- Goals:
-Integration of classified ice images in the working process during two demo cases.
-Collection and analysis of user feedback
-Anticipated is a research and a commercial case. - End Date: June 2023
WP8: User Client
- Contributors: Drift+Noise Polar Services
- Goals:
-Integration of the classified ice images into the backend of the IcySea web service.
-Development of the IcySea frontend to display and deliver classified ice images to the end user.
-Collection and analysis of user feedback - End Date: June 2022