Study On Parameter Retrieval And Type Recognition Of Sea Ice Using Wide-swath And Multi-polarization SAR Data

As part of the cryosphere,sea ice plays an important role in the climate change,thus ice parameters are crucial to the climate forecast.Nowadays,sea ice in the Arctic is enduring significant changes,in both ice extent and ice thickness.Significant decrease in the sea ice extent makes the Arctic shipping active in the past decades and possibly increasingly active in the future,thus ice parameters are useful for navigation safety.Therefore,the ice parameter retrieval of high accuracy is necessary for both the scicentific research and the living.Sea ice parameters include sea ice concentration,ice thickness,as well as ice motion velocity.Remote sensing provides good access to the sea ice parameter retrieval,especially in polar regions.For example,ice concentration products based on radiometers are popular now.However,their low spatial resolution,from several kilometers to tens of kilometers,can not meet the requirements in the climate forcast and shipping activities.With spatial resolution of tens of meters,synthetic aperture radar(SAR)is one alternative for sea ice monitoring.Especially,the development of the advanced SAR technology such as digital beamforming(DBF)and multiple-input and multiple-output(MIMO),makes the occurrence of SAR data with both wide swath and high spatial resolution.This has attracted much attention from people.Wide-swath SAR is the primary mode for sea ice monitoring.Therefore,the sea ice parameter retrieval based on wide-swath SAR data is a hot topic,and the relevant study will benefit the climate study and shipping activities in polar regions.In this dissertation,wide-swath SAR data from multiple spaceborne SAR sensors are used,such as ASAR,RADARSAT-2,TerraSAR-X and Sentinel-1A.They are used for ice concentration retrieval,ice type classification,ice velocity retrieval and ice flux estimation.These paramters are all important to estimate the mass balance of sea ice.Accurate sea ice parameters are significantly usefull in sea ice thermaldynamics and dynamics study,as well as climate change study.The main contents of this dissertation are listed below.First,an automatic ice concentration retrieval algorithm based on wide-swath SAR data is proposed.This algorithm is developed to solve the problem that the large incidence angle range causes the overlap in backscatters between ice and water.For the first time,the incidence angle dependence,along with backscatters and the dissimilarity is exploited for the ice edge detection and the lead detection,and further for the sea ice concentration retrieval.This algorithm is carried out completely automated.Besides,the validation proves that it could be applied in different ice regions,based on HH or VV polarized data from different SAR sensors.Second,a method of ice type classification for removing effects of scalloping and inter-scan banding on sea ice classification is proposed.Textural features are usefull for ice type classification.However,it is found that they could amplify these two types of stripes,which makes it difficult to achieve a good classification result.In this method,the Kalman filter,sea ice concentration,gray level co-occurrence matrices and support vector machine are combined to solve this problem.The result shows that this method is effective in removing effects of scalloping and inter-scan banding on sea ice classification based on VV polarized TerraSAR-X data.Third,a method of ice type classification based on sea ice concentration is proposed.Sea ice concentration from our algorithm is introduced here to discriminate between ice types with similar backscatters in HH and HV polarization.The result shows that this method could improve the classification accuracy,especially the pancake ice.Besides,an application case of ice type classifcation based on sequential SAR data is presented.In this case,ice type information and ice drift velocity are extracted based on SAR data,and wind and currents information are obtained from reanalysis data.Using all these information,the ice dynamic process is interpreted when Xuelong were trapped in the Antarctic in 2014.This study promotes the application of SAR data in the interpretation of ice dynamic process.Fourth,a method of ice area flux estimation based on SAR data has been proposed.It is aimed to solve the problem that the data used in the current method have low spatial resolution.Then the Fram Strait is taken as an example here to show the results using HH polarized Sentinel-1A data.Area flux of Fram Strait in February,April,June,August,October and December of 2016 are analyzed.Methods proposed in this dissertation could provide support for the retrieval of high-resolution ice parameter as well as the application of SAR data in ice dynamics,further promoting the development of the sea ice study,the SAR methodology and the climate change study.