Interpretation Of Sea Ice Geophysical Parameters From SAR In Selected Regions Of The Antarctic And The Arctic

Sea ice is an important indicator of global climate change,impacts,e.g.,the heat flux between ocean and atmosphere,and also feedbacks on the ecosystem.Remote sensing observations reveal that Antarctic and the Arctic sea ice extent have contrasting trends during the past 40 years.The different spatial and temporal variability of sea ice cover in the Antarctic compared to the Arctic results from different sea ice evolution processes.The Antarctic ice area has slightly increased while there is a dramatic decrease in the Arctic under different thermodynamic and dynamic processes.The Antarctic continent is surrounded by sea ice and open water,while the land masses extrude into the Southern Ocean.Arctic sea ice generally grows under quiescent ocean conditions since it is surrounded by land.In contrast,Antarctic sea ice grows under turbulent ocean surface conditions,stimulating deformations into rafting and ridging,and ice motion in response to external oceanic and atmospheric forcing.Another different condition in the Antarctic is the seawater flooding of the snow ice interface when the load of snow cover is heavy enough to depress the surface of the ice floes below the water line.This process is ubiquitous in the Antarctic.The Arctic,on the other hand,was dominated by the thick multiyear ice,which has survived at least one melt season and less prone to flooding.Ice deformation and moistened snow pack subject to flooding of seawater can fundamentally influence the sea-ice scattering characteristics in the microwave range.In the Antarctic,surface melting is not as common as in the Arctic and rarely associated with the presence of melt water on the surface of the ice,i.e.,there are very seldom melt ponds in the Antarctic.Additionally,the ablation season duration in Antarctica is shorter than in the Arctic.Different ambient sea ice conditions with respect to diverse dynamic and thermodynamic interaction between the ocean and atmosphere differs in sea ice physical and microwave properties.Sea ice parameter interpretation from Synthetic Aperture Radar(SAR)has become a hot topic in the sea ice remote sensing community.The mentioned factors make sea ice microwave backscatter complex and it is difficult to derive sea ice geophysical paramters from SAR observations,which is mainly influenced by the roughness of the reflecting surface on scales of the radar wavelength.In order to determine the deformation conditions,sea ice geophysical paramters interpretation including sea ice concentration retrieval,sea ice classification and sea ice ridges detection is constructed from coarse to fine scale in this disertation.The main content is as follows:(1)A summary about principles of sea ice geophysical parameter interpretation is presentd.Sea ice concentration retrieval,classification and deformation feature principles are summarized using different remote sensing sensors:optical,thermal and microwave remote sensing sensors including a description of their advantages and disadvantages.(2)A novel method for sea ice concentration retrieval by combining data from multiple microwave sensors is introduced.Dynamic tie points have been designed for reducing the overestimation error along ice edge and underestimation in some interior areas.The new sea ice concentration product is utilized as input for a fusion method with SAR ice water classification results to reduce the estimation uncertainty.For the ice water classification,statistical distributions have been investigated by exploiting the conditional random fields(CRF)as an additional probability potential at the pixel level.Then,the geolocation registration of sea ice concentration product and ice water classification result has been developed under the maximum a posteriori(MAP)inferring procedures to improve the sea ice concentration retrieval accuracy.RADARSAT-2,Envisat ASAR and Sentinel-1A SAR datasets are used to evaluate the performance of the proposed algorithm.Results show that the proposed algorithm improves sea ice concentration estimates especially in the thin ice zone.(3)During the spring-to-summer transition,ice deformation makes the traditional sea ice classification method deficient in mapping within the thin ice zone.An improved Antarctic sea ice classification algorithms from RADARSAT-2 dual-polarization SAR images are developed using the CRF approach by including multiple features from sea ice concentration,gray-level cooccurrence matrix textures,polarization ratio,backscatter coefficients,and intensity.Five RADARSAT-2 SAR scenes coincident with CHINARE-30th cruise were collected for ice classification into categories such as open water,thin ice,smooth first year ice,deformed first year ice,and old ice.Strategies including statistical distribution and region connection,multiple features and support vector machine(SVM)integrated into the CRF model are proposed to describe the contextual relationships between sea-ice-types.Comparision between the pixel labels from the initial SVM classification result and the segments is used to update the CRF potential to improve the sea ice classification.The best results are obtained with the SVM-CRF algorithm for sea ice classification.For the comparison,three scenes from the Indian Ocean sector and two scenes from Pacific Ocean sectors of Antarctica including medium-resolution with a pixel spacing of 50 m and fine resolution dual-polarization SAR imagery with a pixel spacing of 6.25 m have been used on the proposed five strategies including FEA-SVM,STA-CRF,REG-CRF,FEA-CRF and SVM-CRF.The effects of deformation,rafting,and ridging during the spring-to-summer transition period were overwhelmed by the spatial and contextual CRF models in combination with the rich features extracted for sea ice classification.Results indicate that the SVM-CRF approach has the capacity for improving sea ice classification.(4)A collaborative interpretation of sea ice parameters combining airborne and satellite based data is conducted for detecting sea ice deformation features.So far sea ice ridges have not been determined on large scale from medium resolution wide swath SAR imagery.Pressure ridges can be observed from high resolution SAR imagery,which faciliates the identification of ridges,mainly due to the increased volume scattering inside.A structure tensor is used to describe the ridge features in SAR imagery with intense speckle noise.The ridge extraction algorithm is based on the hypothesis that the bright pixels are ridges with curvilinear shapes.With a Gaussian filter,the bright curvilinear shapes in SAR imagery are enhanced.The dominant orientation of the pixel representing the ridge is determined from the tensor.The structure tensor is applied to extract curvilinear features as ridge with a convolution method.In order to evaluate the effect of spatial resolution on ridge detection,airborne E-SAR data sets have been used to simulate different spatial resolutions from fine to coarse.We follow the simulation procedures presented in the ICESAR 2007 campaign report for satellite-based SAR data simulated from airborne SAR.To investigate how sea ice surface deformation can be observed by SAR,the airborne measurement campaign ICESAR 2007 has been conducted.C and L band E-SAR data from ICESAR 2007 is used to validate the performance of the proposed algorithm on extracting sea ice ridges.The major objective of the E-SAR simulation is to identify ice surface characteristics such as ridges and leads in radar imagery acquired with the airborne multi-frequency E-SAR instrument that resembles the Sentinel-1 SAR satellite system parameters.After a new retrieval algorithm for ridges was developed based on the airborne,higher resolution E-SAR dataset,the algorithm has been transferred to Sentinel-1A satellite data.Reuslts show that the proposed algorithm can effectively detect ridges from multiple resolution SAR imagery.The algorithm is suitable for future to determine the temporal and spatial distribution of ridges and leads on hemispheric scales.