Conference Agenda

With contributions from: Markku Similä, Xi Zhang, Bernat Martinez, Jungang Yang, Jacqueline Boutin, Jin Wang, and their co-authors.

The project is divided into three major parts: (1) development of an algorithm for the retrieval of sea surface salinity based on combined active/passive microwave imagers; (2) data validation and oceanic application of new satellite altimeters and SWIM (“Surface Waves Investigation and Monitoring“, which is a scanning radar altimeter); and (3) devising techniques for sea ice parameter extraction and sea ice monitoring using data from most advanced satellite sensors.

The recent studies in part (1) deal with the development, evaluation and improvement of the sea surface salinity (SSS) retrieval algorithm under unfavorable conditions. A method for considering the effect of rain, implemented by the Chinese team, is based on L-band CAP (Combined Active-Passive) observations. The L-band GMFs (Geophysical Model Functions) are developed and the radiation characteristic of the rough sea surface is analyzed for rain-free and rainy conditions, respectively. Using this approach, the effect of rain-induced roughness can be corrected when calculating the SSS.

The objective of part (2) is the identification of eddies and their spatial-temporal variation characteristics on the Chinese seas using radar altimeters. The recent work of the European team is structured as a three-step approach: (a) improvement of retracker algorithms for the European altimeter missions, (b) development of geophysical corrections adapted to coastal zones, in particular focussing on the wet troposphere, (c) computations of geostrophic currents and derivation of mesoscale eddies. The Chinese partners recently have been working on inter-comparisons of data from Sentinel-3 SRAL, HY-2A RA, and Jason-2. The accuracy of, e. g., significant wave height retrievals as a function of the distance from the coast is analysed. The new algorithms shall be evaluated and validated in the next phases of the project (2018-2019) using in-situ and other reference data.

The work on sea ice monitoring (part 3) has been focussed on (a) the use of radar altimetry and SAR images for calculating ice thickness and sea ice concentration, and (b) a feasibility study concerning the performance of single-pass interferometric SAR for the retrieval of sea ice surface topography. The Chinese team, with contributions from the European partners, developed a new algorithm for retrieving the sea ice freeboard based on a new approximation of the echo waveform. This method resulted in more accurate freeboard estimates. A European group evaluated the achievable accuracy of ice surface height variations for different satellite InSAR configurations and analysed Tandem-X data acquired over sea ice as an example. Another Sino-European team concentrated on the determination of sea ice concentration and thickness from RS-2 ScanSAR images obtained over the Bohai Sea in winter 2012/2013.

In this introductory talk we will provide a brief overview of the different activities, which then will be described in more detail in subsequent presentations.

ESA’s Sentinel-3 mission was successfully launched in February 2016. It carries a new altimeter-SAR Radar Altimeter (SRAL) which works on two modes of Low-Resolution Mode (LRM) and SAR mode. SRAL improves the along-track resolution (approximately 300m) in SAR mode using a delay/Doppler technique. SRAL is expected to provide even better measurements of sea surface height, significant wave height and wind speed than the conventional altimetry, especially in the coastal zone. In addition, Chinese first satellite altimeter-HY-2A Radar Altimeter (RA) has been in orbit more than 4 years, and Sentinel-3 SRAL and HY-2A RA are Simultaneous in orbit currently. In this study, sea surface height data of Sentinel-3 SRAL are analyzed by the comparison of the distribution and time series variability of SLA data between the different altimeters (Sentinel-3 SRAL, HY-2A RA, Jason-2) data and tide gauge data in the China Sea. Significant wave height data of SRAL are evaluated by the comparison with buoys data in the China Sea. The performance of SRAL observation over the coastal regions is analyzed. The accuracy and reliability of SRAL data are analyzed in the different locations with the different distances to the coast. Finally, the preliminary performance of Sentinel-3 SRAL and the evaluation of HY-2A RA in the China Sea are summarized.

The sea surface salinity (SSS) is one of the key parameter for scientific community to understanding the ocean better. Equipped with an L-band radiometer, SMOS and Aquarius provide an unprecedented SSS data set of the global oceans. Enormous efforts are devoted to the development, evaluation and improvement of the SSS retrieval algorithm especially under some unfavorable conditions, i.e., the rain. Rain drops induce freshening and roughness effect to the sea surface. The fact that both mechanisms causing TB to increase makes it challenging to retrieval sea surface salinity when it rains. This presentation describes a method to retrieval the SSS under the rainy conditions based on the L-band CAP (Combined Active-Passive) observations. The L-band GMFs (Geophysical Model Functions) are developed and the radiation characteristic of the rough sea surface is analyzed for the rain free and rainy conditions respectively. The excess emissivity of H/V polarization is found to rise as the rain rate increases with a slope of 0.0003/mm·h^-1, which means a 1-psu error at the rain rate of 10mm/h. The dependence of the sea surface emissivity (sensitive to both roughness and freshening) on the backscatter (only sensitive to roughness) is obtained and the rain-induced roughness effect is corrected. The bias of retrieved SSS shows no clear dependence on the rain rate and the standard deviation of SSS error is about 0.5psu. The above results confirm the feasibility of this new retrieval algorithm for the SSS remote sensing in the rainy weather with the CAP observations.

Scientists from the Chinese National Satellite Ocean Application Service (NSOAS) and the Finnish Meteorological Institute (FMI) have jointly studied two fundamental sea ice parameters, sea ice concentration (SIC) and sea ice thickness (SIT) in the Bohai Sea. The case study was motivated by the risks the sea ice can cause for off-shore activities. Although the ice season is relatively short, usually from December to March, it can cause damage even in mild winters. On average level-ice thickness varies from 20 to 40 cm depending on the weather conditions.

We targeted the winter of 2012–2013 due to the availability of RADARSAT-2 (RS-2) ScanSAR imagery over the Bohai Sea. Our approach was to utilize several satellite sensors to create SIC and SIT charts.

The data set at our disposal consisted of 11 RS-2 SAR HH/HV images, daily AMSR2 imagery, four MODIS images and 31 in-situ measurements collected from the oil field platforms during the period from 2 January to 19 February 2013.

Basis for the calculations was the segmentation of SAR images using the Iterated Conditional Mode optimization and the extraction of different segment-wise features from them. The following texture features turned out to be the most useful in this connection: entropy, HH-polarization spatial autocorrelation, two quantities describing the shape of the variogram and HH-polarized corner points. These features were utilized in the SAR based SIC estimation. The scaled entropy alone yielded the best results in the SIT estimation.

The selection of the features in the SIC estimates was performed automatically by searching through all feature combinations and minimizing the residual error sum. Two different methods were tested: a linear model and a nonlinear Multi-Layer Perceptron neural network. Both approaches provided good results for very low or very high ice concentrations. For this data set, the linear model performed better. As a reference data we have used the in-situ measurements, the MODIS based SIC charts and the optical HJ-1B image.

The total ice volume in the study area was estimated using the thermodynamic sea ice model HIGHTSI with atmospheric boundary layer forcing (the reanalyzed ECMWF data). The ice drift was taken into account using the AMSR2 based ASI SIC chart with a resolution of 3.125 km. The ASI SIC chart was adjusted for the Bohai Sea conditions by the University of Bremen. The spatial distribution of the HIGHTSI SIT model field was then improved using statistics calculated from the AMSR2 data. The result was interpolated into a 1 km spatial grid and used as the background field h_B for the SAR based SIT estimation.

The background field and the SAR features were combined using a liner model that included a modulation term between h_B and each feature. Also here the minimization of the residual sum was the objective criterion in the feature selection. Somewhat surprisingly only a single feature, the scaled entropy, gave the best estimation results. As a reference data we used the MODIS based ice thickness chart as well as the field measurements.

The approach applied in this study has potential to be implemented operationally for the Bohai Sea ice service. However, more in-situ measurements as well as tuning the estimation model parameters are necessary, because the current study covers only one winter season. The obtained results should be regarded as guidelines for further research

Altimeter waveform retracking correction is essential to get accurate sea ice freeboard. Some empirical fitting methods have been proposed, and these methods simulate the waveform by fitting to echo and empirically positioning the retracking point. In this paper, we develop a new method to retrack lead elevation by simulating the lead echo waveform using a composite cubic Bezier curve which is stable. The lead waveform is first divided into several segments, for each segment, the cubic Bezier curve is used to simulate the waveform. For ice, an improved offset center of gravity (OCOG) method is used. A new sea ice freeboard retrieval algorithm by using Cryosat-2 SAR-mode data is then performed combining the Bezier curve waveform fitting and an improved OCOG method.

The sea ice freeboard is calculated in following steps. The first step is to discriminate sea surface and the ice. The discrimination is based on the fact that the shape variation in echo waveform depending on whether the echo is dominated by specular reflections from leads, or by diffuses reflections from ice. The pulse peakiness (PP) and the stack standard deviation (SSD) parameters are used to discriminate the ice/water type. Returns from leads are identified by PP > 18 and a SSD<4, while echoes from ice floes are identified by PP < 9 and SSD > 4. Secondly, for echoes from leads, we use a more sophisticated Bezier curve to fit the echo to get retracking point, and the retracking point is set where rise reached 70% of the waveform maximum. For echoes from ice floes, a modified OCOG is used, and the retracking point is positioned at the point where the rise reaches 95% of the computed OCOG amplitude. Thirdly, to obtain an accurate freeboard, we obtain local sea level height as shown followed. We use a linear interpolation on the difference between lead elevations and mean sea surface height (MSS) for each Cryosat-2 track, and then yield the sea surface anomaly (SSA, the deviation between the actual sea surface elevation and the MSS). At last, the sea ice freeboard can be retrieved by sea ice surface elevations minus MSS and SSA.

We directly compare the freeboard retrievals by proposed method and L2I products of CryoSat-2 to IceBridge data. A mean bias of 1.3cm was found with the uncertainty of 0.025m which is a latitude-dependent gradient. For two IceBridge campaign periods in March 2015 and April 2016, mean differences of 1.57 and 1.09cm are for the freeboard retrievals by proposed method, while mean differences of 1.75 and 1.84cm are found when using Cryosat-2 L2I products. This suggests the proposed method is capable of reconciling the sea ice freeboard from radar altimetry data sets with a high accuracy.

Ocean wave is a kind of fluctuation caused by the wind in the ocean. The ocean waves are very complex phenomena, the study of the ocean waves is of great significance on the marine engineering, marine development, transportation and shipping, marine fishing and aquaculture and other activities. In-situ observations, satellite remote sensing and numerical models are the important means of obtaining ocean waves information. Especially, satellite remote sensing is the important method of obtaining global ocean waves information synchronously. In this study, the significant wave height (SWH) of ocean waves obtained by T/P, Jason-1/2, ENVISAT, Cryosat-2 and HY-2A satellite altimeter and ENVISAT ASAR wave spectrum are used to generate the global ocean wave remote sensing data of 2000~2015 with the spatial and temporal resolution of 0.25° and one day. Firstly, all satellite data are validated and corrected by the comparison between them and the National Data Buoy Center (NDBC) buoys data. Then global ocean wave remote sensing data are obtained by the inverse distance weighting method. Finally, the resulting data are evaluated by the buoys data.

In this paper, sea-ice radial drift is studied based on the single-look complex (SLC) data of Sentinel-1A/B SAR. The method for sea-ice drift retrieval is developed using Doppler shift. The estimated Doppler frequency and predicted geometric Doppler shift are calculated for each unit of the Doppler grid, then the Doppler centroid anomaly can be obtained. According to the analysis of the uncertainties from the orbit, attitude, antenna pattern, topography, etc., the error in the Doppler centroid anomaly can be eliminated as far as possible to acquire the standard Doppler centroid anomaly. Furthermore, the standard Doppler centroid anomaly is converted to sea-ice radial drift velocity. Finally, the developed method will be validated through the comparison with not only a conventional cross-correlation method, but also measurements from a drifting ice buoy.

HY-2A is the first satellite mission for the dynamic environmental parameters measurements of China which has been launched successfully on August 16th, 2011. The multi-bands scanning microwave radiometer (RM) is one of the key sensors onboard the HY-2A satellite with the primary objective of measuring SST, surface wind speed, water vapor and cloud liquid content from space. On the basis of HY-2A RM observations, the sensitivity of some brightness temperature (TB) channels to the rain rate and the wind speed are analyzed. Two TB combinations which show minor sensitivity to rain are obtained. Meanwhile, the sensitivity of the TB combination to the wind speed is even better to the original TB channel. Based on these TB combinations, a wind speed retrieval algorithm is developed and validated. The wind speed retrieval accuracy is better than 2 m/s for rainy conditions, which is evidently superior to the HY-2A RM standard product.

Sea ice is not only an important part of the global climate system, but also affects the development of oil ,gas and mineral resources in the Arctic. In addition, the sea ice can affect the perforation of the "Arctic Waterway" and cause harm to production activities such as sea-related production and maritime traffic. That monitor the sea ice drift and obtain accurate sea ice drift direction and speed is of great significance to the study of climate analysis, safe navigation of ships and management of offshore oil platform.

At present, all the study of sea ice drift monitoring is based on the intensity information of SAR satellite image. However, due to the influence of speckle noise, the Algorithms of ice motion monitoring could not work well. With the development of synthetic aperture radar (SAR) technology, the texture have been shown to greatly improve the classification ability of sea ice with SAR image. Therefore, the texture of sea ice SAR image is expected to improve the sea ice drift detection capability.

Based on the above, The Sentinel-1 SAR data were utilized to detect and analyze the strengthen ability of sea ice drift monitoring by using the SAR image texture. Eight kinds of texture features such as contrast, correlation, dissimilarity, entropy, mean were extracted from Sentinel-1 SAR data. A total of 66 sea ice samples which were divided into new ice, young ice and first year ice three types were selected for the experiment. First, normalized cross-correlation method is used to evaluate whether the texture is suitable for sea ice drift monitoring by comparing the following four results: match position is correct or not, the ratio of the maximum value and the second value of match results, the ratio of the maximum value and the mean value of match results, and the ratio of the maximum value and the triple variance of match results. In addition to that the effect of different sea ice types and resolution on the correctness of matching is analyzed from the results.

Then, the SURF method is used to evaluate the sea ice by comparing the following three results: the matching accuracy rate of from SAR image and eight texture, the correct matching logarithm of feature point, and the accuracy of correct matching feature pairs. The use of the SURF method for sea ice drift monitoring can produce many misaligned feature pairs, we use the following ways to determine the feature pairs is correct or not: The approximate velocity and direction of the sea ice drift are determined by artificial visual judgment, and then the velocity and direction calculated by the SURF feature point are considered correct which is within the error range of the approximate velocity and direction. The experimental results show that the texture feature is more suitable for the sea ice drift monitoring than the SAR image itself.

Ship surveillance plays an important role in maritime traffic control, shipping safety, fishery supervision, maritime accident rescue, and oceanic rights protection. Synthetic Aperture Radar (SAR) has been widely used in ship surveillance as it has a full day, all-weather imaging capabilities. A great deal of research has been done on vessel detection using SAR; however, it is still difficult to detect the "weak" target under terrible sea conditions. The "weak" target with low target-sea contrast (TSC) is easy to be lost, and the existence of strong sea clutter can cause false alarm target. Therefore, this paper mainly focuses on the detection of "weak" vessels with polarimetric SAR.

1) High-resolution SAR sensors have a wide azimuth beam width, as well as a large range bandwidth. During SAR image formation, multiple squint angles and radar wavelengths are integrated to synthesize the full resolution SAR image. Based on this principle, the original single-look complex image SAR data was decomposed into different sub-aperture images in azimuth direction and sub-band images in range direction through time-frequency decomposition method. The scattering difference between the vessel and the ocean in the sub-aperture images or sub-band images is studied. We found that the vessel targets had coherence attributes between different sub-apertures, but the sea surface is significant differences and exists decoherence effect.

2) The complex coherence product of the sub-images polarimetric SAR is introduced and a plurality of sub images are formed into image pairs similar to the interferogram. Polarization coherent optimal parameter named PSCO (polarmetric SAR sub-aperture /spectral coherence optimal) is constructed by combining multiple sub images pairs and the concept of permanent scatterers’ detection in polarimetric interferometry. The performance of PSCO is tested using RADARSAT-2 quad-pol data. By comparing the enhancement ability of ship-sea contrast, the ability of clutter suppression and the calculation time, we choose 3 as the number of sub-images for PSCO. The CA-CFAR method is used to detect the ship for PSCO, and compared with CA-CFAR method based on HV polarization. The results show that the proposed method can suppress the sea clutter well and improve the detection performance. This method is suitable for terrible sea conditions, strong clutter and other complex sea background for vessel detection.

Internal waves occur almost everywhere on the shelves and slope. The generation and propagation of internal waves are closely related to the ocean bottom topography, tidal current, large-scale circulation and stratification. Therefore, internal waves have an obvious local property which shows different characteristics in different areas. Remote sensing data has been widely used in the temporal-spatial distributions, generation and evolution, parameter inversion of the internal waves and is the best instrumentality for detection of internal waves on a near-global scale. The MODIS imagery owns the advantages of large swath area and near-daily global coverage, but influenced by the weather greatly. SAR has the ability of all-weather and day-and-night observation, although data coverage is limited. This paper combines SAR and MODIS images effectively to observe internal waves on a near-global scale. The ASAR and MODIS imagery on global scale over the period of 2011 was processed. The distribution of internal waves is presented by extracting crests of internal waves. Regions of internal waves are frequently observed of Pacific, Indian Ocean and Atlantic of the internal waves are presented. The further analysis of the scale, propagation law and the temporal-spatial distributions characteristics of the internal waves in the typical occurrence area are also conducted.

In this paper, a support vector machines (SVM) regression model is proposed to extract integral ocean wave parameters such as significant wave height (SWH) and mean wave period (MWP) from the ASAR wave mode images. Calibrated ASAR wave images can be applied directly to retrieve SWH and MWP without prior information. The model was established based on the nonlinear relationship between the sigma0, the variance of the normalized SAR image, SAR image spectrum spectral decomposition parameters, SWH and MWP. The input parameters of the SVM regression model are feature parameters extracted from ASAR images and the SWH provided by European Centre for medium range weather forecasts (ECMWF) is the output. A global data set of 30771 pairs of ASAR wave mode images and collocated ERA-Interim data from ECMWF is used to train SVM model. Based on the matching dataset, the particle swarm optimization (PSO) algorithm is applied to optimize the input kernel parameters of the SVM regression model and establish the SVM model. The SWH extracted by this model was compared with the ERA-Interim data and the buoy data. The RMSE of SWH is 0.34m and 0.48m and the correlation is 0.94 and 0.81, respectively. The MWP was also validated by ERA-Interim data and buoy data, the RMSE is 0.68s and 1.08s, and the Scatter Index is 0.04 and 0.09, respectively. The results show that the SVM regression model is an effective method to extract SWH from SAR data. The advantage of this model is that SAR data may serve as an independent data sources to extract SWH, which can avoid the complicated solution process of wave spectrum.

Abstract：As a kind of marine natural disaster, Green Tide has been appearing every year along the Qingdao Coast, bringing great loss to this region, since the large-scale bloom in 2008. Therefore, it is of great value to obtain the real time dynamic information about green tide distribution. In this paper, methods of optical remote sensing and microwave remote sensing are employed in Green Tide Monitoring Research. A specific remote sensing data processing flow and a green tide information extraction algorithm are designed, according to the optical and microwave data of different characteristics. In the aspect of green tide spatial distribution information extraction, an automatic extraction algorithm of green tide distribution boundaries is designed based on the principle of morphology dilation/erosion. And key issues in information extraction, including the division of green tide regions, the obtaining of basic distributions, the limitation of distribution boundary, and the elimination of islands, have been solved. The automatic generation of green tide distribution boundaries from the results of remote sensing information extraction is realized. Finally, a green tide monitoring system is built based on IDL/GIS secondary development in the integrated environment of RS and GIS, achieving the integration of RS monitoring and information extraction.

A series of oil spill accidents has occurred frequently with the development of petroleum industry and marine oil transportation,which lead to serious oil pollution.For instance,the oil spill in Penglai,China 19-3 oil field and Dalian,Xingang,China has caused great ecological and economic losses,Therefore,it is meaningful for monitoring oil spill information in real time to prevent oil spill disasters.Based on the multi-source Synthetic Aperture Radar(SAR)image data,this paper has carried out research on the remote sensing monitoring of oil spill,and the operational monitoring system of marine oil spill based on Geographic Information System(GIS)is developed,which realize the integration of preprocessing multi-source remote sensing image data,extracting oil spill monitoring information,making oil spill thematic map and publishing oil spill monitoring reports.In order to realize the rapid processing of multi-source image data integration,the original data is analyzed based on Geospatial Data Abstraction Library(GDAL)open source raster spatial data conversion library and realized image reading,geometric correction and so on;the algorithm of multi-scale image segmentation and gray contrast characteristic extraction are encapsulated on the basis of the previous step,which achieve automatic processing of SAR image block,segmentation,oil spill identification,merging,parameter assignment,etc,the one-key processing of oil spill extraction is realized;for the application demand of oil spill monitoring,production of various kinds of standardized products, automatically add the elements of cartographic decoration elements,image parameters,oil spill area,the source of oil spill and other kinds of information,real-time generate thematic reports.