Conference Agenda

We analyzed a great deal of SAR images acquired over the German part of the Wadden Sea by the L-, C-, and X-band SARs aboard ALOS-2, Radarsat-2 and Sentinel-1, and TerraSAR-X, respectively. Using this wide range of multi-frequency / multi-polarization SAR data we investigated which combinations of radar band and polarization are best suited for a classification of different Wadden Sea surface types, including sandy and muddy sediments, sea grass meadows, and bivalve beds. New parameters, based on a decomposition of the complex SAR data, were used as input into a UNet-based semantic segmentation network with a texture-enhancement (TE) module to classify intertidal sediments and habitats. The experiment results verified the superiority of TE-UNet model compared with state-of-the-art semantic segmentation models.

Oceanic eddies are widely distributed in the global ocean, they play a crucial role in the global ocean energy cycle, the transport of heat and salt, and the distribution of nutrients. Synthetic Aperture Radar (SAR) is an ideal sensor for studying ocean eddies due to its high spatial resolution and independence of daytime and weather conditions. This paper proposes a method based on the deep learning networks of the YOLO family, named EOLO, to detect and extract geographic information of ocean eddies on C-band spaceborne SAR images. Several key enhancements were made to improve the performance of EOLO, including the introductions of spatial attention mechanism and new up-sampling operator, and improvements of the feature fusion method, loss function, and anchor box size, which contribute to an average precision of 91.5%. We also conducted experiments in the Baltic Sea, the Red Sea, and the Western Mediterranean Sea to verify the generalization of EOLO in different seas, reaching 95.7%, 96.8%, and 96.5% precision respectively. Furthermore, 8569 eddies were extracted by EOLO in the Western Mediterranean Sea in 2021, compared with the eddies from the altimeter data, the results show that the SAR eddies based on EOLO can detect at least 45% of the ocean eddies invisible to altimeters and are more realistic.

We introduce a neural network for the automatic detection of waterlines on Sentinel-1A/B SAR imagery of intertidal flats. The neural network is able to segregate water from intertidal flats for certain weather conditions in the German North Sea coast with high precision for calm weather conditions.
To simultaneously detect large structures and achieve a high resolution, the neural network consists of two stages of resolution. The neural network is structured as an image-to-image network and takes radar images and an ordinary land-water mask as input. The first stage produces a low resolution (640x640 m) allocation, whether a depicted area contains mostly land, mostly water or approximately equal parts. The allocation from the first stage is added to the input for the second stage, clarifying where islands and tideways are located. The second stage considers small parts of the radar image in a high resolution, accurately segregating water from intertidal flats at the resolution of the radar image (10x10m).
After re-merging the image, flood filling is performed to eliminate minor inaccuracies. However, relevant parts can be negated by this procedure. A more detailed land-water mask might mitigate that problem.

The tangential wind speed increases from the center to the eyewall of tropical cyclones (TC) along the radial direction and begins to decay when it extends outward. The tangential wind profile model is one of the most effective and widely used methods to reconstruct the TC radial wind speed. This paper proposes a parametric tangential wind profile (TWP) model based on high-spatial-resolution SAR imagery. The new model functions are piecewise with maximum tangential wind speed as a threshold, and all of them are designed as nonlinear. Notably, the derivative at the segmentation threshold is zero to ensure a smooth transition of the estimated wind speed profile. With the SAR-derived azimuth-averaged wind speed, we can determine the model parameters and get the tangential wind speed. The TWP model outperforms the commonly used SMRV model, as it better resolves the tangential wind profile shape as depicted by both SAR-derived winds and hurricane hunter Stepped-Frequency Microwave Radiometer (SFMR) derived winds. A comprehensive analysis of the TWP model parameters is carried out by fitting tangential winds for 620 hurricane hunter flights. Interestingly, the tangential wind profiles for major hurricanes show a similar shape. The proposed TWP model can be used for improved TC characterization and forecasting purposes.

Although The SAR is unanimously recognized as a key operational remote sensing instrument for oil spill surveillance and damage assessment owing to its all-day and almost all-weather observation ability together with its fine spatial resolution, the estimation of ancillary information related, e.g., to oil’s thickness and fraction water content is still a challenging task. Crude oil slicks and their emulsions can form thick layers ranging from micrometer to millimeter and can even reach centimeter thickness in the case of fresh oil under low sea state conditions [5]. The oil thickness distribution is as important as the spatial distribution of oil slicks, which is beneficial for proper choice of response methods and spatial allocation of response resources, as well as for legal purposes for prosecution.

In this study, the potential of the electromagnetic scattering model to retrieve quantitatively parameters of oil spills on the sea surface is investigated using the Artificial Neural Network (ANN) technique. In the SAR image plane, oil slicks appear as dark patches compared with ambient clean sea surface. This is due to both geometrical and electrical effects of the oil slick which, one on side, damps the short-gravity and capillary sea waves resulting in a lower backscattered signal and, on the other side, if the slick is thick enough or emulsified it can also change dielectric properties of the upper sea layer. To extract information about the oil slick from SAR imagery while limiting the effects of sea state conditions and incidence angle, the damping ratio (DR), i.e., the slick-free to slick-covered backscattering ratio, has been widely adopted. The backscattering DR of the oil slick is predicted using the AIEM augmented with MLB, effective dielectric constant model, and the composite medium model to include the effect of an oil slick. The simulated DR and their corresponding oil parameters, namely the oil thickness and seawater volume fraction, are utilized to train and test the NN with a relatively simple four-layer structure. The rationale consists of overcoming the lack of ground information using a forward scattering model. The adequately trained NN is then applied to the L-band UAVSAR image collected during the DWH oil spill accident to retrieve the slick thickness and volumetric fraction of seawater in the oil layer. The inversion results show that the thick emulsions are in the middle portion of the slick with 2 – 4 mm thickness surrounded by thin films with thickness less than 1mm. And the water content of the DWH oil slick is about 20% - 30%. Results are contrasted with optical data and previous studies of DWH oil spill and show qualitatively good agreement.

This study is supported by the ESA-NRSCC Dragon-5 cooperation project “Monitoring harsh coastal environments and ocean surveillance using radar remote sensing sensors” (ID 57979).

Ocean surface current (OSC) is one of the key marine dynamic elements which dominates the global circulation of carbon and heat. Measurements of the OSC are of particular significance for the studies and applications of marine environment monitoring, global climate change forecasting, marine search and emergency response, etc. [1-3]. By modulating ocean surface topography and roughness, the ocean currents could be characterized on synthetic aperture radar (SAR) images [4]. By far, two main-stream technical routines for OSC measurement based on SAR have been proposed, i.e., along-track interferometric SAR (along-track INSAR, ATI) [5, 6] and Doppler centroid anomaly (DCA) technique [7, 8]. Essentially, these techniques utilize the surface Doppler information to retrieve the corresponding velocity which are confronted with the challenge of separating the contributions from waves and currents. This requires a physical modeling of radar scattering from ocean current surface. Thus, in this study, we aim to numerically investigate the polarimetric SAR imaging responses to two-dimensional ocean surfaces with currents and waves. The well-developed radar imaging model (RIM) proposed by Kudryavtsev et al. [4] is employed to conduct the numerical simulations under various frequencies, incidence angles, wind speeds and full polarizations. The current surface with a typical internal wave phenomenon generated by the MITgcm numerical mode is used, which has resolution of about 1/200° in longitude direction and 1/60° in latitude direction. The current modulation of wave spectrum is considered in the KHCC03 spectrum. Experimental results indicate that current modulation on ocean scattering is more significant at lower wind speeds. It is also noted that the current modulation of ocean scattering performs most remarkable at cross-polarization, while has least effects at VV-polarization. The current modulation effect is nonlinear to current velocity. At large current velocity, the current modulation effect could be saturated, particularly for co-polarizations. More detailed numerical results will be given in the presentation.

Sea surface wind field is a parameter of key importance for several applications that span from weather forecast up to recreation activities. In addition, it also plays a role in the context of climate change as one of the effects of global warming is the increasing occurrence of strong winds that threaten coasts and critical infrastructures.

The wind can be estimated using either in situ or remotely sensed measurements. In the latter case, the main satellite remote sensor is the scatterometer, i.e., a meso-scale radar that is specifically designed to make normalized radar cross-section (NRCS) measurements that can be converted into the wind field using tailored geophysical model functions (GMFs).

There is an increasing interest in the exploitation of finer-spatial resolution NRCS measurements acquired by the Synthetic Aperture Radar (SAR), i.e.; an imaging radar that can achieve meter or sub-meter spatial resolutions even when operated by satellite. The main challenge relies on that fact that the SAR is not meant to be operationally used for wind field estimation. In fact, it can only provide one NRCS measure for each resolution cell and this makes the inversion of the sea surface wind field an under-determined problem.

To cope with this problem, two strategies are possible that rely on a wind direction information that is introduced into the inversion problem from either external source (e.g., scatterometer, ECMWF) or directly estimated from the SAR scene. This latter option is very promising since it allows forcing the wind field estimation process with a wind direction input on the same spatial grid of the wind to be estimated.

SAR-based wind direction has been addressed using different methods: a) the analysis of the orientation of the wind-induced streaks, using Fourier transforms [1],[2],[3], the discrete wavelet transform [4], the continuous wavelet transform [5] and gradient methods local [6]. In [7] a deep residual network is used to estimate the wind direction from the SAR images even in small turbulent areas, in the absence of streaks on the SAR images and in the presence of ships.

In this study, a sensitivity study is carried out to analyze the performance of neural network (NN) wind direction retrieval on Cosmo-SkyMed X - band SAR imagery. The main objective is to estimate the wind direction from the COSMO-SkyMed SAR imagery with artificial intelligence techniques. This is physically possible by exploiting wind-generated patterns in the SAR image, therefore by estimating the orientation of these patterns.

Experiments are carried out using 19 Cosmo-SkyMed SAR scenes that are augmented with ancillary spatially and time-collocated scatterometer information acquired between February and august 2022 in the North Sea.

References

[1] Wackerman, C., C. Rufenach, R. Schuchman, J. Johannessen, and K. Davidson, 1996: Wind vector retrieval using ERS-1 synthetic aperture radar imagery. IEEE Trans. Geosci. Rem. Sens., 34, 1343-1352.

[2] S. Lehner, J. Horstmann, W. Koch, and W. Rosenthal, “Mesoscale wind measurements using recalibrated ERS SAR images,” J. Geophys. Res., vol. 103, no. C4, pp. 7847–7856, Apr. 1998.

[3] P. W. Vachon and F. W. Dobson, “Validation of wind vector retrieval from ERS-1 SAR images over the ocean,” Global Atmos. Ocean Syst., vol. 5, pp. 177–187, 1996.

[4] Yong Du, Paris W Vachon & John Wolfe (2002) Wind direction estimation from SAR images of the ocean using wavelet analysis, Canadian Journal of Remote Sensing, 28:3, 498-509, DOI: 10.5589/m02-029

[5] Zecchetto, S., De Biasio, F., Della Valle, A., Quattrocchi, G., Cadau, E., Cucco, A., 2016a. Wind fields from C and X band SAR images at VV polarization in coastal area (Gulf of Oristano, Italy). IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing 9 (6).

[6] Koch, W., 2004. Directional analysis of SAR images aiming at wind direction. IEEE Trans. Geosci. Remote Sens. 42 (4), 702–710.

[7] A. Zanchetta, S. Zecchetto, “Wind direction retrieval from Sentinel-1 SAR images using ResNet,” Remote Sensing of Environment, Volume 253, 2021, 112178, 2021.

Coastal macroalgae blooms have a profound influence on marine ecosystem balance, tourism, and aquaculture. Since 2008, the western coasts of the Yellow Sea have been damaged every summer by a green tide caused by the overgrowth of ulva prolifera. Remote sensing has been the primary tool for monitoring this green tide. With the emergence of more free synthetic aperture radar (SAR) images with high resolution and the ability to image in cloudy conditions, SAR images play an increasingly important role in green tide monitoring. Deep learning is a powerful method in remote sensing images classification. However, current studies mainly focus on the image's backscattering coefficient, while ignoring the morphological characteristics. Moreover, the proportion of algae-pixels and seawater-pixels is significantly unbalanced, which will reduce the learning ability of the deep learning method. To address these issues, we propose a deep learning method to detect ulva prolifera. The proposed model is designed on four sides: 1) We add texture features to extract the morphological characteristics of green algae. 2) We design a new loss function to maintain the learning ability of the proposed deep learning method. 3）we build a texture-enhanced path called texture concatenation to help extract Ulva prolifera with fuzzy boundaries. 4）we embed the convolutional block attention module (CBAM) after each convolution layer. For texture features, we calculated four representative gray level co-occurrence matrix (GLCM) maps of the VV polarized images, i.e., ASM, entropy, correlation, mean. Thus, the input dataset includes two polarization channels and four texture channels. For the new loss function, we used the combination loss of binary crossentropy and focal loss via different weights. To construct the proposed model, we labeled 6317/2124 pairs of Sentinel-1 SAR image patches as the training/testing dataset. All of the used images were preprocessed through radiometric calibration, speckle filtering, terrain correction, and incident angle effect correction. Experiments show that when binary crossentropy is weighted for 0.70 and focal loss for 0.30, the model performs best, with a mean intersection over union (mIOU) of 86.31%, outperforming the well-known segmentation models using the identical dataset and hyperparameters. In addition, we also used the proposed model to analyze the interannual variation of green tide in 2019-2021, and found that 2019(2020) has the longest (shortest) bloom duration and biggest (smallest) coverage area; 2021(2020) has the biggest (smallest) nearshore damage to the southern coastlines of the Shandong Peninsula.

The detection and monitoring of ground surface changes using multi-temporal, remotely-sensed images represent the most important applications of remote sensing (RS) technologies. Various applications have extensively used optical RS sensors for change detection (CD). The CD process involves analyzing two or more images captured over the same geographical area at different times to identify significant land cover changes. While optical sensors have been widely used for CD, microwave RS images acquired by synthetic aperture radar (SAR) have been less exploited. However, SAR images in CD are attractive for operational purposes since SAR sensors are active instruments that can operate in any atmospheric and sunlight conditions. Remotely sensed data collected by several constellations of SAR sensors, such as the twin Sentinel-1A/B sensors of the European (EU) Copernicus, enable fast mapping of changes of Earth's surface and allow the time monitoring of areas prone to geohydrological disasters.

Coastal flood risk is a global challenge, as about 40 million people living in coastal port cities will likely be subject to one significant coastal flood event per century. SAR remote sensing is a valuable tool for detecting and monitoring flood phenomena and can differentiate between inundated and non-inundated areas. Flood risk increases due to urban growth, ground subsidence, and climate change. Identifying areas more prone to extreme floods helps optimize urban planners' civil protection actions and evaluate the damage. Recent advances in RS technology have allowed the generation of rapid damage prediction maps and associated models that are helpful in the occurrence of a flood event.

This work focuses on the impacts of floods and extreme weather events on coastal areas' cultural heritage preservation, particularly the case of the monumental city of Venice and the whole Venice Lagoon area. The Venice Lagoon represents Italy's most extensive lagoonal system, one of the largest in the Mediterranean Sea, and one of Italy's most important industrial areas. The lagoonal system includes Venice, an extraordinary archaeological, urban, architectural, artistic, and cultural heritage masterpiece. The Venice Lagoon ecosystem is subject to various drivers of change, such as land-based feeding activities, heavy metal extraction, ground-water extraction, etc., causing multiple environmental impacts on the Lagoon. The subsidence phenomenon of the terrain is one of the most critical drivers of change. A CD study was conducted to analyze the ground deformation (subsidence) that occurred in the Venice Lagoon in recent years using the multi-temporal interferometric Small Baseline Subset (SBAS) technique. The study examined the interlinked effects between the underlying subsidence in the area and the recent extreme flood events that occurred in November 2019. The time series of backscattered S-1 signals were analyzed to identify the extent of the flooded regions and the impact of the floods on the city. The study also leveraged the potential of a newly developed AI method based on Random Forest.

This methodology uses coherent/incoherent SAR change detection indices (CDIs) and their mutual interaction in a single corpus to rapidly map surface changes. The method has shown great success in quickly mapping land surface changes of areas affected by enormous wildfires in Sardinia and Sicily in the summer of 2021 and flooded areas in Houston and Galveston Bay due to Hurricane Harvey in 2017.

We conducted a detailed analysis using 180 Sentinel-1 images acquired from January 2017 to December 2021 to investigate ground deformation in the Venice Lagoon. We generated a stack of 1736 short baseline (SB) interferograms and computed the lagoon's time series of deformation and mean deformation velocity map. We also used a series of pre- and post-flood event acquisitions on 12 November 2019 to perform a change detection analysis of Venice using temporal multi-looked sigma nought maps and Coherence Changes Indexes (CCI) from an AI-based methodology.

For the change detection analysis, we selected a time series of acquisitions with a temporal baseline of ±6, ±12, ±18 days before (11 November, 5 November and 30 October), during (17 November), and after (23 and 29 November, and 5 December) the flood event. Each SAR image of the time series underwent post-processing using a de-speckling noise filtering algorithm and co-registration using Enhanced Spectral Diversity (ESD) to the 17 November acquisition. We computed sigma nought differences (), and we used triplet of SAR images with temporal baselines of ±6, ±12, ±18, ±30, ±36, and ±42 days with respect to the 12 November pre- and co-disaster InSAR () CCI to determine the coherence changes indexes.

The results of the interlinked analysis showed that only a tiny region of the emerged lands in the north and south of the Venice Lagoon is affected by subsidence. Flood events represent a severe threat to the integrity of these areas. The deformation analysis of the city of Venice showed no significant subsidence phenomena. Still, some spot regions over the low-lying lagoon terrains were affected by remarkable signals associated with sea level rise (SLR), which can seriously impact the hydrogeology of the area.

Coastal regions are with dense population, buildings and infrastructures, and vulnerable to natural disasters. Frequent natural disasters will cause huge economic losses and human casualties. Shanghai is a costal mega city and located in low-elevation coastal zones of the Yangtze River Delta. The city is frequently affected by typhoon and storm surge. Besides, The geological foundation of the city consists of soft alluvial deposits, including clay, silt, and sand. Due to its geological conditions, Shanghai is vulnerable to ground subsidence, flooding, and other geohazards.

In order to prevent, resist and reduce the impact of disasters, this study assesses the regional disaster reduction risk (DRR) capacity of a district of Shanghai with Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) and established a machine learning aided evaluation models. We also retrieved long-term and recent ground deformation of the coastal areas of Shanghai with Small Baseline Subset (SBAS) technology and multi-sensor Synthetic Aperture Radar image time-series. We also simulate the possible flood inundation extent under different scenarios based on LISFLOOD-FP simulation model in coastal regions. For towns with weak DRR capacity, we analyze sensitivity of evaluation indicators to explore key indicators which affect improvement of DRR capacity. Finally, we proposed optimal strategies which could improve DRR capacity based on the assessment results of DRR capacity and regional disaster characteristics.

Wetland is known as the "kidney of the earth", and its ecological function, biodiversity, and various values ​​are irreplaceable. However, due to the rapid urbanization process, wetland resources are decreasing day by day, so the investigation of wetland resources is very important. The Yangtze estuarine wetland provides various ecosystem services. However, affected by human activities (upstream sediment reduction) and natural background (sea-level rise, SLR), the saltmarsh of the estuarine wetland is undergoing dramatic changes in recent years.

As one of the important ways of wetland monitoring, remote sensing technology is widely used in wetland monitoring. Traditional wetland monitoring remote sensing technology only uses optical remote sensing data for information extraction, but optical remote sensing data has high requirements for weather and specific characteristics of land cover. For wetland areas with more shallow water and higher rainfall than other areas, it is difficult to effectively obtain information on land cover for long time series using optical images. Comparing with optical remote sensing systems, Synthetic Aperture Radar (SAR) has all-weather, synoptic views of large areas, and day-night imaging capability. Microwave electromagnetic energy can penetrate shallow water and vegetation. In recent years, scholars have conducted extensive research on the information extraction of wetland characteristics based on SAR images.

In this paper, Chongming Dongtan Wetland is taken as the study area, and the multi-temporal Rardarsat-2 full-polarization SAR data and Sentinel-2A medium-resolution optical data are used as the data source. According to the particularity of the estuary wetland in the study area, different characteristic parameters are calculated, including vegetation index, water body index, spectral feature, radar feature, texture feature, and time feature. Six multi-dimensional feature data sets containing different feature parameters have been formulated. We perform object-oriented multi-scale inheritance segmentation on six feature data sets and use segmentation parameter optimization tools to select the optimal segmentation parameters. Combined with field investigation and visual interpretation of high-resolution remote sensing images, we build a classification system of Chongming Dongtan Wetland Saltmarsh Vegetation, which mainly includes three types of wetland saltmarsh vegetation: Phragmites australis, Spartina alterniflora, and Scirpus mariqueter. Different vegetation training samples and verification samples are selected on the segmented images, and the information on saltmarsh vegetation is extracted based on the random forest machine learning algorithm. To further study the interannual variation characteristics and phenological characteristics of wetland saltmarsh vegetation, this research also uses Sentinel-2, Sentinel-1, and Landsat-8 fusion images based on Google Earth Engine (GEE) to construct a medium-resolution long-term median image dataset, to obtain the spatiotemporal distribution results of saltmarsh vegetation in Chongming Dongtan Wetland from 2017 to 2022 and the quantitative migration of saltmarsh vegetation.

According to the results of this research, wetland saltmarsh vegetation protection and tidal flat utilization of Chongming Dongtan Wetland can be scientifically supported for Outline Development Plan for the Chongming world-class ecological island.

Change detection (CD) is to quantitatively analyze and determine the characteristics and process of earth surface change based on remote sensing data in different periods. It is widely used in disaster dynamic detection, urban planning, and other fields. Compared with optical remote sensing images, Synthetic Aperture Radar (SAR) images have the unique advantage for CD. As active remote sensing systems, SAR has all-weather, day-night imaging capability, and permits synoptic views of large areas. The backscatter intensity information of SAR images in urban areas is easily affected by comprehensive factors such as building layout, orientation, and surface materials, resulting in overall low radar echo intensity in some building areas. The introduction of coherence maps into change detection can improve the recognition of changed areas because urban buildings have high phase stability and coherence maps provide reliable information for building change detection (BCD). In this study, we mainly focus on two problems in practical application. First, speckle noise as an inherent characteristic of SAR images greatly influences the performance of BCD. Second, how to fuse intensity information and coherence information effectively is of significance for extracting the spatial features of changed areas. Thus, this study proposes a deeply supervised pseudo-siamese attention-guided network (DSPANet) for BCD, in which convolutional blocks have a strong ability in noise reduction owing to the large receptive fields, and the adopted pseudo-siamese structure does well in extracting intensity information and coherence information with the same network branch but different weights.

This study uses high-resolution TerraSAR-X (TSX) images covering Shanghai from a descending pass track. A set of four TSX images is acquired between 16 June 2019 and 10 September 2021. Before training the network, we pre-process TSX images to get intensity information and coherence information. Intensity information is obtained by pre-processing single-look complex (SLC) images, including radiometric correction, multi-looking processing, and geocoding. Coherence information is obtained by performing interferometry and computing coherence values of interferograms. DSPANet model is utilized to distinguish between changed and unchanged areas. It mainly consists of an encoder and a decoder. In the encoder, the convolutional block attention module (CBAM) is included to assign accurate labels to each pixel through the attention mechanism, thus enhance the feature learning of the network. In the decoder, the multi-level feature fusion (MFF) blocks with residual structure are used to avoid the gradient disappearance and gradient explosion caused by the deepening of the network, and the output of the encoder and decoder is concatenated with the skip connection for retaining shallow features. Besides, considering unstable weight updates and unsatisfactory performance while increasing the depth of the network, the deep supervision idea is introduced in DSPANet. It adds some auxiliary classifiers to realize gradient back-propagation by assisting middle layers in extracting features.

We evaluate the accuracy of the DSPANet model using five different metrics. The results show that the method is reliable in BCD and performs better than other advanced change detection methods (such as U-Net, FC-EF, and FC-Siam-Diff). The Precision, Recall, F1, Kappa, and Accuracy reach 88.02%, 84.13%, 86.03%, 85.09%, and 98.24% respectively.

Keywords: building change detection; Synthetic Aperture Radar (SAR); deep learning; coherence information

Shanghai is located in the coastal area of the Yangtze River Delta in eastern China, facing threats from various disasters such as typhoons, storm surges, flooding and ground subsidence. As a mega-city with an area of 6,340 km² and a population of over 24.75 million, Shanghai is densely populated with disaster-bearing bodies such as buildings, roads, and metro, which are easily affected by natural disasters. Frequent natural disasters will cause significant population and economic losses. Therefore, the objective of this study is to build a disaster-bearing body Comprehensive risk assessment model based on natural disaster scenario data and disaster-bearing body attribute data. Through analyzing the distribution of the disaster-bearing bodies’ comprehensive risk level, we can detect hazards in advance and reduce the potential impact of natural disasters.

Firstly, we constructed a comprehensive risk assessment indicator system according to three dimensions of hazard, vulnerability and exposure, based on the location and attribute information of buildings, and extracted the calculation indexes. We also simulated hazard Scenarios related to buildings and roads, including ground subsidence, typhoons, floods, and storm surges. Ground subsidence information is retrieved with 2017-2021 Sentinel-1A data and Small Baseline Subset (SBAS) technology. The comprehensive risk assessment model of disaster bearing bodies was constructed by weighted calculation input indexes, and realized automatic comprehensive risk assessment of the disaster-bearing bodies in large-scale. The weights of input indexes are determined by summarizing historical disasters data and scoring by experts. We use this model to evaluate the comprehensive risk level of the disaster-bearing bodies under multiple disaster scenarios, and analyze the distribution of risk levels. Finally, the regional disaster reduction risk (DRR) capacity of Shanghai and the comprehensive risk level of the disaster-bearing bodies were combined by disaster matrix to determine which high-risk disaster-bearing bodies are located in areas with low DRR capacity.

Synthetic aperture radar (SAR) Doppler shift measurements consist geophysical and non-geophysical contributions. The former is composed of the sea state (wind waves and swell) and underlying surface current. The latter contains geometry and scalloping errors, antenna electronic miss-pointing and unknown biases. In this letter, for the first time, we attempt to retrieve ocean surface radial current velocities from Sentinel-1A SAR Doppler shift observations acquired over Hurricane Maria. Doppler shifts caused by scalloping error are first estimated using linear fitting method. Doppler shifts arising from electronic miss-pointing and unknown biases are then calculated from mean value of Dopper observations over the land. Finally, we compute sea-state-induced Doppler shifts based on our recent ocean surface Doppler velocity model (so-called DPDop). The retrieved ocean surface radial current velocities are compared with the collocated high-frequency radar measurements, showing a bias of 0.02 m/s and a root-mean-square error (RMSE) of 0.19 m/s. These results suggest that the Doppler velocity model has potential to correct wave bias and can be used to derive reasonable radial current velocities under high wind conditions.

Synthetic aperture radar (SAR) is broadly known for its high-resolution imaging of the ocean surface under all weather conditions during day and night. The Doppler centroid anomaly (DCA) derived from SAR imagettes has been evidenced to well capture the line-of-sight component of ocean current velocity. Several studies have reported the analytical basis of DCA method and the monitoring of major current systems. Its applicability under tropical cyclone (TC) events is not yet examined. In this study, we focus on demonstrating the spatial features of DCA obtained over TC Maria (2017) and Cimaron (2018) as well as it relation to the surface winds. We found that the Doppler velocity provides a promising spatial representation of the surface flow under tropical cyclone wind forcing. The Doppler velocity exhibits an asymmetric feature similar to the radial wind speed with larger velocity-to-winds ratio in the front than in the rear quadrant. The combined Doppler velocity resulted from a tropical cyclone and the Kuroshio Current is distinct, particularly over the regions of encountering flow. The results shall shed light on the SAR observational capability of TC-induced surface velocity and extends our understanding of how the winds and current are coupled under TC.

Due to signal degradation or saturation within tropical cyclones, accurate estimation of inner-core high winds using satellite remote sensing is still challenging. To address this, we propose a deep learning-based approach that leverages generative adversarial networks (GANs) to reconstruct the inner-core high winds from satellite remote sensing data. Our deep learning-based model integrates dilated convolution and attention mechanisms to improve this underestimation issue of synthetic-aperture radar (SAR) data in tropical cyclones. We also tackle the scarcity of SAR data by developing a GAN model that uses Hurricane Weather Research and Forecasting (HWRF) data as a proxy to simulate missing SAR data via transfer learning. To improve the transfer learning process, we explore different pre-trained models and expand the HWRF dataset used for training the deep learning-based models. Additionally, we aim to investigate other machine learning algorithms to enhance the accuracy of scatterometer wind products (e.g. Chinese Haiyang-2 and CFOSAT). We also employ machine learning approach to fuse multi-source data for synergistic monitoring of ocean dynamic environment, specifically in the context of tropical cyclones. In summary, our study demonstrates the potential of deep learning technology for tropical cyclone reconstruction and monitoring using satellite remote sensing data，which can contribute to improving the accuracy of wind products in the context of tropical cyclones.

The CFOSAT (Chinese-French Oceanic SATellite), carrying the first Ku-band scatteromenter (SCAT) with rotating fan beam, was successfully launched in October 2018. The preliminary quality assessment of CFOSAT SCAT wind data is carried out through the comparison for the period from Jan 2019 to Jun 2021 operationally released products with in situ measurements. The reference winds include in situ measurements from offshore (> 50 km) meteorological buoys of the National Data Buoy Center (NDBC) and serval research vessels. All in situ winds were converted to the 10 m equivalent neutral winds using the coupled ocean atmosphere response experiment (COARE) bulk algorithm. The temporal and spatial differences between the CFOSAT SCAT and the in situ observations were limited to less than 30 min and 12.5 km. For CFOSAT SCAT wind speed products, the comparison and analysis using the NDBC buoys yield a bias of 0.34 m/s, a root mean square error (RMSE) of 1.24 m/s. Although less accurate of CFOSAT SCAT wind direction at low winds, the RMSE of 19.76 deg with a bias of 1.13 deg is found for wind speeds higher than 4 m/s. Moreover, CFOSAT SCAT winds were evaluated against anemometers in situ onboard R/Vs , whose cruise were distributed globally. The comparison results against R/V winds are found consistent with those by the widely used NDBC buoys. The encouraging assessment results show that wind products from CFOSAT SCAT satisfy the mission specification and will be useful for scientific community.

The Surface Waves Investigation and Monitoring (SWIM) instrument onboard the China–France Oceanography Satellite (CFOSAT) can provide wave spectra using its off-nadir beams. Although SWIM shows a reasonable performance for capturing spectral peak, the accuracy of mean wave periods (MWPs) computed directly from the SWIM spectra is not satisfying due to the high noise level of the spectra. SWIM can also provide good-quality simultaneous wind speed (U10) and significant wave height (SWH) like an altimeter. The MWP can also be estimated using a U10-SWH look-up table presented in previous studies. However, the accuracy of this method is also limited as the U10-SWH look-up table is only applicable for wind-sea-dominated conditions. The two MWP retrieval methods are independent of each other, and their error properties are complementary to each other. Therefore, this study further presents a merged MWP retrieval model combining the nadir U10-SWH and the MWP from the off-nadir spectrum of SWIM using a simple artificial neural network. After training against some buoy data, the model reaches unprecedented accuracy for MWP retrievals (RMSEs of ~0.36 s for zero up-crossing periods, ~0.41 s for mean periods, and ~0.60 s for energy periods), demonstrating the usefulness of SWIM in the studies of ocean waves.

Satellite Radar Altimetry (RA) missions are one of the important means of global ocean observation and global and regional sea level changes monitoring. Satellite radar altimetry technology can provide continuous, all-weather, nearly whole coverage observations of global ocean. Atmospheric Wet Tropospheric Delay (WTD) is one of the error sources in satellite altimetry. The WTD with the range of 0~50cm is related to the variabilities of tropospheric water vapor and cloud liquid water in the radar signal propagation path, and varies spatially and temporally. The HY-2C satellite is the third China's marine dynamic environment monitoring satellite, which carries Radar Altimeter (RA), Scanning Microwave Radiometer, Microwave Scatterometer and Calibrated Microwave Radiometer (CMR). The CMR can provide WTD data for the correction of RA sea surface height. Due to the pollutions of coastal land, sea ice, rainfall and anomalies of instrument, CMR wet tropospheric delay sometimes has large errors or even is missing. In order to solve the problem of missing or low accuracy of CMR WTD data, the Wet Tropospheric Correction (WTC) of HY-2C RA is carried out by GNSS data, reanalysis data and other microwave radiometer data in this study. Taking the WTD correction data obtained by ERA5 reanalysis data as the background field, and combining the effective CMR WTD data along the ground track of the HY-2C altimeter, the nearshore WTD data obtained by the GNSS data and the WTD data obtained by other satellite microwave radiometers, multi-source data fusion was carried out on the basis of retaining the effective CMR WTD data by using spatiotemporal matching and objective analysis methods. Eventually, the missing CMR WTD data are filled and the accuracy of sea surface height measurement of HY-2C RA is improved to meet the growing data demand.

Internal tides are internal waves with tidal frequency those occur in the stratified ocean and are generated by barotropic tides flowing through steep topography. The internal tide is an important body of energy transfer, propagation and dissipation of barotropic tide in the ocean, and is also one important factor of driving the vertical transport of ocean nutrients and influencing ocean structure and ocean circulation. The study of the propagation and dissipation of internal tides at different spatial and temporal scales has been one of the important directions in marine science. Although both the theory and numerical models of internal tides indicate that they have obvious seasonal variations, the multiscale temporal modulation of internal tides is still poorly understood. This is very important to study the dissipation mechanism of internal tides and their interaction with other oceanic dynamic processes. With the development of satellite altimetry technology and the update of internal tide information extraction technology, satellite altimetry can obtain the characteristics of internal tide with large spatial and temporal coverage compared to in situ measurement. The Luzon Strait is one of the important generation sources of internal tides in the global ocean due to its steep ridges and strong barotropic tides. In this study, the seasonal variations of two major internal tides in the Luzon Strait region, M2 and K1, are analyzed based on multi-source altimeter data. Using the multi-satellite altimetry data from 1992 to 2020, a two-dimensional plane wave fitting method is used to extract the internal tide information and establish the internal tide model for different seasons. Furthermore, the origin, propagation direction, integrated energy flux and seasonal variation of the internal tide are analyzed, which is important for the parameterization of internal tide mixing in the numerical simulation.

Ocean waves are seawater fluctuation phenomena that occur on the ocean surface and are closely related to human beings. The study of ocean waves has great significance in shipping, offshore platform construction, national defense and military affairs. Though the traditional means of ocean wave observation have accurate results, the costs are high and the spatial-temporal coverage is sparse. Satellite remote sensing such as altimeter and SAR provide a new way for ocean wave observation, but their observation coverage is also sparse and large-area synchronous observation data of ocean wave cannot be obtained. The scatterometer can obtain large-area synchronous sea surface wind field data, and there is a nonlinear relationship between the sea surface wind field and ocean wave. HY-2C is the third marine dynamic environment monitoring satellite of China, and radar altimeter (RA) and microwave scatterometer are equipped which can simultaneously obtain ocean wave height and sea surface wind field. In order to obtain more ocean wave observation data with large-area synchronous spatial-temporal coverage, the wide swath significant wave height from HY-2C scatterometer is retrieved by the deep learning method in this study. The significant wave height data of HY-2C, sea surface wind field data of HY-2C and sea surface temperature data are used as the training set. For high sea state situations with relatively small data volume, the training data set is expanded by data augmentation methods. The HY-2C wide swath significant wave height is intelligently extracted by using the recurrent neural network, and the retrieval accuracy of the HY-2C wide swath significant wave height is evaluated by the data of buoys and other satellite altimeters.

Oceanic mesoscale eddy is a kind of vortex-current motion that is approximately in geostrophic balance, which are characterized by dynamic height anomalies in the surface. Vigorous mesoscale eddies are broadly distributed in the global ocean and can be readily observable from sea surface height anomaly (SSHA) field. The vertical temperature and salinity information of the subsurface ocean can be obtained from Argo floats. Combining sea surface observation (SSHA field) provided by satellites and the vertical temperature/salinity profiles provided by Argo floats, the vertical temperature and salinity structure of mesoscale eddies in the global ocean are analyzed. The result shows that cyclone eddies generally dominate the negative temperature anomalies, and anticyclonic eddies generally dominate the positive temperature anomalies in the global ocean. In the regions with strong current variation, eddy activities are vigorous, the temperature anomalies within eddies are more significant. The global distribution of vertical salinity anomalies of eddies is more complicated than that of temperature anomalies. Both the cyclonic eddies and anticyclonic eddies show positive salinity anomalies and negative salinity anomalies. Research suggests that the differences in the temperature and salt structure of mesoscale eddies in different regions of the global ocean are mainly caused by the differences in the temperature and salt characteristics of local water masses.

Laser altimetry has been shown to be able to provide high-resolution observations of ocean properties, which are crucial for global oceanographic monitoring. The Ice, Cloud, and land Elevation Satellite 2 (ICESat-2) has demonstrated the ability to distinguish between individual ocean surface waves. This provides physical observations far from coastal regions, where most in-situ gauges are located. The CRYO2ICE campaign, which started in the summer of 2020, has provided periodic coincident orbits between ICESat-2 and CryoSat-2, allowing for the validation of ICESat-2 observations with radar altimetry. However, the data available was restricted to the northern hemisphere. Since the summer of 2022, the CRYO2ICE campaign has performed a realignment, to get coincident orbits in the southern hemisphere, enabling ocean observations in a far larger area, as well as observing the overall higher significant wave height (SWH) in this region of the oceans. This is an opportunity to extend our dataset of observed sea states and get a better understanding of the performance from ICESat-2 at extreme wave heights. In this study, the validation of ICESat-2 thereby includes data from the southern hemisphere, where we are validating two methods: individual wave observations and a statistical model. The former uses the high-resolution data from ICESat-2 to directly observe the individual surface waves and estimate the SWH. The statistical model leans closer to the conventional estimate for the SWH, which observes the general surface variance, and uses this empirical relationship to estimate the SWH. By analyzing this new data, we seek to gain insight into ICESat-2's performance at these extreme wave heights, as well as improve the accuracy of our models.

In the waters west of the Luzon Strait, a phenomenon of alternating Type A and Type B waves sometimes occurs. Previous explanations for Type A and Type B waves mostly relied on in-situ measurements at a daily or weekly scale, with fixed-point observations, short observation times, and considerable randomness due to factors such as sea conditions. Satellite SAR remote sensing, with its all-day, all-weather, large-scale observations, and relatively low-cost data acquisition, has become an important data source for ocean internal wave research.

This study uses Envisat, ERS, and Sentinel 1A/B satellite SAR remote sensing images as in-situ data and employs the MITgcm model to investigate the generation mechanisms of internal waves in the South China Sea. The model utilizes GEBCO real depth measurements as topographic input, employs the TPXO forecast model for tidal currents at the eastern ridge of the Luzon Strait and applies least squares method to extrapolate these currents to the model boundary as boundary conditions, and uses WOA13 (World Ocean Atlas 13) data as stratification input. The research results show that the moment when the model-generated internal waves reach the specified location is consistent with the moment when the internal waves captured in SAR images arrive at the same location. The numerical model results are satisfactory when verified with in-situ measurements, providing favorable conditions for investigating the generation of internal waves in the South China Sea. Furthermore, the MITgcm results are processed, and combined with internal tidal beam theory to explore the generation and evolution of Type A and Type B waves in the time series. By utilizing methods such as Hovmuller diagrams, the study traces back the timing of Type A and Type B wave generation and offers reasonable explanations.

In the previous research, we proposed a satellite SAR internal wave amplitude inversion algorithm based on Euler numerical simulation. This method simulates a steady internal solitary wave by giving an initial flow field, and then obtains the corresponding amplitude and surface flow field. The algorithm constantly modifies the initial flow field to calculate the correlation coefficient between SAR internal wave profile curve and surface flow field gradient. When the correlation coefficient reaches the maximum, the internal wave amplitude is the amplitude of inversion.

However, the initial flow field simulated in previous studies is set by the initial solution of the KdV equation. Since the internal wave theory of KdV is only valid under the condition of weak nonlinear and weak dispersion, there is a large gap between the internal solitary wave flow field given by the KdV equation and the actual internal solitary wave flow field under the condition of strong nonlinear and large amplitude, so when it is used as the initial flow field, it takes a long time for the numerical model to get a stable internal solitary wave solution.

However, the MCC internal wave theory contains high-order nonlinear terms, and the internal solitary wave solution of MCC is closer to the actual internal solitary wave in the case of large amplitude, so it can be used as the numerical model of the initial flow field to obtain the steady internal solitary wave faster. Therefore, the MCC internal wave theory is adopted to set the initial flow field, and the steady-state internal solitary waves under different stratification conditions are simulated by numerical models. The simulation results of MCC and KdV are compared.

The results show that the numerical simulation results using MCC are basically the same as those using KdV, but in the case of strong nonlinear large amplitude, the time of the numerical simulation using MCC is shorter.

The maximum wave height (Hmax) is an extremely important factor that has a significant impact on the safety of maritime activities. However, the Hmax is much less investigated than significant wave height (SWH) in the wave remote sensing. Nowadays, radar altimeters and CFOSAT provide the SWH operational but without Hmax products. A method of obtaining the Hmax from CFOSAT SWIM Level 2 parameters is presented. The buoys are the most reliable way to observe the Hmax, but the collocations between buoys and CFOSAT tracks are too few to perform the supervised learning training. The ERA5 wave reanalysis from the European Centre for Medium-Range Weather Forecasts (ECMWF) is one of the most accurate datasets. However, the obvious bias and scatter index of Hmax are found from the comparison between ERA5 and buoys located west of France. A machine learning model is firstly built to reduce the error of ERA5 Hmax. Then the corrected ERA5 Hmax is collocated with CFOSAT observations and used for the training target of SWIM Hmax retrieval. The SWIM parameters both from SWIM nadir and boxes, including the SWH, wavelength and wave partition information, are used to obtain the Hmax based on machine learning. The CFOSAT data in 2021 are used to train the Hmax machine learning model while the data in 2020 are used to perform the independent validation. The bias, RMSE and scatter index of CFOSAT Hmax are 0.01m, 0.51m, 16%, while 0.77m, 1.09m, 19% are for the ERA5. Therefore, this study provides a perspective to obtain the Hmax from satellite remote sensing for further applications such as marine forecasts.

Rivers provide key ecosystem services that are inherently engineered and optimised to meet the strategic and economic needs of countries around the world. However, limited water quality records of a full river continuum hindered the understanding of how river systems response to the multiple stressors acting on them. This study highlights the use of Sentinel-2 MSI data to monitor changes in water colour in two optically complex river systems: the Yangtze and Danube using the Forel-Ule Index (FUI). FUI divides water colour into 21 classes from dark blue to yellowish brown stemming from the historical Forel-Ule water colour scale, and has been promoted as a useful indicator showing water turbidity variations of water bodies. The results revealed contrasting water colour patterns in the two rivers on both the spatial and seasonal scales. Spatially, the FUI of the Yangtze River gradually increased from the upper reaches to the lower reaches, while the FUI of Danube River declined in the lower reaches, which is possibly due to the sediment sink effect of the Iron Gate Dams. The regional FUI peaks and valleys observed in the two river systems have also shown to be related to the dams and hydropower stations along them. Seasonally, the variations of FUI in both systems can be attributed to the climate seasonality, especially precipitation in the basin and the water level. Moreover, land cover within the river basin was possibly a significant determinant of water colour, as higher levels of vegetation in the Danube basin were associated with lower FUI values (7.9±0.6), whereas higher FUI values (9.3±1.5) and lower levels of vegetation were observed in the Yangtze system. This study furthers our knowledge of using Sentinel-2 MSI to monitor and understand the spatial-temporal variations of river systems and highlights the capabilities of the FUI in the optically complex environment.

Harmful algal blooms (HABs) occur due to a proliferation of phytoplankton within a body of water, resulting in deterioration of the aquatic environment which affects human and animal health. HABs are now a global issue with their frequency and severity increasing significantly due to anthropogenic activity and climate change. As with any natural or anthropogenic induced hazard, it is vital that efficient and effective monitoring strategies are put in place. Ocean colour satellites are effective in monitoring HABs around the world over long-term scales, however, species identification in dynamic coastal waters is challenging. Phytoplankton size class (PSC) is suggested to be a good indicator of cell size and considered to reflect the ecological and biogeochemical functional role of the phytoplankton present in the water column. Thus, it is important to be able to monitor PSCs, particularly in dynamic coastal waters where there are frequent changes in nutrients and phytoplankton community structure.

Our research draws on Sentinel-2 MSI and Sentinel-3 OLCI which differ in spatial, spectral, and temporal resolution. The objectives of this study are to develop and validate HAB detection and PSC algorithms for near-shore and coastal waters, with better generalisation capability and lower computational overload that could improve the identification of the optical characteristics directly associated with phytoplankton properties.

The study will be focused on four optically diverse regions of interest; The Danube Delta and Black Sea Coastline (Romania), Galician Coast (NW Spain), Shandong Peninsula Coast (China) and the Northern-South China Sea (China). Here, we will present results from the Galician coast. We used in-situ data such as hyperspectral Remote Sensing Reflectance, Chlorophyll-a concentration, phytoplankton abundance and taxonomy, along with fractionated chlorophyll-a and particle absorption properties to develop and test the algorithms. We focus on the detection of Alexandrium minutum from Sentinel-2 MSI and Sentinel-3 OLCI data, through the characterisation of the spectral properties directly associated with the bloom. Alexandrium minutum exhibit a range of spectral signatures depending on which optically active constituents are present in the water. We use an unsupervised Random Forest classification algorithm to develop clusters of similar reflectance spectra and propose a new indicator which can detect Alexandrium minutum from diatom dominated waters. Existing PSC retrieval algorithms based on pigment cover, chlorophyll-a abundance, and phytoplankton absorption for coastal and transitional waters will be tested in dynamic coastal waters. In addition, atmospheric correction models such as Polymer, Acolite and C2RCC were tested against in-situ hyperspectral data and their performance was evaluated.

We will present results on the optical characteristics of Alexandrium minutum and the potential of MSI and OLCI for their remote detection. We will discuss our plans for the development of Super Learners for HAB indicators and PSC and the evaluation of the PSC algorithms.

Within the joint Sino-European project “Remote Sensing of Changing Coastal Marine Environments” (ReSCCoME) we are developing techniques for the use of Synthetic Aperture Radar (SAR) data for the monitoring of European and Chinese coastal areas. We demonstrate that a classification of sediments on exposed intertidal flats is possible, when complex SAR data acquired at different radar bands is used. Single-band SAR data can already be used to generate Digital Elevation Maps (DEM) through an identification of waterlines at different water levels. Here, two approaches, including a new neural network, are used to are yielding promising results. We further demonstrate that SAR wind fields yield a useful and robust tool to assess the potential of possible future wind farms, and to demonstrate the impact of existing windfarms on their surrounding environment, particularly the deficit in local wind speed.

The project aims at exploiting microwave satellite measurements to generate innovative added-value products to observe coastal areas characterized by harsh environments, even under extreme weather conditions. The following added-values products are addressed: water pollution, intertidal area monitoring, ship and metallic target observation, NN methods to retrieve wind direction from SAR imagery.

The following activities have been addressed:

Water pollution

Previous activities: Theoretical scattering models (under monostatic and bistatic configurations) have been developed to predict sea surface scattering with or without surfactants. In the monostatic case, theoretical predictions have been contrasted with actual measurements collected by the Synthetic Aperture Radar.

New activities: A model has been developed to shed light in the prediction of oil-sea contrast using different combinations of scattering (AIEM and two scale BPM) and damping (Marangoni and MLB) models.

Target detection

Previous activities: Multi-polarization backscattering from a known ship observed at different incidence angles. The analysis is carried on using metrics based on both power and phase information.

New activities: A new metric is defined, namely the polarization signature of the degree of polarization, that can be used to better asses the scattering variability at the variance of incidence angle for both sea and targets.

The polSAR backscatter from PAZ imagery acquired over the Robin Riggs wind farm is analysed to estimate blade rotation using sub-aperture analysis

Intertidal area monitoring

New activities: A data set that consists of X-band (CosmoSkyMed and PAZ), L-band (ALOS-2) and C-band (RadarSAT-2 and Sentinel-1) polarimetric SAR scenes has been acquired in the Scottish Solway Firth intertidal area to discuss the variability of the polarimetric scattering against SAR frequency and incidence angle over a common area.

Wind speed

Previous activities: SAR and ancillary scatterometer and model-based information are used to estimate the wind vector from SAR scenes under moderate and extreme weather conditions.

New activities: A new processing chain that exploits NN to estimate wind direction from the SAR imagery is proposed and tested using X-band CosmoSkyMed SAR imagery augmented with ancillary ASCAT and ECMWF info.

All this matter will be detailed in the proposed piece of study.

Our project aims to develop and validate innovative products for inland, transitional and coastal waters to support and improve the water ecosystem services, sustainable management and security. We have made some progress on the algorithms and applications of optical remote sensing images on oil spill detecting and water quality retrieving.

Firstly, we have made some progress on optical remote sensing image preprocessing. We developed an OWT (Optical Water Types) based method for flagging land-affected signal. The developed method improved the retrieval of water quality parameters. Results show a seasonality in the land-affected signal driven mainly by sun geometry and land cover. Besides, we tested different atmospheric correction models against in-situ hyperspectral data and evaluated their performance over coastal waters.

Secondly, we applied different kinds of satellite data to detect oil spills. We assessed the performance of Ultraviolet Imager (UVI) onboard Haiyang-1C/D (HY-1C/D) satellites by the following aspects: image features of oils under sunglint, sunglint requirement for spaceborne UV detection of oils, and the stability of the UVI signal. The results indicated that in UVI images, it is sunglint reflection that determines the image features of spilled oils, and the appearance of sunglint can strengthen the contrast between oils and seawater. Besides, we proposed an object-based spectra comparison (OBSC) approach to extract emulsified oil slicks from Balikpapan Bay, Indonesia, using optical imagery from Sentinel-2 Multispectral Instrument (MSI) and PlanetScope. We used optical imagery from Landsat-8 OL to detect oil slicks on the ocean surface through spatial analysis and spectral diagnosis in the northern South China Sea (NSCS). We demonstrated the capability of medium-resolution optical imagery in monitoring regional oil spills.

Thirdly, we developed several algorithms for retrieving water quality parameters, including CDOM (Colored Dissolved Organic Matter), Chla (chlorophyll-a), and water clarity. We proposed a blended CDOM algorithm based on OWT classification. Results showed that the blended algorithm has higher accuracy in CDOM estimating than a single algorithm for all waters. We also proposed an optical classification algorithm to exclude highly turbid waters, and then to estimate Chla in the less turbid waters only. We constructed an exponential estimation model based on Rrs(NIR)/Rrs(red), and applied the model to Landsat TM and OLI images in Lake Taihu to analyze its Chla spatiotemporal distribution. We also proposed a modified model of the quasi-analytical algorithm to retrieve the water clarity of inland waters across Hainan Island, China using Sentinel-2 multispectral instrument data. Based upon this, the first spatiotemporal analysis of recent water clarity in Hainan Island was conducted.

Coastal regions are vital places for the economy, sustainability, and environmental care of entire nations with severe impacts on a global scale. However, coastal regions are vulnerable to natural disasters. The coastal regions are particularly exposed to extreme events and the effects of global climate change. Remote sensing (RS) technologies play a significant role for: i) monitoring disturbances of public/private infrastructures, ii) helping cultural/natural heritage preservation, iii) handling and maintain effective and updated disaster risk management plans, and iv) managing efficient agriculture processes. In this context, the joint European Space Agency (ESA) – Ministry of Science and Technology of China (MOST) DRAGON V GREENISH project was designed to develop and apply conventional and new algorithms for the detection and mapping of flooded areas, the analysis of urban climate-related threats and the anthropogenic disasters (e.g., ground subsidence in coastal areas and over reclaimed-land platforms), to improve the knowledge and develop innovative RS methods. GREENISH is the result of international cooperation between some European and Chinese research centers that operate in the remote sensing (RS) sector. The main project goals are: i) to detect and study the ground deformations in coastal/deltaic regions using conventional and novel interferometric synthetic aperture radar approaches; ii) to monitor changes through coherent and incoherent change detection analyses; iii) to study coastal erosion, using high-resolution optical and SAR images; iv) to assess sea level rise (SLR) and hydrogeological risks in urban coastal areas; v) to train Young Scientists (YS).

Within this framework, SAR remote sensing is a valuable tool for detecting and monitoring flood phenomena, allowing the differentiation between inundated and non-inundated areas. This work and the presentation planned at the next D5 symposium aim to summarizes the project's key achievements during the recent years and provide insights on the forthcoming activities. A special focus will be on the application/derivation of new RS techniques, also aided with artificial intelligence tools and methods. More specifically, starting from a sequence of calibrated, co-registered SAR acquisitions, the family of used methodologies for change detection analyses of Earth’s surface consists of different modules that span from the generation of proper change detection indices to the integration of these pieces of information with those achievable using novel interferometric SAR approaches, also aided by AI and multi-grid techniques. Moreover, methods about evaluation of regional disaster reduction risk capacity are also developed. Accessibility and location of emergency shelters in coastal mega city under extreme waterlogging disasters are also analyzed.

It is presented in this paper some recent progresses of ESA-MOST China Dragon Cooperation Program “Synergistic Monitoring of Ocean Dynamic Environment from Multi-Sensors (ID. 58009)” including: (1) Assessment of ocean swell height observations from Sentinel-1A/B Wave Mode against buoy in situ and modeling hindcasts; (2) Quantifying uncertainties in the partitioned swell heights observed from CFOSAT SWIM and Sentinel-1 SAR via triple collocation; (3) Up-to-Downwave asymmetry of the CFOSAT SWIM fluctuation spectrum for wave direction ambiguity removal; (4) Validation of wave spectral partitions from SWIM instrument on-board CFOSAT against in situ data; (5) Quality assessment of CFOSAT SCAT wind products using in situ measurements from buoys and research vessels; (6) Direct ocean surface velocity measurements from space in tropical cyclones; and (7) Deep learning-based model for reconstructing inner-core high winds in tropical cyclones using satellite remote sensing.

Polar lows are small and intense high latitude maritime cyclones and frequently induce typical ocean disasters such as strong winds, high waves and heavy rainfall. They remain difficult to observe and forecast due to their short lifetime (<48 hours) and small horizontal scales (200~1000 km). Satellite remote sensing is an important manner to monitor polar lows because of sparse synoptic observing network existing in subarctic and Arctic oceans. Previous studies subjectively identified polar lows by visual inspection of satellite thermal infrared imagery. However, this subjective visual analysis method is time-consuming and inevitably involves error in polar low detections. In this study, we present an automatic procedure to objectively detect and track a polar low occurring in Greenland Sea using spaceborne synthetic aperture radar (SAR) and passive microwave radiometer data. Based on the marker-controlled watershed segmentation method and the morphological image processing algorithm, Sentine-1A and RADARSAT-2 high-resolution SAR images and successive total atmospheric water vapor content field observations from multiple radiometers (e.g., AMSR2, SSM/I, GMI) are used to fix the center location of polar low. The track of this polar low is further determined from detected centers. The polar low detections are confirmed by the presence of cloud vortex signatures visible on the AVHRR and MODIS thermal infrared imagery, and the SAR-retrieved ocean surface high wind speeds. The results show that the proposed method has potential to efficiently detect and track polar low from multi-sensor data.

Satellite altimeter is one of the important means for remote sensing observation of ocean dynamic processes. Satellite altimetry data is abnormal in the coastal areas where the echo waveform of radar altimeter is affected by the land and the accuracy of the geophysical correction of sea surface height calculation is low in the coastal areas. The second ocean dynamic environment monitoring satellite of China named HY-2B is equipped the radar altimeter. The waveforms are retracked and the range corrections are reprocessed in the coastal areas in order to solve the issues that the accuarcy of HY-2B altimetry data degrades and have few effective measurements in coastal areas. According to the characteristics of echo waveform of HY-2B radar altimeter in the coastal areas, a coastal waveform retracking algorithm based on effective trailing edge and small noise leading edge is developed to reduce the influence of land pollution on HY-2B altimetry waveform retracking. The comparisons of waveform retracking results of HY-2B altimeter by different retracking algorithms show that the proposed coastal waveform retracking algorithm based on effective trailing edge and small noise leading edge has obvious advantages in waveform retracking and can obtain more accurate sea surface height observation in the coastal areas. For the degraded low accuracy of HY-2B altimetry range corrections in the coastal areas, the sea state bias correction, wet tropospheric correction, ionospheric correction and ocean tide in the coastal areas are improved, and the errors of range corrections are reduced. The comparison between the results with range correction reprocessing and the original data shows that the coastal range correction reprocessing presented in this study can effectively improve the accuracy of HY-2B altimeter data. According to the proposed coastal waveform retracking algorithm based on effective trailing edge and small noise leading edge and the range correction reprocessing method in the coastal areas, the HY-2B altimeter data in the China seas and their adjacent waters (105~135 °E, 0~42 °N) from December 2018 to May 2022 is processed. Comparisons and analyses of the sea surface height difference between the processed results and standard products show that the precision and availability of the processed HY-2B altimeter data are improved compared with the standard product.

Detection of rain on C-band synthetic aperture radar (SAR) images of the ocean is a challenging task, since several processes contribute to the radar signature of rain, which are often ambiguous: 1) surface scattering from the sea surface whose roughness is modified by impinging raindrops, and 2) volume scattering and attenuation by hydrometeors in the atmosphere. Understanding the signature that rain imposes on SAR images of the sea surface is of relevance for interpreting other features visible on SAR images of the sea surface correctly. Rain disturbs other radar signatures, e.g., those of wind patterns and of internal waves. While the contribution of surface scattering to the radar signatures of rain over the ocean has been studied intensively, the contribution of volume was often considered negligible at C-band. One mechanism that was identified only recently as an important contributor to radar signatures of convective precipitation system over the ocean, is radar scattering at hydrometeors in the melting layer (ML). Building on a previous paper, we investigate this contribution in more detail by analyzing Sentinel-1 SAR images showing radar signatures of different types of convective rain over the northern part of the tropical South China Sea. We compare them with the dual-polarized weather radar data of the Hong Kong Observatory (HKO), with and data of the Global Precipitation Mission (GPM) and with radiosonde data. The comparison shows the radar signatures due to radar scattering at hydrometeors in the ML occurs in areas where updraft has carried moist air up to the freezing level. This occurs usually near the center of the rain cell, but in one case, we have observed it also at the rim of a downdraft pattern. Here, the updraft is so strong that it reaches the height of the freezing layer, which in this case had a height of 5325 m. Our analysis has also revealed that radar scattering at hydrometeors in the melting layer does not only give rise to the often observed patches or blobs of strongly increased NRCS values at co-and crosspolarization, but also to less strong increased NRC values which lie in the range of NRCS values caused by wind. Thus, such ML-related radar signatures can easily confounded with wind signatures. Furthermore, we point out that the theory describing the radar scattering at hydrometeors in the ML, which is applied in this paper to the C-band on board the Sentinel-1 satellites, is also applicable to L-band SARs, like the one flown on Seasat. Finally, we show examples how rain disturbs the radar signature of internal waves

HY-1C and HY-1D are the two ocean color satellites in China which play the important role in routine work of global marine environment monitoring launched separately in 2018 and 2020. The overall objective of HY-1 serial satellite is to monitor global ocean color and SST (Sea Surface Temperature), as well as the coastal zones’ environment. The China France Oceanography Satellite (CFOSAT) and Haiyang-2B (HY-2B) satellites were successively launched in China in 2018. As missions for measuring the dynamic marine environment, both satellites can measure the nadir significant wave height (SWH). Sentinel-2A/B satellites were launched in 2015 and 2017 separately. In this project, all these satellites data have been used to monitor marine disaster and environmental changes. Based on the various methods and different data types, satellite remote sensing monitoring research have been conducted in several typical marine disasters and dynamic environment changes. The results show the advantages both in new algorithms and multiple satellite data applications. The main developments in 2023 of the project are as follows:

1) Based on the time series HY-1C/D satellite data in 2019-2021, the long-term oil spills detection has been conducted in China Seas and coherent areas. The results show that it’s possible to distinguish the various spill types, for example the emulsified and non-emulsified oils, using the CZI satellite data in the condition of different sun-glint reflections which also displays the outstanding advantages of HY-1C/D data applications. According to the 3 years data analysis, the spatial patterns of oil spill distributions have been conducted for the first time in the China Seas.

2)Using HY-1C/D and MODIS satellite data, this project investigates the green tide biomass in the Yellow Sea and East China Sea. According to the characteristics of different spatial resolution data, we develop a comprehensive method to classify the difference of monitoring results using various satellite data which could improve the accuracy of greed-tide detection and coherence the green-tide bio-mass evaluations resulted from different satellite data. The results show that : 1)Compared with both of pixel area and cover area, the uncertainty of biomass estimations is the least one which could reduce the scale differences involved in the area estimations evidently and could be used to quantify the green-tide monitoring . 2) Based on the CZI and MODIS data in 2021, the comprehensive monitoring of green-tide using biomass-like method has been conducted to display the reasonable spatial distributions as well as the evolution tendency with high accuracy.

3) A new method is proposed to compare and verify ocean wave spectrum by remote sensing and in situ measurements at the spectral level. Under different sea conditions and sea surface conditions，mean directional wave height spectra from surface waves investigation and monitoring (SWIM/CFOSAT) are compared at the spectral level to the buoy counterparts, in different classes of the sea state. Under medium and high sea conditions, 8 ° and 10 °SWIM spectra have a high consistency with buoy observations.Under low sea conditions, bias between SWIM and buoy observation mainly due to parasitic peak, non-linear surfboard effect and a slight underestimation of speckle noise spectral density.

4）In this project, the HY-2B altimeter and CFOSAT nadir SWHs have been validated against the National Data Buoy Center (NDBC) buoys and the Jason-3 altimeter SWH data, respectively, which resulted in CFOSAT nadir SWH having the best accuracy and HY-2B having the best precision. The SWHs of the two missions are also calibrated by Jason-3 and NDBC buoys. Following calibration, the root mean square error (RMSE) of CFOSAT and HY-2B are 0.21 and 0.27 m, respectively, when compared to Jason-3, and 0.23 and 0.30 m, respectively, compared to the buoys. Our results show that the two missions can provide good-quality SWH and can be relied upon as a new data resource of global SWH.

5）Icebergs are big chunks of ice floating on the ocean surface, and melting of the icebergs contribute for the major part of the freshwater flux into ocean. Dynamic monitoring of the icebergs and accurate estimation of their volume are of great importance to predict the trend of freshwater budget of the Southern Ocean. Prydz Bay in Antarctica with a large number of icebergs is selected as the study area. In this work, a normalized shadow pixel index (NSPI) is designed to identify iceberg shadows with different shapes in HY-1C/D CZI and Sentinel-2 MSI images. Besides, the iceberg freeboard can be determined with considerable precision (~1.13 m). Moreover, the basal melting of icebergs has been preliminarily assessed according to the variation of iceberg freeboard using repeated MSI observations. The results indicate that icebergs in Prydz Bay were with a mean freeboard of ~56 m in early December 2022, and experienced a reduction in freeboard of ~1.89 m within two months, in correspondence with the Antarctic seasonal trend. The new methodological framework, therefore, turns out to be a reliable complementary approach to studying the iceberg freeboard in polar regions.

The use of Significant Wave Heights (SWH) on the swath of scatterometers satellite missions has been shown to be of great interest for monitoring wave propagation in storm conditions and improving wave forecasting in coastal areas. In this work, the production of swath significant wave heights for all HY2B, 2C and 2D satellite missions is pursued and assimilation tests of these data have been implemented and evaluated with independent buoys and altimeters wave data. Morever combined assimilation experiments of swath SWH jointly with CFOSAT and Sentinel-1 directional spectra have been performed with the latest CFOSAT level 2 processing (IPF-6). This latter provides improved antenna gain and directional wave spectra with better data quality filtering.
We will also present in this work the development of maximum wave height retrieval using deep learning technique for HY2 missions and their use for the detection of dangerous waves by combining with parameters computed on the directional wave spectra of CFOSAT and Sentinel-1. Extreme events with strong wave-current interactions and wave propagation in cyclone conditions have been investigated.

This work investigates the recent implementation of significant wave height from SAR wave spectra by machine learnig technique. Assimilation experiments have been performed by using SWH from Sentinel-1 and SAR and SWIM directional wave spectra. The validation with buoy wave data indicates very good consistency in terms of bias and scatter index of SWH.

Further results related to impact of the assimilation of these new wave products on the coupling with ocean model and what consequences on upper ocean mixed layer.