The Royal Canadian Navy produces a semiweekly map of major water mass boundaries in the Western North Atlantic using temperature measurements from several data sources, including satellite sea surface temperature (SST) images from the Advanced Very High Resolution Radiometer (AVHRR). Temporal–spatial detail that can be provided by AVHRR of the location of important SST boundaries such as the Gulf Stream North Wall is limited due to pervasive cloud cover. The ability of satellite-borne synthetic aperture radar (SAR) to image SST front signatures unrestrained by cloud cover makes it a potentially significant additional data source. The Spaceborne Ocean Intelligence Network project has developed an automated procedure to detect candidate SST front signatures in RADARSAT-2 SAR images of the ocean surface and classify them with greater than 80% accuracy.
The Royal Canadian Navy's Meteorology and Oceanography Centre (MetOc) Halifax currently produces a semiweekly ocean-feature analysis (OFA) to provide the fleet with the location of major water mass boundaries in the Western North Atlantic Ocean, such as the Gulf Stream North Wall (GSNW). OFAs are generated using temperature measurements from a network of buoys, temperature profile measurements from ships, satellite sea surface temperature (SST) images from the Advanced Very High Resolution Radiometer (AVHRR), and SST climatology. An automated analysis integrates all of the data using geophysical interpolation techniques (kriging) to fill in gaps, such as those due to cloud cover. Cloud cover is prevalent in the region, with clear-sky percentage varying from about 50% in late summer to less than 5% in winter (according to the International Satellite Cloud Climatology Project), hampering OFA production.
The synthetic aperture radar (SAR) sensor on the RADARSAT-2 satellite measures variations in the roughness of the ocean's surface unrestrained by cloud cover. These variations manifest as dark (low microwave backscatter) regions and bright (high microwave backscatter) regions on the SAR image, and often include high-contrast edges or brightness fronts that are signatures of SST fronts. RADARSAT-2 images therefore represent a potentially significant additional data source for OFA production.
SST FRONT SIGNATURES IN SAR.
SAR images reveal spatial variations in small-scale ocean surface roughness, forced primarily by near-surface winds and modulated by longer waves, currents, and surfactants. A positive buoyancy flux within the marine atmospheric boundary layer (MABL) on the warm side of an SST front can force convective processes that transport momentum from the upper MABL toward the surface. Furthermore, the atmospheric temperature gradient often present across an SST front can produce a corresponding cross-front atmospheric pressure gradient. These processes can lead to an enhancement in near-surface wind speed on the warm side of the front, with concomitant intensification of surface roughness compared to the cold side. The processes can therefore lead to a brightness front in the SAR image that is collocated with the SST front.
SPACEBORNE OCEAN INTELLIGENCE NETWORK.
The Spaceborne Ocean Intelligence Network (SOIN) is a six-year project, supported by the Canadian Space Agency via its Government Related Initiatives Program, mandated to develop procedures that can automatically detect ocean SST front signatures in RADARSAT-2 images. The centerpiece of the project is a detection-and-classification algorithm that uses an edge detector to identify brightness fronts in SAR images that may be SST front signatures. A statistical classification algorithm is then used to discriminate SST front signatures from horizontal wind shear signatures not associated with SST fronts. This automated process achieves classification accuracy greater than 80% in the vicinity of the GSNW.
SUMMARY OF SOIN EDGE DETECTION AND CLASSIFICATION RESEARCH.
SOIN fostered the research presented in 2012 and 2013 articles published by Jones et al. in the Journal of Atmospheric and Oceanic Technology, the latter of which is outlined here.
A total of 1,227 RADARSAT-2 ScanSAR Narrow A images (C-Band, VV polarization, 300 km by 300 km) were provided to the SOIN team by MetOc Halifax. Those images were preprocessed prior to edge detection. This consisted of the application of a land mask, a filter to remove signatures of strong target sources such as ships, block averaging to reduce image size (and thus computational cost) from 12,000 by 12,000 pixels (corresponding to pixel spacing of 25 m) to 1000 by 1000 pixels spacing (with pixel spacing of 300 m), and radiometric flattening to account for systematic variations in backscatter due to changes in viewing geometry across the image. A Canny edge detector was then applied to each image to identify brightness fronts.
A set of 495 Canny edges consisting of 302 SST front signatures and 193 signatures of horizontal wind shear not associated with an SST front—all located in the vicinity of the GSNW—was collected from 252 RADARSAT-2 images. Comparing each RADARSAT-2 image with concurrent MODIS SST and surface weather analyses validated the identities assigned to the Canny edges. Edges identified as SST front signatures were those matching the location, orientation, and shape of an SST front in a MODIS image. Edges aligned with an atmospheric front in a surface analysis chart were identified as being signatures independent of SST fronts. Edges not included in the analysis were those associated with readily identifiable signatures of processes such as atmospheric or oceanographic gravity waves and large-scale convective squalls, as well as edges that could not otherwise be identified with confidence due to lack of validating SST data.
The 495 Canny edges were subjected to statistical analysis with the objective of identifying textural or contextual measures near the fronts that could be used for statistical classification. The only informative measure identified was the mean angle between a Canny edge and the wind direction. Using nearest-in-time near-surface QuikSCAT Level 3 data (25 km in resolution) obtained from NASA's Physical Oceanography Distributed Active Archive Center (PO.DAAC) website (http://podaac.jpl.nasa.gov), it was found that signatures of horizontal wind shear not associated with SST fronts tended to coincide with near-surface winds oriented along the Canny edge, typically at angles less than 23° from the along-edge direction. In contrast, near-surface winds tended to be oriented across SST front signatures at angles greater than 23°. Classification of the 495 Canny edges using mean wind angle with a decision boundary of 23° resulted in a classification accuracy of 0.83 (Table 1). Using Fisher's Exact Test, the frequency of agreement between validated identity and assigned class was found to be significantly greater than that expected by chance alone (p value < 0.0000).
In the SOIN context, a practical operational approach to SST front detection in RADARSAT-2 images will entail automated detection and classification of brightness fronts followed by manual input and evaluation. First, an image will be processed to produce candidate SST front signatures using the Canny edge detector. To reduce the number of edges to a manageable number, only those on the order of 80 km or more in length will be retained for analysis. The mean angle between the near-surface wind direction and each edge will be calculated, and edges with a mean angle of more than 23° will be classified as SST fronts. Winds from the Global Environmental Multiscale (GEM) Model, obtained from the Canadian Meteorological Centre, will be used to determine the average angle between the wind and a detected brightness front.
A manual analysis will then be carried out by MetOc personnel trained to use contextual cues (e.g., the previous OFA, visual cues in the SAR image that are commonly associated with SST fronts near the GSNW, surface weather analysis) to correct misclassifications and/or to check for potential SST front signatures not found by the edge detector. Water mass boundaries will then be drawn based on all available information (see Fig. 1). The efficacy of this semiautomated approach will be tested once the operational procedure is in place.
Although SAR observation of the ocean surface is less constrained by cloud cover than infrared radiometers, strong atmospheric processes can sometimes obliterate signatures produced by SST fronts. The question remains as to whether an OFA that includes SST fronts derived from SAR images is more informative than an OFA that does not. To answer this question, the probability that an SST front signature exists in a RADARSAT-2 image of a region near the GSNW, at a time when cloud cover prevents front identification using satellite SST, must be determined. An estimate of this probability cannot yet be provided because in the present study, cloud-free SST images were required for validation. It is expected that this probability will be revealed as data accumulates during the upcoming operational phase of SOIN.
In addition to the methods described in this communication, members of the SOIN team have explored a second approach for identifying SST front signatures in SAR images that makes use of the well-documented linear relationship between SST gradients and perturbations in near-surface winds. Using select cases, Kuang et al. (2012) demonstrated the potential of a measure derived from the wind stress components along and across a brightness front to uniquely identify SST front signatures. Once this methodology has been statistically validated using a large dataset, it may prove to be an important additional component of the classification algorithm.
The SOIN team thanks the Canadian Space Agency, which provided funding for SOIN via its Government Related Initiatives Program.