12.7 OCEAN FRONTS AROUND ALASKA FROM SATELLITE SST DATA Igor M. Belkin*, Peter Cornillon and David Ullman University of Rhode Island, Narragansett, Rhode Island ABSTRACT long-term variability of fronts to be studied. Long-term time series of satellite observations are Satellite-retrieved sea surface temperature (SST) used to survey ocean surface thermal fronts in the data have been widely used for regional frontal Gulf of Alaska, Bering, Chukchi and Beaufort Seas studies since the 1970s (e.g. a global survey of as well as in the upstream region of the Alaskan Legeckis, 1978). Earlier studies (see extensive Current and farther south along the North bibliographies in Belkin et al., 2003a,b,c) have American shelf to include British Columbia waters been focused mainly on fronts associated with and the Columbia River Plume area, thus covering western boundary currents such as the Gulf 45°N-75°N, 160°E-120°W. The Cayula-Cornillon Stream, Kuroshio, Brazil-Malvinas Confluence, algorithms for front detection and cloud screening Agulhas Current, and East Australian Current; were applied to the Pathfinder twice-daily 9-km fronts associated with eastern boundary currents resolution SST images from 1985-1996. A number and coastal upwellings have been studied in such of new frontal features have been identified in the areas as the California Current, Peru-Chile Alaskan Seas; some previously known fronts have Current, Canary Current and Northwest African been systematically studied for the first time. upwelling, Benguela Current, and Leeuwin Frontal frequency maps are provided for each of Current; in the open ocean, the North Atlantic four Alaskan Seas. In the Gulf of Alaska and Subtropical Front and the North Pacific Subtropical Bering Sea the SST fronts are best defined in Front received most attention, while high-latitude spring (May) and fall (November), while being fronts masked by persistent cloudiness have masked by surface heating in summer. In the nonetheless been studied in the Nordic Seas and Chukchi and Beaufort Seas the SST fronts are the Southern Ocean. best seen in summer (August-September), when Satellite observations of surface fronts in high- both seas are typically ice-free. Seasonal latitude seas are hampered by seasonal ice cover evolution of SST fronts is noted off the Oregon- and persistent cloudiness. Nonetheless, several Washington coasts and Vancouver Island, in studies have demonstrated the great potential of Hecate Strait and Dixon Entrance. remote sensing, including infrared imagery, in observing surface manifestations of oceanic phenomena (fronts, eddies, upwelling etc.) such INTRODUCTION as the Warm Coastal Current in the Chukchi Sea Ocean fronts are relatively narrow zones of (Ahlnäs and Garrison, 1984), coastal upwelling off enhanced horizontal gradients of physical, St. Lawrence and St. Matthew islands in the chemical, and biological properties that separate Bering Sea (Saitoh et al., 1998), the St. Lawrence broader areas of different vertical structure Island Polynya (SLIP; Lynch et al., 1997), and (stratification). The fronts are crucial in various spring blooming in the Bering Sea (Maynard and processes that evolve in the ocean and at the Clark, 1987; Walsh et al., 1997). ocean interfaces with the atmosphere, sea ice and The above studies, being very important in ocean bottom (Belkin, 2003a, b). Until the 1970s, elucidating physics and geography of individual frontal studies were based on ship data (Fedorov, fronts, didn’t amount however to a regional 1986). The main problem in using ship data for synthesis, which requires a unifying approach to climatological purposes is extremely non-uniform, be consistently applied to a long-term data set of patchy, and mostly sparse coverage provided by thoroughly calibrated measurements. The present ship data. On the contrary, satellite data, at least, work summarizes the most important results of in principle, are spaced regularly and allow a fairly such a project undertaken at the University of dense global coverage to be attained. The satellite Rhode Island (URI), where advanced algorithms data sets extend back to the early 1980s, thus for front detection and cloud screening have been encompassing nearly two decades and allowing a developed earlier, described in the next section. *Corresponding author address: Igor M. Belkin, The present work is essentially an exploratory Univ. Rhode Island, Grad. School of study. The main goal is to describe all the robust Oceanography, 215 S. Ferry Road, Narragansett, (persistent) frontal features noticeable in the data, RI 02882, email: [email protected] regardless of the spatial scale of the features, and compare them with previously available term frontal frequency maps, quasi-synoptic frontal observations. As such, the work presents a brief composite maps, and long-term frontal gradient provisional compendium of thermal fronts maps. The long-term frontal frequency maps show observed in the Alaskan seas. the pixel-based frequency F of fronts normalized on cloudiness: For each pixel, F = N/C, where N is DATA AND METHOD the number of times the given pixel contained a front, and C is the number of times the pixel was Because the fronts are originally defined as cloud-free. Thus, the frequency maps are best high-gradient zones, most objective computer- suited for displaying most stable fronts. At the based approaches to front identification are based same time, frontal frequency maps understate on gradient computations (e.g. Kazmin and some fronts associated with widely meandering Rienecker, 1996; Yuan and Talley, 1996). The currents such as the Gulf Stream Extension, North approach used in this study is based on Atlantic Current, and Azores Current. In such histogram analysis. Since a front is a boundary cases quasi-synoptic frontal composite maps are between two relatively uniform water masses, most helpful because they present all of the histograms of any oceanographic characteristic synoptic snapshots of the "instant" fronts detected (e.g. SST) in the vicinity of the front should have in individual SST images within a given time two well-defined modes that correspond to the period (e.g. week, month, or season), without any water masses divided by the front, while the latter averaging or smoothing. The frontal composite corresponds to the frequency minimum between maps thus allow one to detect the most unstable the modes. This basic idea has been implemented fronts that are not conspicuous in the frontal by Cayula and Cornillon (1992, 1995, 1996) and frequency maps. The long-term frontal gradient Ullman and Cornillon (1999, 2000, 2001); the maps show two scalar quantities, gradient reader is referred to these works for pertinent magnitude and gradient direction, associated with details. The fronts used for this study were derived each frontal pixel in the long-term frontal from the NOAA/NASA Pathfinder SST fields frequency maps. Gradient visualization by color (Vazquez et al., 1998) for the period 1985-1996, mapping gradient magnitude and direction (e.g. covering almost entire World Ocean, from 75°N to Ullman and Cornillon, 1999) has a clear 75°S. These fields were obtained from the AVHRR advantage over vector mapping, namely Global Area Coverage data stream (two 9.28-km resolution. Scalar mapping allows even tiny details resolution fields per day) and are available from to be preserved, down to a single pixel, whereas the Jet Propulsion Laboratory. SST fronts were vector mapping typically requires subsampling, obtained from the cloud-masked SST fields with otherwise vector maps become crowded and the multi-image edge detection algorithm (Cayula illegible. and Cornillon, 1996; Ullman and Cornillon, 1999, 2000, 2001). The cloud masking and front GULF OF ALASKA detection algorithms were applied to each of the 8364 SST images in the 12-year sequence. The Fronts in the Gulf of Alaska vary strongly with frontal data were aggregated over months (e.g. 12 season and year. In late fall-winter, the Shelf- Januaries taken together), and seasons (e.g., the Slope Front (SSF) is observed in the northern winter climatology is obtained from all Januaries, and northwestern Gulf that peaks in February-April Februaries, and Marchs taken together). The front between 140°W and 165°W, extending from detection and tracking is conducted at three levels: Queen Charlotte Islands in the east up to window, image and a sequence of overlapping Shumagin Islands in the west (Figure 1). This images. The optimum window size is important. front is associated with the Alaskan Stream (e.g. Based on a series of numerical experiments with Reed and Schumacher, 1987)). In winter, the various window sizes, Cayula and Cornillon (1992) Alaskan Stream seems to be bounded by two have arrived at the optimum window size of 32 by parallel fronts. Such situations were clearly 32 pixels. The front detection algorithm uses all recorded in March 1987, April and December pixel-based SST values within each window to 1992, January and March 1995, and March 1996; compute a SST histogram for the given window. less clearly, in February 1986 and March-April For each window that contains a front (a relatively 1990. The SSF seems to sporadically form a large narrow zone of enhanced SST gradient), the meander off Kodiak Island at 148-150°W, first corresponding SST histogram would have a described by Musgrave et al. (1992) who correctly frequency minimum identified with the front. Three (albeit tentatively) mapped it from a single basic types of maps are used in the analysis: long- hydrographic section in April 1988 augmented by drifter observations. Our frontal composite maps have not been described before. It is not clear at for April and May 1988 revealed the same this point if there is any relation between these meander and confirmed the mapping by Musgrave fronts. Both fronts might have some relevance to et al. (1992). Moreover, we have identified similar the southwestward outflow from the Shelikof Strait meanders in February 1989, December 1992, and observed from drifters (Schumacher and Kendall, March 1993.