Decorrelation Scales of Sea Surface Temperature Imagery
for the Gulf of Maine
Proposal to the
Maine Space Grant Consortium
School of Marine Sciences
University of Maine
Orono, ME 04469
Satellite observations of sea surface temperature in the Gulf of Maine have corroborated many aspects of our understanding of the Gulf of Maine’s complex physical circulation. Additional information is required to help interpret the many features in sea surface temperature (SST) images who’s spatial and temporal scales are sharp or transient and cannot be placed into a physical context. This project proposes to make measurements of spatial and temporal decorrelation scales of SST features. Robust statistical measurements of the decorrelation timescales features will improve our ability to link image features to the underlying physics. This has important implications for future comprehension of the Gulf of Maine renowned biological productivity
Summary of proposal
The Gulf of Maine is a biologically productive, partially enclosed shelf-sea with strong bathymetric relief and complex physical circulation. The general circulation, consisting of large-scale anticyclonic circulation through the Gulf of Maine, provides the background for a system of linked gyres over Jordan and Georges Basin in the eastern gulf, and a vigorous coastal current system. There is a strong temporal variation to this system, inherited from the seasonally variable input of Atlantic Slope Water SLW and Scotian Shelf Water and from the vernal evolution of the coastal current. Climatic changes to the Gulf’s circulation result from the proximal influence of the Labrador current and the Gulf Stream, which originate at polar and tropical latitudes. Numerous ship-board and moored physical and biological observations have repeatedly shown a surprising variability to the circulation, often punctuated by abrupt changes in the flow regime This proposal seeks to use satellite remote sensing to make novel measurements of the temporal and spatial scales of these intriguing and unexplored transitions.
Satellite remote sensing of sea surface temperature (SST) is a powerful tool for large scale observation of the ocean and coastal regions. Sea surface temperature results from a complex history of surface heating, air-sea exchange, lateral advection and exchange between surface and deep water. SST images have been used successfully to understand many processes such as development and advection of geostrophic eddies, frontal genesis, and the dynamics of freshwater plumes on the continental shelf. Much of our present knowledge about the Gulf of Maine’s circulation has evolved in conjunction with satellite observations of sea surface temperature. Images of SST in the Gulf of Maine exhibit dramatic temperature changes over very short spatial scales of a few kilometers. These sharp features frequently evolve and disappear in a few days time. Field observations of these features rely on fortuitous encounters (e.g. Pettigrew, et al., 1997), and there is a lack of evidence to guide us in their interpretation. We are therefore unable to place many of the fine-detailed SST features into their proper physical context. An urgent goal of this project is to measure the regional variation in decorrelation timescales of SST patterns, as these timescales are indicative of underlying physical mechanisms. Robust statistical measurements of dominant time and spatial scales can be extracted from the high resolution details in SST imagery using standard mathematical techniques. These scales are hypothesized to vary between sub-regions and, in particular, between the eastern GOM and western GOM.
The results of this research will advance the confident use of SST images as a diagnostic tool for study of the Gulf of Maine circulation. By evincing the temporal and spatial scales of SST image patterns, we improve future field sampling efforts to capture features like transient or seasonal fronts, coastal plumes and eddies, which may make important contributions to the Gulf of Maine’s biological productivity. In addition, this work may augment studies underway here at University of Maine in Penobscot Bay, the Eastern Maine Coastal Current and Georges Bank. Additional benefits will accrue by fostering the growth of a University-operated remote sensing capability which is dedicated toward advancing research in marine science.
Description of Proposal
The Gulf of Maine is a biologically productive, partially enclosed shelf-sea with strong bathymetric relief and complex physical circulation. Our present understanding of the kinematics and the physical dynamics of the Gulf’s circulation has evolved considerably through observational and modeling studies. The major pathways of water to the Gulf of Maine are identified as surface inflow from the Scotian Shelf Water(SSW), deep inflow of nutrient rich Atlantic Slope Water (SLW) at great depth through the Northeast Channel, with outflows through the Great South Channel and via the anticyclonic gyre around Georges Bank. Recent evidence from hydrographic and satellite tracked drifter observations has advanced a more complete conception of the Gulf’s general circulation (Pettigrew and Hetland, 1995). The overall anticyclonic circulation from Nova Scotia to western Massachusetts provides the background for a system of linked gyres over Jordan and Georges Basin in the eastern gulf, and a vigorous coastal current system extending along the coast from the Bay of Fundy into the western Gulf. There is a strong temporal dependency in this system on seasonal variability in SLW and SSW input and the vernal evolution of the coastal current. Climatic changes to the Gulf’s circulation result from proximal influence of the Labrador current and the Gulf Stream which originate at polar and tropical latitudes, respectively. Numerous ship-board and moored physical and biological observations by Pettigrew and Hetland (1995), Townsend (1992), Brooks (1987), and many others have repeatedly shown the surprising variability in the general circulation, punctuated by locally abrupt changes in the flow. This observed behavior may be the rule, rather than the exception. In particular, Brooks (1987) has studied intrusions of Gulf Stream ring streamers, Bisagni, et al. (1996) and Townsend (1994) have observed rapid advection of SSW onto Georges Bank, Pettigrew et al (1997) cataloged the vernal eruption of an EMCC plume into the central GOM, and Pettigrew and Hetland (1995) have described complicated and unstable trajectories of satellite-tracked drifters around the linked Georges and Jordan Basin gyres. These observations indicate important processes that may promote delivery of nutrient-rich waters to the surface, stimulating biological productivity. The urgent need to understand these facets of circulation has increased with the growing recognition of the linkages between the Gulf’s large primary productivity and potential yield to fisheries and aquaculture industries. In this proposal, we seek to clarify the importance of these features by quantifying their dominant temporal and spatial scales using satellite images of sea surface temperature.
Remote sensing from satellites is an important tool for rapid, large scale observation of the ocean and coastal regions. Sea surface temperature results from a complex history of surface heating, air-sea exchange, lateral advection and exchange between surface and deep water. SST images have been used successfully to understand many mesoscale (10-100 km) processes. Much of our present knowledge about the Gulf of Maine’s circulation has evolved in conjunction with satellite observations of sea surface temperature, which can reveal the great contrast in surface water temperatures between circulation systems in the Gulf. Images of SST in the Gulf of Maine can exhibit dramatic temperature changes over very short spatial scales of a few kilometers. In many cases these thermal gradients are limited in their spatial extent and often in their temporal persistence. Examples of this include the sporadic eruption of large mesoscale eddies from the Eastern Maine Coastal Current (EMCC), ‘streamers" in the terminal end of EMCC, and cold SST bands between the Jordan and Georges Basin Gyres.
Previous analysis of GOM SST patterns has focused on the annual variation of large scale features. Bisagni (1994) produced a long-term auto-correlation function from three years of images covering the entire GOM. This function related the persistence timescale of the most dominant features in the SST. This was used to aid in optimal interpolation over temporal gaps in image time series. This has proved useful for examining large-scale circulation features (e.g. identification of cold water band, Bisagni (1994)), but the interpolated images no longer contained small scale features (< 11 km) or those persisting less than five days. Bisagni pointed out that the exact form of the autocorrelation function at every location in the GOM was still unknown and likely influenced by local circulation features in different parts of the Gulf. An important goal of this project that builds on Bisagni’s work is to measure the regional variation in decorrelation timescales of SST patterns.
Another important and unresolved SST pattern is the observed development of thermal contrast between the eastern and western Gulf of Maine. By late spring, the difference in the surface temperature of the stratifying GOM can be seen as a distinct front separating colder surface waters of the eastern GOM from warmer waters to the west. Seventy years ago Bigelow (1927) remarked on this puzzling feature, noting it implied a rate of warming that decreased from east to west by more that could be accounted for by regional differences in downward mixing of heat (from an east-to-west gradient in tidal mixing strength). Recent evidence from Pettigrew and Hetland (1995) of hydrographic and satellite tracked drifter observations has shown a distinct surface temperature difference and a large degree of drifter segregation between east and west. These new observations have advanced revised conceptual models of the general circulation, which further suggest that the east-west SST front may separate significantly different oceanographic circulation regimes.
Rational of this project
Images of SST in the Gulf of Maine commonly exhibit dramatic temperature changes over very short spatial scales of a few kilometers. These thermal gradients are limited in their spatial extent and often in their temporal persistence. We are unable to place many of these fine-detailed SST features into their proper physical context, because we lack sufficient information to identify the responsible physical mechanisms. To date, field measurements from ship-borne and moored sensors have relied on fortuitous encounters with transient features. These observations suffer from under sampling due to the abrupt appearance and limited spatial extents of these features. This has placed severe restrictions on our ability to interpret these features in their proper oceanographic context. If we are to reconcile the patterns of satellite observed SST with the growing knowledge of GOM circulation, we must answer two urgent questions: 1) what are the decorrelation timescales of these SST images features?; and 2) do decorrelation timescales exhibit regional differences that reflect different physical circulation regimes?
To answer these questions requires a better statistical record of the dominant time and spatial scales of the SST features in parts of the Gulf of Maine. Satellites provide the advantage of synoptic, two dimensional coverage spanning hundreds of square kilometers. Multiple images are collected each day, making an ideal data set for measuring the rate of pattern change as a function of spatial dimension over large areas of the Gulf. Using standard techniques of cross-correlation and spectral analysis, robust statistical measurements of dominant temporal and spatial scales can be extracted from high resolution details in SST. These scales are hypothesized to vary between sub-regions and in particular between eastern GOM and western GOM. Spectral analysis of satellite imagery can be used to describe spatial variability, with great success (e.g. Denman and Abbott, 1994). Due to the Gulf’s temperate latitude, we encounter occasional storms, fog banks and atmospheric haze, which reduces the frequent collection of cloud-free images. This can be mitigated by restricting attention to sub-regions of the SST where abrupt variations in SST patterns may reflect underlying physical processes (candidates include the Eastern Maine Coastal Current, the Jordan and Georges Basin gyres, and along important fronts).
A statistically robust estimate of timescales of SST pattern evolution and their degree of regional variation will provide more confidence for using satellite-observed SST as a diagnostic tool. By evincing the temporal and spatial scales of SST image patterns, we can improve future field sampling efforts to capture features like transient or seasonal fronts, coastal plumes and eddies, which may make important contributions to the Gulf of Maine’s biological productivity. By improving our ability to use SST images as a diagnostic tool, this work may augment studies underway here at Univ. Of Maine and affiliated institutions. For example, monitoring of the Penobscot Bay outflow, red-tide algal bloom studies, studies of EMCC and Georges Bank GLOBEC projects. As the launch date of the SeaWifs ocean color sensor approaches (late spring, 1997), it is possible that the products of this work will also improve interpretation of future patterns of satellite observed chlorophyll pigment in the productive Gulf of Maine.
Tools and Computational Techniques
The School of Marine Science owns and operates a TeraScan satellite receiving station built by Sea space Corp., San Diego, Ca. This powerful tool is being used to receive and archive satellite SST images from the GOM region. The dish receives a telemetry signal from NOAA satellites passing within a range specified by the user. Processing of the imagery involves extracting the five channel information from the Advanced Very High Resolution Radiometer (AVHRR) and converting it to a surface temperature (in degrees Celsius) using a multiple channel sea surface temperature algorithm (MCSST) method. All images from polar-orbiting satellites are distorted due to earth curvature, earth rotation, spacecraft attitude (pitch, roll, and yaw), optical sensor limitations, and small perturbations in satellite orbit. These small errors contribute to offset of computed pixel location relative to fixed geodetic point on earth’s surface. The effect of this is seen where coastline overlays (based on actual geodetic position) show an offset from pixels which are supposed to represent those points. After initial correction of an image, using up to date orbit ephemeris, further corrections are carried out by linear correction to selected ground control points, improving the image to map registration (Emery and Ikeda, 1984). This navigation process is critical for using images to determine correlation timescales of features like a thermal fronts, which may extend across only a few pixels. Cloud cover can be removed from the image using correction algorithms. Calibration of the temperature is done with up-to-date calibration factors from the satellite telemetry, ensuring a minimal errors in SST. Pixel resolution depends on satellite orbit relative to the dish location, with the best resolution at about 1.1 km square per pixel. Based on our experience, we receive at least 4 of these high angle, best resolution passes per day.
By summer of 1997 we will have a complete year of archived daily images. This process will continue throughout the next few years. The invariable gaps in our data set can be supplemented by daily imagery from NOAA’s Coastwatch facility and by newly reprocessed SST imagery from the Pathfinder mission for 1985-1991 (Thomas, pers. comm.). The software to run the Sea space Terascan system includes powerful computational tools, which will be used for this project.
A number of techniques have been applied successfully to extract spatial and temporal information from SST imagery (e.g. Chelton and Schlax, 1990; Denman and Freeland, 1985; Deschamp et al, 1981). Temporal scales of variability can be determined for each sub-region identified in the Gulf of Maine without sacrifice of original high-resolution of SST features. Cross-correlation between successive images is sensitive to feature shape, size and orientation and can be used to measure preferred orientation and length scale of sub-image features. Temporal changes in image features will degrade the cross-correlation. From the irregular sequence of SST images over the course of a year, or a season, we can assemble image pairs with a range of lags in time. We then compute the cross-correlation between image pairs with increasing lag. This yields a measure of decorrelation time, or the maximum length of time that image features persist over the entire record length. This technique has been successfully used by Abbot and Denman (1994) to characterize the relationships between sea surface temperature and pigment images from a dynamic upwelling region along the west coast. They found that pattern decorrelation timescales within an upwelling jet were twice as long as those outside the jet and demonstrated that pigment patterns lagged thermal patterns in a manner consistent with a developing bloom of phytoplankton in the nutrient-rich water. The decorrelation timescale for selected sub-regions of the Gulf of Maine may distinguish preferred scales of motion, and suggest the relevant controlling physical mechanisms. For example, the timescales of geostrophic eddy will be longer than those of transient frontal regions.
Objective measures of spatial extant of image features can be made by use of the spatial structure function. The spatial structure function for any variable (for example, temperature), represents as a mean square difference the statistical influence of one point on other points at a distance h (Deschamp et al, 1981). Features that tend to be grouped into specific spatial scales produce a relative peak in the function at that lag h. (Denman and Freeland, 1985). This method has been successfully applied to estimates of thermal patterns and mesoscale (10-100 km) variations in geopotential height, temperature and chlorophyll. Using the structure function, Denman and Freeland showed the existence of a 60 km diameter stationary eddy off Vancouver Island, which was present throughout a series of sporadic research cruises. Identifying the preferred spatial scale of patterns in SST data, for different regions of the Gulf, may provide a deep insight into the interpretation of satellite data during periods when it is not possible to collect data at sea. In addition, applying this technique to sub-regions and determining the largest spatial extent of features that evolve in the areas of interest should reveal the best size of sub-regions for the temporal cross-correlation analysis.
Proposed benefits to State of Maine and NASA
We are, as yet, unable to place many of the fine-details in SST image features into their proper physical context. In the past, oceanographic field surveys have suffered from an accepted degree of under sampling with respect to the transient and abrupt nature of patterns seen in satellite SST images. This has placed fundamental limitations on our ability to interpret these features in their proper oceanographic context. A statistically robust estimate of timescales of SST pattern evolution and their degree of regional variation will increase confidence for using satellite-observed SST as a diagnostic tool. Quantifying decorrelation timescales and dominant spatial scales of features in Gulf of Maine SST images will help develop better relationships between observations by ship-board survey and the features observed in large-scale satellite images. This work may also support studies underway here at University of Maine on of circulation in Penobscot Bay, red-tide algal bloom studies, further studies of EMCC, and Georges Bank. The products of this work may also improve interpretation of future patterns of satellite observed chlorophyll pigments.
The State of Maine has recently committed to supporting research in the marine sciences. This project will help to foster the continued growth of a University-operated remote sensing capability with specific application to marine research in the Gulf of Maine. Results of this and related marine research projects will advance our comprehensive knowledge of the Gulf of Maine’s complex circulation and its renowned productivity. It may improve our ability to forecast the oceanographic "weather" in the Gulf of Maine’s circulation system, and determine the pathways and fates of nutrient-rich water, which has important consequences for site selection of aquaculture industry. As fisheries practices continue to evolve, we may be able to better use SST imagery to identify and study preferred habitats and spawning grounds of fish stock.
NASA’s commitment to the Mission to Planet Earth emphases global climate processes and their underlying mechanisms, over a broad range of time and space scales. Study of the decorrelation timescales in SST images provides a basis for unifying large, global scale satellite imagery with observations of shelf-scale circulation features. The Gulf of Maine is uniquely suited to this study, because it’s complex circulation is strongly influenced by local physics of the continental shelf and by the proximity of the offshore Labrador and Gulf Stream currents, who’s variability is inherently linked to global processes.
Bigelow, H. B., Physical Oceanography of the Gulf of Maine, Department of Commerce, Washington, D.C., 1927
Bisagni, J. J., Gifford, D. J. and C. M. Ruhsam, The spatial and temporal distribution of the Maine Coastal Current during 1982, Cont. Shelf Res. 16, 1, 1996
Bisagni, J. J., R. C. Beardsley. C. M. Ruhsam, J. P. Manning, and W. J. Williams, Historical and recent evidence of Scotian Shelf water on Georges Bank, Deep Sea Res., 43, 1439, 1996
Brooks, D. A., The influence of warm-core rings on slope water entering the Gulf of Maine, J. Geophy. Res.92 (C8), 8183, 1987
Chelton, D. B. And M. G. Schlax, Estimation of time-averaged chlorophyll concentrations from irregularly-spaced satellite observations, J. Geophys. Res. 95, 1990
Denman, K. L. And M. R. Abbott, Time scales of pattern evolution from cross-spectrum analysis of AVHRR and CZCS imagery, J. Geophys. Res.99(C4), 7433, 1994
Denman, K. L and H. J. Freeland, Correlation scales, objective mapping and a statistical test of geostrophy over the continental shelf, J. Mar. Res., 43, 517, 1985
Deschamps, P.Y., Frouin, R. and L. Wald, Satellite determination of the mesoscale variability of the sea surface temperature, J. of Phys. Oceanogr. 11, 864, 1981
Emery, W. J. and M. Ikeda, A comparison of geometric correction methods for AVHRR imagery, Can. J. Rem. Sens., 10, 1, 46, 1984
Pettigrew, N. R. and R. D. Hetland, The cyclonic gyre of the eastern Gulf of Maine, Abstract GLO-5, Proceedings of the Oceanography Society, 1995
Pettigrew, N. R., D. W. Townsend, J. P. Wallinga, P. J. Brickley, and R. D. Hetland, Observations of the Eastern Maine CoastaL Current and its offshore extensions in 1994, J. Geophys. Res (submitted), 1997
Townsend, D. W. and N. R. Pettigrew, The role of frontal currents in larval fish transport on Georges Bank, Deep Sea Res. 43, 1773, 1996