SEPTEMBER 1985 NOTES AND CORRESPONDENCE 401NOTES AND CORRESPONDENCESea Surface Flow Estimation with Infrared and Visible Imagery ANDREW C. VASTANO Department of Oceanography, Texas A&M University, College Station, TX 77843 STEPHEN E. BORDERSSlO/Satellite Oceanography Facility, University of California, San Diego, La Jolla, CA 92093 RUTH E. WITTENBERG Visibility Laboratory, University of California, San Diego, La Jolla, CA 92093 26 October 1984 and 11 January 1985ABSTRACT Sequential Nimbus-7 CZCS infrared and visible images obtained on orbits 3157 and 3171 during 9-10 June1979 have been used to derive sea surface flow from advective sea surface pattern displacements and elapsedtime. Individual analyses with infrared (11 microns) and visible blue/yellow ratio (0.443 and 0.550 microns)pairs of images yielded coherent velocity distributions over an oceanic region near Georges Bank. A compositeof eighty flow vectors illustrates a seaward diversion of cold surface water off Northeast Channel, Gulf of Maineby a northeastward intrusion of Gulf Stream water along the continental slope, These results demonstrate thatinstances arise when infrared and visible surface pattern changes can be used jointly to compose flow regimes.A sea surface topography map derived from the composite vector distribution has a range of 20 cm and anexpected repeatability of 0.39 cm.1. Introduction Satellite observations can yield quantitative measurements pertinent to mesoscale dynamics and studiesof the total time rate of change of scalar properties ofthe oceanic surface layer. Surface flow estimates canbe extracted from data gathered by a number of sensorsand are a topic of current research. Radar altimetricsensings of the surface result in surface topographicmeasurements adequate for flow computations. However, the narrow sensor footprint and orbital characteristics have heretofore produced inadequate arealcoverage and aliasing of moving mesoscale features.The NOAA (AVHRR)~ and Nimbus (CZCS) satellitesensors can provide the necessary spatial and temporalresolution to define mesoscale features. The patterndisplacements apparent in sequential AVHRR andCZCS scenes have revealed motion and led to simplemeasures of flow (e.g., Vastano and Bernstein, 1984).Image processing algorithms that provide quantitativesea surface motion estimates are a recent development(Kelly, 1983; Vastano and Borders, 1984). Kelly hasused temperature fields computed from AVHRR datato derive both cross- and along-isotherm componentsof motion and compose a total vector field. ThisAdvanced Very High Resolution Radiometer.method is an automated one after small feature displacements in sequential images are used to evaluatea free parameter in an analytic model of the velocityfield. A uniformly computed vector field can be produced over the image by this method. Vastano andBorders have applied an interactive algorithm that requires sequential images and objectively identifies smallfeature displacements in the direction of the temperature gradient. The result is a nonuniform field of vector estimates. Both techniques have been applied toNOAA-7 AVHRR images with successful productionof sea surface vector fields that represent horizontalmotion. Infrared or visible image analyses can be hamperedby the presence of atmospheric interference. Cloudsover a target area often partially or completely obscureviews of the sea surface and mesoscale features. However, observations by more than one polar-orbitingsatellite can take advantage of atmospheric motion thatmay clear clouds during the interval between satellitepass times. Whether in order to alleviate interferenceproblems or to increase the temporal resolution withina given day, the use of multiple satellites can producea greater number of sequential sea surface patterns. Insuch instances, the patterns from different satellites maybe related to different scalar quantities. This note reports an application of the interactive method of flowvector computation to pairs of coincident CZCS inc 1985 American Meteorological Society402 JOURNAL OF ATMOSPHERIC AND OCEANIC TECHNOLOGY VOLUME2frared and visible images and presents the resultingcomposite flow estimate. The objective is to show therelation between infrared and visible vector flow estimates.2. Sequential image analyses Nimbus-7 CZCS scenes of the region immediatelyeast of the Gulf of Maine and south of Nova Scotiawere taken on orbits 3157 and 3171 during 9-10 June1979. The 11-micron infrared channel and two visiblechannel [0.443 (blue) and 0.5513, (yellow) microns] images for each scene were registered to within 0.6 km,yielding 512 x 512 pixel images (423 x 423 km) forflow analyses. Sequential pairs were formed by the infrared images and computed, visible ratio images (blue/yellow). The patterns exhibited in the images correspond to the sea surface temperature distribution (infrared) and an uncorrected approximation to the pig /!FIG. 1. Flow vectors and dynamic topography overlayed on the infrared image for 9 June 1979.SEPTEMBER 1985 NOTES AND CORRESPONDENCE 403ment concentration or chlorophyll distribution (blue/yellow). Similar pattern shapes, structures and displacements were apparent in these four images. Thealgorithm for flow vectors (Vastano and Borders, 1984)was applied to the infrared and visible pairs and produced a total of eighty estimates. In order to provideindependent assessments of pattern displacement andvector estimates, each pair was treated separately bydifferent investigators. Figure 1 presents the results ofthe flow computations overlayed on the infrared imagefor 9 June 1981. The vector estimates from the infraredimage pair (red) and the visible image pair (yellow) canbe visually compared and jointly form a coherent setthat indicates the intrusion of Gulf Stream water alongthe continental shelf east of Georges Bank. Immediatelyeast of the intrusion, cooler waters move west-southwest and southwest over the one-day interval betweenscenes. The composite set of infrared and visible imageshas been used to compute the associated streamfunction representation (Vastano and Reid, 1985). Thismethod assumes a nondivergent flow approximationand trigonometric streamfunction expansion basisfunctions. The contours shown in Fig. 1 are sea surfacetopography derived from the streamfunction by takinga geostrophic approximation. The range of topographyis approximately 20 cm and the expected repeatabilityis 0.39 cm. The similarity of the vector estimates from the infrared and visible image pairs and the coherent fieldthey portray demonstrates that instances arise whenflow estimates based on different types of sensors maybe combined. These results infer that separate flowfields derived from different sensors can be used together to examine the evolution of mesoscale featuresin a region. Experimental verifications of these methodsare in progress. Acknowledgments. The image processing and analyses were carried out at the Scripps Satellite Oceanography Facility, Scripps Institution of Oceanography,La Jolla, California. Andrew C. Vastano was supportedby Grant OCE 80-26037 of the National ScienceFoundation and the Office of Naval Research underContract N00014-75-0537.REFERENCESKelly, K. A., 1983: Swirl and plume or application of statistical methods to satellite-derived sea surface temperature. CODE Tech. Rep. No. 18, SIO Ref. No. 83-15, 210 pp.Vastano, A. C., and R. L. Bernstein, 1984: Mesoscale features along the First Oyashio Intrusion. J. Geophys. Res., 89, 587-596.--, and S. E. Borders, 1984: Sea'surface motion over an anticyclonic eddy on the Oyashio Front. Remote Sens. Environ., 16, 87-90.--, and R. O. Reid, 1985: Sea surface topography estimation with infrared satellite imagery. J. Atmos. Ocean. Tech., 2, 393-400.