StormWatch 97-98

1. Purpose

StormWatch is a comprehensive effort to explore the potential of spaceborne SAR to yield high resolution wind field estimates in the coastal zone. The project is built around a systematic collection of RADARSAT imagery along the NE coast of North America from Cape Hatteras to the Grand Banks, but also uses supplementary estimates of surface wind, waves, and (derived) marine boundary layer stability from several readily available sources. By interpreting spatial backscatter variations from calibrated RADARSAT imagery in the context of the supplementary estimates, it is anticipated that SAR will indeed provide the extra information necessary to construct high resolution (order of 1 km) estimates of the wind field all the way to the coast, and even into inland bays and estuaries.

2. 1992 ERS-1 Data Set

StormWatch builds on lessons learned from two previous, but smaller, SAR data sets collected over the region between Cape Hatteras and Cape Cod by ERS-1 in 1992 and RADARSAT in 1996-97. Both of these previous data sets were collected to explore the sensitivity of spaceborne SAR to various geophysical parameters such as the wind and wave field and the stability of the marine boundary layer. However, the 1992 ERS-1 data set was severely limited both by the sparsity of auxiliary information, and by the very limited space/time sampling of the ERS-1 SAR. Only a few relatively cloud-free AVHRR images were collected, and the ERS-1 orbit and narrow swath allowed adjacent spatial coverage only every 17 days. Nevertheless, the complete 800 km by 800 km SAR mosaic that resulted provided much insight into the effects of various geophysical variables on the SAR oceanic signature. The complete (uncalibrated) ERS-1 data set of 87 frames can be accessed at various pixel sizes down to 100 m on the ERS-1 branch of this web site.

3. 1996-97 RADARSAT Data Set

The RADARSAT (narrow ScanSAR) data set of 1996-97 was a substantial improvement over the ERS-1 collection in many ways: the wider SAR swath of 300 km permited adjacent spatial coverage every 3 days; temporally averaged AVHRR imagery routinely produced at JHU/APL substantially alleviated the cloud contamination problem; a high spatial resolution (one-degree grid) wind and wave analysis was provided routinely every 6 hours by FNMOC. Since the SAR surface signature is influenced not only by the surface wind field, but also by the large waves, local bathymetry, wave-current interactions at the edge of major currents, and the stability of the marine boundary layer, an accurate colocation of all these auxiliary data sets with respect to the SAR is an essential aid to the interpretation of SAR imagery. A colocated data set with the Radarsat 96-97 (uncalibrated) imagery compiled at JHU/APL can be accessed on the RADARSAT branch of this web site.

4. Expanded Areal Coverage of Stormwatch

Stormwatch is much more comprehensive than the two aformentioned SAR data sets in several important ways. With a wider swath (450 km), more cycles per season (minimum of six), and more passes per cycle (at least ten), the seasonal areal coverage is 10 to 20 times the 96-97 RADARSAT data set, and more than 40 to 100 times the 92 ERS-1 data set. (See Figure 1). The probability of observing a severe winter storm is enhanced by even more than the ratio of areas, since the additional coverage to the northeast will sample a region of higher wave climate. In the 96-97 data set, surface winds exceeded 10 m/s on only 2 passes out of 12; in StormWatch, we might expect many more such passes, mostly to the northeast, as the winter storms develop and mature. We might even reasonably hope for a few occasions when the surface winds will exceed 30 m/s. Such high wind conditions will be useful for exploring the behavior of SAR wind algorithms in extreme conditions.

5. Expanded Auxiliary Data Set accompanying Stormwatch

An improved and expanded auxiliary data set will include a number of additional sources of winds, waves, and air and sea temperature, colocated with the SAR overpasses (every third day), as well as a much denser reference set from GOES 8 (every 30 minutes). Some of these additional sources (see Figure 2) will aid directly in the conversion of SAR backscatter intensity to wind speed. Others will provide independent estimates of the wind field. And still others will help construct the complete picture of the weather patterns occuring over the entire winter. For example, the FNMOC wind direction fields can be used as a first guess to guide the application of conventional scatterometer algorithms. The SAR-derived wind field can then be embedded in independent wind field estimates from ERS-2, the FNMOC WAM analyses, and possibly other sources, all of which may give additional insight about potential biases. Corrections for marine atmospheric boundary layer (MABL) stability can be estimated by using the merged (temporally averaged) AVHRR, the FNMOC SST analysis, and the NORAPS air temperature analysis. GOES imagery and SSMI rain rate fields will be useful for tracking frontal dynamics.

6. Toward an Operational System

The present plan includes the collection of two seasons of StormWatch data: StormWatch 97-98 (01 Nov 97 to 28 Mar 98), and StormWatch 98-99 (27 Oct 98 to 23 Mar 99). Using lessons from both data sets, and acquiring access to near-real time imagery, we hope to be in a good position to generate near-real time high resolution coastal wind estimates within three hours after SAR overpass time. Clearly, there will be errors in the estimates whose magnitude will depend upon knowledge of the "true" SAR wind algorithm, especially upon knowledge of the relative importance of the various "contaminating" influences, and the accuracy of their estimates. Nevertheless, with some skill, we might expect that the accuracies of the resulting SAR wind estimates will exceed anything currently available in near coastal regions, especially if nested within some of the other open ocean fields. One strategy for generating an operational product could make use of 12 hour WAM wind forecasts (rather than analyses) from FNMOC (or NOAA), a recent MABL stability analysis, and an iterative application of a scatterometer algorithm. However, one can imagine alternative strategies that might reduce the potential errors even more. The further development of some of these strategies will no doubt be an active area of research for some years to come.
r.beal@jhuapl.edu
© The Johns Hopkins University Applied Physics Laboratory