The industry's recognized authority for design, engineering and application of equipment and services in the global ocean community
Issue link: http://sea-technology.epubxp.com/i/477967
www.sea-technology.com March 2015 / st 37 determine the error introduced in the results. It was surprising to fnd that the mean error in the predicted cur- rent speed only exceeded 5 centime- ters per second (deemed an accept- able level of error) when 87 percent of the data was randomly removed from the time series. Since missing or removed HF radar data can occur continuously for extended periods of time (days to weeks), additional testing was performed by randomly remov- ing "chunks" of data of various lengths (ranging from one week to months) in addition to random points. This analy- sis indicated that depending on the size and number of missing "chunks," acceptable harmonic analysis results could usually be achieved after remov- ing 60 to 70 percent of the data. Thus, it was decided that a minimum thresh- old of at least 40 percent data return must be achieved over one year for the tidal current predictions to be in- cluded in the Web product for a given grid point. Assessing Harmonic Analysis Quality Another challenge with HF radar current observations is verifying the overall performance of the harmonic analysis and the resultant tidal cur- rent predictions. How can we assure the tidal current predictions are rea- sonable? Although many of the same methods for verifying an analysis from ADCP data can be used, diffculties with HF radar data lie in the large number of time series to analyze (e.g., more than 240 for the lower Chesa- peake Bay) and the added complex- ity of spatial variability in the current feld. A modifed set of spatial and sta- tistical diagnostics were implemented to more easily assess performance over the large number of spatially varying current observations. Some examples of these new di- agnostics used to quality control the results include ellipse plots of the fve major tidal current harmonic constitu- ents and the fraction of the total cur- rent variance accounted for by the tidal current predictions. For instance, grid points with signifcantly differ- ent constituent properties than their neighboring points were more closely analyzed and potentially removed. Ad- ditionally, grid points where the tidal current predictions did not account for at least 50 percent of the total current variance were in most cases removed since these predictions would not pro- vide a very reliable estimate of the total currents. These and other diagnostics were utilized to identify grid points with poor harmonic analysis results, which are then excluded from the fnal Web product. Next Steps By leveraging our tidal current and harmonic analysis expertise, as well as the HF radar data centralization and standardization provided by IOOS, the new surface current Web product described here is now available on the CO-OPS website. However, the release of this product does not signify the end of develop- ment, but rather the beginning. In con- cert with the product release, a robust outreach and communications effort has begun to solicit feedback regarding expansion of the product to additional locations and new technical enhance- ments. Expansion of the product to addi- tional locations can be accomplished with only moderate effort for pre-exist- ing HF radar deployments. New York Harbor was scheduled to be added to the product in the fall of 2014. Poten- tial expansions in the near term could include Long Island Sound, Delaware Bay and southern California, depend- ing on stakeholder interest and re- quirements. Technical enhancements being considered include adding a nautical chart map layer, providing warnings of instances and locations of high, non- tidal residual current, and showing multiple HF radar data resolutions on the same map (e.g., 2-kilometer, 1-ki- lometer and 500-meter resolutions for San Francisco Bay). We are also inves- tigating the inclusion of a statistical model to provide a short-term predic- tion of the nontidal residual current, which could substantially improve the accuracy of near-term (5 to 10 hours) predictions. In addition to product- specifc enhancements, CO-OPS will be exploring the integration of related current products (e.g., HF radar, opera- tional hydrodynamic numerical mod- els and ADCPs) to provide users with a seamless data visualization experi- ence. CO-OPS has been coordinating with a variety of users throughout the development and release of this new HF radar Web product. These include regional harbor safety groups and pi- lots, researchers, recreational boaters info@edgetech.com USA 1.508.291.0057 Clearly Superior Imaging SONAR SYSTEMS SUB-BOTTOM PROFILERS BATHYMETRY SYSTEMS SIDE SCAN SONARS The Leader in Underwater Technology