QuikSCAT Impacts on Coastal Forecasts and Warnings: Operational Utility of Satellite Ocean Surface Vector Wind Data (2024)

a. WFO operations

Within each WFO, forecast and warning responsibilities are grouped by user needs into so-called desks. The groupings are location specific, depending on regional user community demands, staffing constraints, and meteorological and/or geographical features that might be unique to a given WFO region (e.g., tropical storms, coastal jet phenomena, sea ice extent, etc.). Forecasters are trained for the specific responsibilities at each desk, and they rotate on a locally determined schedule from desk to desk within each WFO. A typical partitioning of forecast and warning responsibilities might include desks with specific focuses on local weather (sometimes called the public forecast desk), aviation, and marine issues. For coastal WFOs with larger than normal areas of responsibility (e.g., Juneau, Alaska; Honolulu, Hawaii; etc.) there can be a “satellite” desk that serves near-real-time (NRT) imagery needs for the other desks within the WFO.

The forecaster shift at a given desk includes several reporting deadlines for issuing analyses, forecasts, and possibly generating, updating and/or canceling warnings. Analyses, forecasts, and warnings are conceived, and often automatically generated from graphical displays in the Interactive Forecast Preparation System (IFPS), which is a tool within the Advanced Weather Interactive Processing System (AWIPS; for an introduction to IFPS, see the Web site www.weather.gov/ndfd/ifps2.htm; for background information on AWIPS, see the Web site www.nws.noaa.gov/ops2/ops24/awips.htm and related links). The AWIPS environment can be customized by each forecaster according to personal preferences and/or as needed for particular meteorological events. Forecasters can chose to overlay analysis and forecast fields with observations from a variety of sources, including SVW data from QuikSCAT. Alternatively, NRT SVW data from QuikSCAT are displayed on another monitor, adjacent to the AWIPS station. In this alternative case, the NRT SVW data are usually accessed from the NOAA/National Environmental Satellite, Data, and Information System (NESDIS) Web site (http://manati.star.nesdis.noaa.gov). The NRT SVW data are made available on the Web site within 3 h of a QuikSCAT overflight. An additional latency might be introduced in the transfer from NOAA/NESDIS to AWIPS displays at coastal WFOs. However, this potential delay did not arise as an issue in surveys and observations of WFO operations.

In IFPS, forecasters work with interactive graphical displays to adjust isolines on maps handed forward from the previous forecast shift. Gradients are sharpened or weakened to provide better agreement with current observations and the latest NWP forecasts. Once a map has been manually updated, an automatic text generator is used to provide an initial interpretation, and the forecaster edits this text before releasing the given product (e.g., a zone forecast, a 1-day marine forecast, or warning, etc.) to the public. The overlay–edit–text generation cycle runs continuously at each desk.

b. SVW data

Specifics of scatterometer SVW retrievals are carefully elucidated for the SeaWinds scatterometer that operates on QuikSCAT in the Algorithm Theoretical Basis Document by Freilich (1996). In broad overview, radar signals of precisely known properties (e.g., frequency, amplitude, polarization, and geometry with respect to the ocean surface target) are emitted from a spacecraft in low-earth orbit. These signals are scattered by gravity–capillary wave roughness elements on the ocean surface that are known to be in equilibrium with the shear stress imparted on the ocean by the surface wind. The backscatter signals are detected at the spacecraft (so-called radar backscatter cross sections, σ_o) and carefully referenced to earth surface locations from which they emanate. Composites of the georeferenced σ_o (i.e., over so-called wind vector cells or WVCs) are used to retrieve SVW data via a geophysical model function. The area of the WVC impacts the number of σ_o entering each SVW retrieval, and this defines the spatial resolution of the SVW data. The QuikSCAT data have been provided at 25-km resolution for the duration of the mission, and most applications make use of QuikSCAT data at this resolution.

A 12.5-km resolution SVW product was introduced experimentally and, as of mid-2006, has become a standard QuikSCAT product as well. Smaller WVC area in the 12.5-km product, and therefore fewer σ_o per WVC, leads to a lower signal to noise ratio in the SVW retrieval. Also, despite higher spatial resolution, the 12.5-km SVW retrievals are not provided any closer to shore than in the 25-km resolution data (i.e., same land mask). This limits the value added from the perspective of the operational user at a coastal WFO.

Details of the SVW retrievals differ for NRT applications (i.e., Chang et al.; information online at http://manati.star.nesdis.noaa.gov/quikscat/nrt_release_notes.pdf) versus science applications (i.e., Ocean Vector Winds Science Team at the NASA Jet Propulsion Laboratory; see the Web site http://podaac.jpl.nasa.gov/ocean_wind/quikscat/L2B/doc). Hoffman and Leidner (2005) provide an excellent review of the NRT retrieval of SVWs from backscatter observations on QuikSCAT. Practical summary descriptions of the SVW retrieval algorithms for both the NRT and science retrieval cases also appear in recent papers with an operational focus in Chelton and Freilich 2005 and Chelton et al. 2006. Minimizing the turnaround time, or data latency period, from overflight to SVW retrieval drives the operational retrievals and product suite development for NRT SVWs. NRT processing algorithms composite σ_o differently than in the case of the science data algorithms; they employ a slightly different rain flag algorithm, and ambiguity removal is based on a different numerical weather prediction product. Otherwise, the vector wind retrieval algorithm is similar to the science data retrieval case (i.e., cf. Figs. 2 and 3), and the error properties of the NRT data are remarkably similar as well. NOAA/NESDIS also produces and posts a 12.5-km SVW NRT product, but at the time of this report, the 12.5-km data were not being provided via AWIPS to the coastal WFOs.

The first panel in Fig. 1 is an AWIPS presentation typical of the SVW data access at most U.S. coastal WFOs. Figure 2 is the corresponding NRT product as it would appear on the NOAA/NESDIS Web site. Figure 3 is a representation of the science data product from JPL (latest release or “R2”). The more commonly used 25-km resolution data are displayed in Figs. 1 –3. Note the rain flag indicators in Fig. 2 (black wind barbs) and Fig. 3 (red SVWs) near 42°N, 125°W. The direction interval retrieval with threshold nudging option (DIRTH; Stiles 1999) in the standard science data product is activated for the sample swath in Fig. 3. The DIRTH option removes small-scale noise that occurs in wind direction retrievals for WVC in the midswath regions. In the AWIPS display (Fig. 1), the NRT SVW retrievals overlay high-resolution (1 km) satellite infrared (IR) imagery that coincides with the QuikSCAT overflight time to within 30 min.

As summarized above, scatterometry is an approach to SVW remote sensing that involves active microwave instrumentation and retrieval techniques. A passive polarimetric approach has been pursued for SVW remote sensing as well. The passive polarimetric technology is less well developed than is scatterometry, and has yet to demonstrate an operational utility. The WindSat instrument on the Coriolis spacecraft was launched in January 2003 with a goal of proving the feasibility and practicality of the passive-polarimetric approach for SVW retrieval from a spaceborne system. The WindSat SVW retrieval algorithms continue to be refined and stabilized more than 4 yr post launch. At the time of our surveys and site visits, WindSat data were available via the NOAA/NESDIS Web site, but not available via AWIPS to the U.S. coastal WFOs. While this is still the case for WFOs, soon after our surveys and site visits were completed, WindSat data became available to the NOAA/National Centers for Environmental Prediction (NCEP; e.g., TPC and OPC) via the NCEP AWIPS (N-AWIPS) system.

Surface wind retrievals from synthetic aperture radar (SAR) sensors will eventually play a role in operational applications at coastal WFOs as well, but it is premature to investigate the operational role of SAR wind retrievals at this time. Operational SAR products (e.g., for ship locations, oil spill and environmental monitoring, surface waves, etc.) are already available for high-latitude environments from U.S., Canadian, and European government agency sources. SAR SVW retrievals exhibit an order of magnitude finer spatial resolution, and more precise determinations of coastal boundaries, than can be achieved from scatterometer data. However, wind direction retrieval algorithms for SAR data require ancillary information (e.g., scatterometer winds). Future international plans for wide-swath and constellation SAR missions will alleviate present-day issues of instantaneous spatial coverage and revisit intervals that limit the operational utility of SVW retrievals from SAR data at mid- and low-latitude coastal locations now.

Written surveys and in-person site visits of U.S. coastal WFOs were conducted over a period from Northern Hemisphere late summer 2005, through the 2005/06 winter season. Surveys were mailed to 33 coastal WFOs, and 16 multiday site visits took place over this period (Fig. 4). The surveys and site visits were designed to measure WFO forecaster familiarity with SVWs, identify forecaster access to SVW data, and to quantify the role satellite SVWs play in local weather and marine analyses, forecasts, and warnings issued by each WFO. A science operations officer (SOO) survey was also mailed to each coastal WFO, and site visits included presentations to the WFO staff to describe the details of the satellite SVW retrievals and SVW validation studies, particularly with respect to rain flags.

In the next section, the survey design and results are described. Impressions from the site visits are documented and discussed in section 3. A summary and recommendations are provided in a final section.

a. Survey design

Multiple choice format surveys were designed for coastal WFO forecasters and SOOs. Prototype designs of the SOO and forecaster surveys were tested in the WFOs in Miami, Florida, and San Juan, Puerto Rico, on the initial site visits in late August 2005. Based on the responses and our interpretations of these sample surveys, minor modifications were implemented and the final form surveys were mailed to 33 U.S. coastal WFOs in October 2005. The multiple choice format was designed to minimize interruptions in the forecasters’ duty cycle. The multiple choice format also facilitates a variety of summaries as demonstrated below.

Design goals for the forecaster survey included the following:

• identifying data delivery and technology transfer pathways for satellite SVWs in the WFO operational environment;

• distinguishing among QuikSCAT and WindSat familiarity and impacts;

• ranking satellite SVW data with respect to other input data used in local weather and marine forecasts and warnings; and

• identifying satellite SVW data limitations and prioritizing potential improvements in the data stream.

b. Response rate

Cooperation from WFO forecasters and SOOs resulted in a 73% response rate by May 2006. Returns of 125 forecaster surveys and 24 SOO surveys accounted for average return rates of about 5 forecaster surveys and 1 SOO survey per responding office. The geographic distribution of coastal WFOs represented in the survey responses, and the variety of offices visited in person (16 WFOs), provide a realistic sample of the diversity of marine forecast situations in which satellite SVWs can be used (see Fig. 4).

c. Pathways

Forecaster introduction to and familiarity with satellite SVW data can be gained via several pathways, including training by the WFO SOO or other forecasters in the local office; review of a Cooperative Program for Operational Meteorology, Education and Training (COMET) training module (Freilich 2007); or via self-education using scattered resources on the Internet. For the Northern Hemisphere fall, winter, and early spring seasons of the survey, satellite SVW data from QuikSCAT were most commonly available to WFO staff via AWIPS, and secondarily via the NOAA/NESDIS Web site (http://manati.star.nesdis.noaa.gov/quikscat).

Five questions on the forecaster survey asked if QuikSCAT and/or WindSat data were used in general, and/or for specific purposes. Table 1 demonstrates that WindSat data were not used at the time of this survey. As described above, the limited access to WindSat data, and the absence of any WindSat operational training materials, rendered this dataset a noncontributor to the results of this study. Thus, unless otherwise specified (e.g., Figs. 5 and 6), the generic term SVW data refers exclusively to the QuikSCAT data for the remainder of this report.

Table 1 also demonstrates that more than half of the survey respondents use QuikSCAT data, at least “sometimes” and more than 15% use it “usually.” Of the usually respondents, the majority used QuikSCAT for sea-state forecasts and warnings, followed by swell forecasts, wind waves, and as a validation of features in NWP forecasts and analyses.

d. Rankings

Forecasters were asked to rank the input dataset’s importance on a numeric scale from 0 to 5, where 5 is “highly important.” Input dataset comparisons with satellite SVW data were made for the combined set of desk responsibilities, but broken down by lead-time category (i.e., “short term,” 1–2 days, and “long term,” 3 days and longer) The input sets listed in the rankings included aviation routine weather reports (METARs) and other surface observations, ship and buoy reports, aircraft reports, radar [e.g., Weather Surveillance Radar-1988 Doppler (WSR-88D)], satellite imagery (e.g., visible and IR images), other satellite data (e.g., soundings, cloud vector winds, etc.), SVWs from QuikSCAT, SVWs from WindSat, numerical weather prediction (NWP) analysis fields (e.g., from the Local Analysis and Prediction System, LAPS; the Rapid Update Cycle Model, RUC; etc.), and NWP forecast fields [e.g., from RUC, the North American Mesoscale Model (NAM), the Global Forecast System (GFS)].

Figure 5 demonstrates that for short-term forecast purposes (i.e., day 1 and 2 products) the most important datasets, in order of number of level 5 rankings (rightmost red bars in each dataset cluster), are local observations from METARs, ships and buoys, satellite imagery, radar, and NWP forecasts. The QuikSCAT data utility for day 1 and 2 forecast products falls in a second tier of importance (i.e., levels 2–4) that includes aircraft and other satellite data, and NWP analyses. Figure 5 shows the rankings for QuikSCAT data for day 1 and 2 products were based on 12 responses at level 5 (red), 14 at level 4 (gold), 38 at level 3 (beige), 21 at level 2 (green), 10 at level 1 (blue), and 11 at level 0 (purple).

The NWP forecasts are, by far, the most important datasets to survey respondents with respect to forecast products for day 3 and longer. Figure 6 demonstrates that the NWP forecasts obtained more than 80 responses at level 5, and no other dataset reached 20 responses in this category. In fact, with 17 responses at level 5, satellite imagery was the only other dataset with over 10 responses in this category. The QuikSCAT data relative importance profile for long-range forecast applications matches those of NWP analyses and all other near-real-time observational datasets (e.g., METARs, ships and buoys, radar, other satellite observations; Fig. 5). Not surprisingly, the operational QuikSCAT data are more useful for short-term forecast and warning applications than they are for the long-term products.

The distributions for the relative importance of the input datasets for special forecast and warning products (not shown) are nearly identical to the distributions in Fig. 5. The special products include marine and aviation forecasts, severe weather forecasts, and warnings. The most important datasets are satellite imagery and local observations from METARs, radar, and ships and buoys. NWP forecasts round out the datasets in the first tier of the importance rankings for specialized products. QuikSCAT data, other satellite data, and NWP analyses compose a second tier of importance rankings for these applications.

e. Satellite SVW data limitations and potential improvements

Specific questions in the forecaster survey sought to identify critical limitations in the data from QuikSCAT in terms of rain flag effects, timeliness of the SVW data, and nearshore coverage. Follow-up questions sought to quantify the importance of potential improvements in these areas, and in improved displays on AWIPS. Two open-ended questions were posed to allow forecasters to input desired improvements for each of the SVW data systems (i.e., QuikSCAT and WindSat).

Figure 7 is a pie chart quantifying survey responses with respect to QuikSCAT data latency limitations. The QuikSCAT swath and orbit configuration is designed to maximize the global ocean coverage in minimum time. The system achieves about 90% coverage of the ice-free global ocean in 24 h. The “maximum coverage in minimum time” design goal impacts the frequency of the sampling at any given location. The revisit interval for QuikSCAT sampling is on the order of 12 h. Given current technology, the revisit interval cannot be reduced beyond this range for a broad-swath scatterometer instrument aboard a single platform in low-earth, polar-orbit orientation. For the brief period ofApril–October 2003, QuikSCAT and the second Advanced Earth Observing Satellite-2 (ADEOS-2) both carried SeaWinds scatterometer instruments and they were in coordinated polar orbits such that the midlatitude revisit interval was about 7 h. ADEOS-2 failed on orbit in October 2003.

Forecasters were asked when QuikSCAT data were too old to be useful for their tasks. Figure 7 demonstrates that data latency is indeed a limitation of the QuikSCAT data. Greater than 80% of the survey respondents found QuikSCAT data to be outdated for their purposes in a time period shorter than the revisit interval. However, it appears that about 50% of the survey respondents would be satisfied with SVW data latencies corresponding to revisit intervals achievable by two, coordinated, polar-orbiting platforms [i.e., O(6 h)].

The SVW retrieval from scatterometer observations of radar backscatter can be confounded by the presence of rain. Rain contamination occurs in the following three ways: 1) the backscatter signal can be attenuated by rain in the atmospheric column between the sensor and the sea surface; 2) conversely, the rain in the atmospheric column can be a source of radar backscatter; and 3) raindrops impacting the ocean surface can alter surface roughness element amplitudes and densities that create radar backscatter. Wind vector cells (WVCs) over which radar backscatter returns are accumulated are flagged for potential rain contamination using a multidimensional histogram algorithm developed by Huddleston and Stiles (2000). Milliff et al. (2004) demonstrate that the rain flag algorithm in what had been the standard science product retrieval (R1) was overly conservative in that useful wind information existed in many rain-flagged WVCs. This has been alleviated in the latest retrieval (R2), and the entire QuikSCAT SVW record is now being reprocessed.

Forecasters were asked if rain flags affected their use of SVW data from QuikSCAT. Figure 8 is a pie chart documenting the response. Only 17% of responders either did not answer the question or stated that rain flags prohibited their use of the data. As such, rain flags do not appear to be a major impediment to the operational utility of satellite SVWs.

The forecasters were asked to rank possible improvements to the satellite SVW data stream. These included SVW retrievals within 10 km of shore, twice the number of SVW retrievals in the coastal WFO region of responsibility, three times the number of SVW retrievals, improved graphic presentation on AWIPS, and additional SVW data on AWIPS. Note that improvements in temporal resolution (i.e., reducing the revisit interval) were not included as possibilities in this question. Temporal resolution issues are dealt with in another question as described above (Fig. 7).

The bar chart in Fig. 9 summarizes the responses. The greatest importance (greater than 60 responses in the “critical” and “very” categories) was placed on nearshore retrievals. Improving graphical displays on AWIPS was next in importance, with more “very important” rankings than “critically important” (Fig. 9). The least important possible improvement was increasing the data density of SVW retrievals (i.e., 2×, 3× higher resolution) within the region of responsibility.

f. SOO surveys

The survey questionnaire for WFO SOOs consisted of nine questions. Two questions were aimed at determining SOO familiarity with SVW data from QuikSCAT and WindSat, and one question concerned SVW data training at the WFO. Several questions asked for SOO impressions of WFO staff usage of satellite SVW data in an overall office summary context. SOOs were also asked to rank the potential importance of possible improvements in data timeliness, coverage, and presentation. Not surprisingly, SOO responses did not differ from the forecaster responses in this regard.

Table 1.

Several survey questions provided information to quantify forecaster usage of WindSat and QuikSCAT data. The responses for these questions were pooled from 125 forecaster surveys to yield the responses tabulated here. More than half of the responding forecasters used QuikSCAT data at least sometimes. WindSat data were not having an impact on forecasts and warnings at U.S. coastal WFOs at the time of the survey. In addition to 275 respondents who answered “no,” 138 respondents chose not to answer any questions regarding WindSat data usage.