Verification of NCEP Marine Wind Forecasts for the Isle of Shoals C-MAN During Gale and Marine Storm Events

Steven J. Capriola

National Weather Service Forecast Office

Portland, Maine


Through experience, forecasters at WSFO GYX have found that NCEP generated marine wind forecasts (AFOS and N-AWIPS product MRPCWN) have routinely underforecast wind speeds at the Isle of Shoals (IOSN3) site, in particular during strong wind events. Strong wind events included gales (winds 34 knots to 47 knots) and storms (winds 48 knots or more).

This paper will verify MRPCWN wind forecasts at the IOSN3 for observed gale and storm events. Since gale and storm warnings are preferably issued for the second forecast period (ie a 12 to 24-hour forecast), the MRPCWN guidance used will be for this time frame. This paper will verify for type of event (defined by wind direction), cycle forecast (00Z or 12Z) and season (warm vs cool). Data was collected from October 1997 through May 1998 which included the 1997-1998 winter (cool) season and part of the 1998 warm season.

Another concern to local forecasters is the timely availability of data and the availability of a complete data set. Hourly data of wind direction, speed and gusts are available on the Automation of Field Operations and Services (AFOS) and the National Advanced Weather Interactive Weather Processing System (N-AWIPS), however other technology, such as the Internet (NWS Tallahassee, FL), provides this data in a more timely fashion as well as providing 10 minute winds and peak hourly wind gusts not available in AFOS (and N-AWIPS). This paper will also compare the peak winds reported in AFOS vs the peak winds available via the Internet.

Finally, this paper will not attempt to verify warnings issued (or not issued when needed) by the NWSFO or to verify model forecasts (NGM in particular).


The Isle of Shoals platform is a Coastal-Marine Automated Network (C-MAN) type using the Data Aquisition, Control and Telemetry (DACT) payload system, developed in 1981. This C- MAN is located approximately 0.1 deg lat (6nmi) off the New Hampshire coast (42 deg 58 min N LAT 70 deg 37 min W LON - see Fig. 1) and was installed in September, 1984. The site elevation is 10.7 meters (43 feet) ASL. The height of the wind sensor is 19.3 meters above the site elevation or about 30 m ASL.

The National Weather Service (NWS) standard wind sensor height is 10 meters. The 10 m wind is the best estimate of the surface wind (usually when the wind sensor height is above 20 m). Winds are adjusted to 10 m and 20 m elevations and are reported along with observed winds at wind sensor elevation. However, the difference in wind speed from site elevation to 10 m is usually only 1 or 2 knots and can often be zero for the Isle of Shoals.

Wind direction and speed is averaged over a 2 minute interval for the hourly reports. They are available at 10 minute intervals via the Internet (but only reported for the 10 minute interval ending at the top of the hour in AFOS). Wind gusts are a peak 5-second gust measured during the 2 minute interval and peak wind gusts are the maximum 5-second wind gust during the measurement hour, reported at the last hourly 10-minute segment.

The reporting resolution for wind speed is 2 knots (for real time sustained wind and gusts). The reporting accuracy for wind direction is +/- 10 deg (true) and for wind speed the reporting accuracy is +/- 2 knots.

Data quality control is performed by the National Data Buoy Center (NDBC). Data is updated hourly in AFOS and contains the sustained (average) wind and peak wind gust during one 10 minute interval (H+50 to H+00) and is available about half an hour after observation time (H+00). On the Internet data is updated hourly as well with a sustained and peak wind during the latest 10 minute interval (H+50 to H+00) but also includes sustained winds for the other 5 10-minute segments during the hour plus a continuous peak wind for the entire hour. Data availability varies from under a half hour to on a rare occasion more than an hour after observation time.


The Nested Grid Model (NGM) based coastal wind guidance is available twice daily (0000 and 1200 UTC) for 6 hour increments out to 48 hours for selected coastal C-MAN and buoy sites (NWS Tech. Procedures Bulletin No. 390).

Development of the guidance followed these steps:

$ Used a single station approach to forecast winds, i.e. a set of equations was developed for each site. These equations were also derived for each cycle and for the warm (April- September) and cool (October-March) seasons.
$ Assumed the Limited Area Fine Mesh (LFM) 00-hour analyses and 06-hour projected data were perfect and related them to the appropriate observed data to develop forecast equations.
$ NGM model data was applied to these equations to determine forecasts from 6 to 48 hours.
$ Predictand data included wind speed and direction from records over the period from January 1978 through December 1988 (only available since September 1984 for IOSN3) and stratified into warm and cool seasons.
$ Predictor data included time of year and 1000, 850, and 500 MB data which included: u- and v- wind (and upper level geostrophic )wind component, wind speed and relative vorticity plus 500 MB 6-hour height change.
$ There was no inflation of wind speed forecasts to adjust for extreme wind conditions.


The data was verified for a 12 to 24-hour MRPCWN forecast (ie for the second forecast period) for winds (sustained or frequent gusts) of 34 knots or greater (gale force or stronger) for a window of the strongest observed winds during the event (a selected hour plus/minus 3 hours). The data was also verified for the cycle (0000 UTC vs 1200 UTC), season (cool (October through March) vs warm (April through September)), and storm type. Storm type was defined by wind direction. This included northeast (NE), northwest (NW), southeast (SE) and southwest (SW) which accounted for (though not necessarily in every case) offshore flow (NW) vs onshore flow (SE), cold air advection (NW) vs warm air advection patterns (SW and SE) and rapidly developing cyclones (NE). On occasion the time of the highest forecast wind did not coincide exactly with the window of observed highest winds, but was verified just the same. This took into account small (a couple hours or less) temporal errors sometimes found in model wind forecasts.

A multiplication factor was calculated that could be applied to the MRPCWN guidance to determine a forecast for sustained winds and the maximum (hourly) gust (as reported in AFOS).
A percentage difference in the maximum (hourly) gust (AFOS) and peak wind for the event (Internet) was calculated to determine a percentage to apply to the maximum (hourly) gust to predict a peak wind for the event. Finally the data was plotted on a forecast vs observed graph to determine a best fit line.

5. EXAMPLE - 28 MARCH 1998

On this date the observed highest winds were during the 6 hour period from 4 PM to 10 PM (7 PM plus/minus 3 hours which included 7 hourly observations). Observations showed a south wind averaging 28.0 knots, a maximum hourly gust of 31 knots and a peak wind of 35 knots.

The MRPCWN product from the 0000 UTC cycle on the 28th showed a 24-hour forecast (valid for 7 PM on the 28th - 0000 UTC 29 March) of a southerly wind (180 degrees) at 12 knots.

The multiplication factor was determined by the observed wind speed divided by the forecast wind speed for both the average observed wind and the maximum gust, ie

Average observed / forecast = 2.3 Maximum gust / forecast = 2.6

The percentage difference from max gust to peak wind was calculated by subtracting the max gust from the peak wind then dividing by the max gust. This value is then multiplied by 100%.

((35 - 31) / 31) X 100% = 12.9 %


Based on 29 events from October 1997 through May 1998 during expected gale or storm situations the 12 to 24-hour MRPCWN forecast for the IOSN3 can be adjusted by multiplying the MRPCWN forecasted wind speed by 1.5 to determine an expected sustained wind and multiplied by 2.0 to determine an expected maximum (hourly) wind gust (see table 1). On average, adding 12.8 percent to the maximum hourly gust forecast will determine a peak wind for the entire event.

Nearly half (41.4%) of the 29 events were NE. Most events were either NE or NW with only about 1 in 4 (24.1%) either SE or SW. The warm season events were nearly equally divided between NW, NE and SE with zero events SW.

Model cycle time, season, and storm type (i.e. wind direction) generally made no difference in the results with the following exceptions. During the warm season the guidance underforecasted wind speed slightly more than in the cool season. Also, for SW wind events higher multiplication factors are needed to determine sustained wind and wind gusts forecasts. In both these cases the data sets were very small. It was noted during the gathering of results that as the sample size increased the variability in results decreased. An even larger sample size would be expected to lower the variability even more. Since the results showed small variability for cycle time, season and storm type the overall results could be used for any type event for either season or cycle time, assuming a larger data sample would continue the trend toward less variability in the results.

Individual results showed some variability. The largest difference between observed and forecast wind speeds occured on 23 April 1998 when the multiplication factors were 2.4 and 2.9 for sustained winds and max gusts, respectively. The least difference occurred on 29 December 1997 with multiplication factors of 1.1 and 1.4 for sustained winds and max gusts, respectively. Interestingly both these events were NE.

The results for sustained winds were plotted (Fig. 2). The data was analyzed to determine a best fit line, which had a slope equal to the multiplication factor (1.5). Of the 29 plotted events, 13 were above the best fit line, 13 below the line and 3 on the line. This graph can be used to calculate expected sustained winds, given an MRPCWN wind forecast. The same could be done for wind gusts.

Multiplication factors and percent adjustments applied to forecast winds for given observed winds, hourly maximum wind gusts and event peak winds.

Number Multiplication. factor Percent increase from
of events average wind / max gust hourly max to peak gust.

By cycle time:

00Z 18 1.6 / 2.0 10.2 %
12Z 11 1.5 / 2.0 17.1 %

By season:

Cool 21 1.5 / 1.9 12.0 %
Warm 8 1.7 / 2.2 14.9 %

By storm type:

SE 5 1.5 / 1.9 18.3 %
NE 12 1.6 / 2.0 13.0 %
NW 10 1.5 / 2.1 10.2 %
SW 2 1.9 / 2.2 9.7 %

All events:

Total 29 1.5 / 2.0 12.8 %

Table 1. Multiplication factor (applied to MRPCWN forecast) to determine observed average winds and maximum hourly wind gusts. Percent increase applied to observed maximum hourly wind gusts to determine observed peak wind for each event. Results obtained for events based on season and storm type and for model cycle time.


A ficticious synoptic situation suggests that gale force winds are likely for the coastal waters of Maine and New Hampshire. The guidance for C-MANs MDRM1 and MISM1 supports it. The forecast for the IOSN3, as is usually the case, forecasts much weaker winds.

The MRPCWN forecast for IOSN3 was a northwest wind at 18 knots (3218). Using the multiplication factors of 1.5 and 2.0 (for all type events) for sustained winds and maximum hourly gusts, respectively gives sustained winds of 27 knots (1.5 X 18 knots), frequent gusts of 36 knots (2.0 X 18 knots) and a peak wind (considered occasional gusts) of 41 knots (36 knots plus 12.8% of 36 knots). This results in a forecast of a NW wind at 27 knots with freqent gusts of 36 knots and a peak wind of 41 knots.

The adjusted (CWF) forecast, if using only the IOSN3 forecast, could be written as a northwest wind 25 to 35 knots with occasional higher gusts (up to 40 knots) or a northwest wind 20 to 30 knots with frequently higher gusts (of 35 to 40 knots). Either of these forecasts would require the issuance of a gale warning for the coastal waters at the Isle of Shoals based on NWS guidelines.


On 23 January 1998 a marine storm warning verified for the coastal waters of Maine and New Hampshire. The MRPCWN guidance 18-hour forecast from the 1200 UTC cycle on 23 January 1998 was for an east southeast wind (110 deg) at 28 knots valid for 0600 UTC 24 January 1998 for the Isle of Shoals. This is a stronger wind forecast for this location than is typically seen, but still falling just short of gale warning criteria. However, applying the results from this study would prompt a forecaster to issue at least a strong gale warning, if not a storm warning. The calculations give a sustained wind forecast of 42 knots (1.5 times 28 knots), hourly wind gusts of 56 knots (2.0 times 28 knots) and a peak wind of 63 knots (56 knots plus 12.8% of 56 knots). A forecast for this location would be on the order of 35 to 45 knots with either occasional or frequent gusts of 50 to 60 knots.

The observed sustained winds were no stronger than what was forecast and gusts were not much higher (table 2). Typically with coastal storms a coastal front develops over the coastal waters of Maine and New Hampshire, extending north or northeast from the parent intensifying low pressure system to the south. The coastal front (Bosart et al, 1972 and Bosart, 1975) delineates warm air and east, southeast or south winds on the east side of the pressure trough from cold air and northeast, north or northwest winds on the west side. The coastal front can move west (inland) or east (drift further out to sea) depending on the track and intensification of the asociated low pressure system and the strength and location of the surface high pressure system to the north.
During this event a coastal front did form over the coastal waters of Maine and New Hampshire. It is evident from the observed data that the coastal front moved east of the IOSN3 during the event (between 0500 and 0600 UTC), just prior to the onset of the strongest winds. Observations to the east (and east of the coastal front) over the coastal waters of Maine (MDRM1 and MISM1) verified the very strong winds that reached storm force (tables 3 and 4).

Observed winds from the IOSN3 from 0200 UTC to 1000 UTC 23 January 1998

Winds (in knots)
Time (UTC) Wind Direction Sustained Gusts Peak (Direction and speed)

0200 E (090) 27 29 E 34
0300 E (090) 26 27 E 33
0400 E (090) 21 21 E 29
0500 E (080) 16 16 E 23
0600 NNE (030) 12 14
0700 ENE (070) 14 16 N 26
0800 SE (140) 22 23
0900 N (360) 15 17 SE 26
1000 N (010) 16 17

Table 2. Observed winds from the IOSN3 buoy from 0200 UTC to 1000 UTC 23 January 1998. Wind speed in knots, wind direction using 8 points of the compass and true direction.

Observed winds from the MISM1 from 0200 UTC to 1000 UTC 23 January 1998

Winds (in knots)
Time (UTC) Wind Direction Sustained Gusts Peak (Direction and speed)

0200 ESE (110) 34 37 ESE 42
0300 E (100) 37 41 E 46
0400 E (100) 42 47 E 52
0500 ESE (110) 44 49 E 54
0600 ESE (110) 43 49 ESE 55
0700 ESE (120) 46 54 ESE 58
0800 SE (130) 47 52 ESE 61
0900 SE (130) 40 45 ESE 60
1000 SE (140) 40 45 SE 49

Table 3. Same as table 2 but for the Matinicus Rock (MISM1) buoy.

Observed winds from the MDRM1 from 0200 UTC to 1000 UTC 23 January 1998

Winds (in knots)
Time (UTC) Wind Direction Sustained Gusts Peak (Direction and speed)

0200 ESE (110) 29 31
0300 ESE (120) 31 33
0400 ESE (110) 32 35 ESE 39
0500 ESE (110) 33 35 ESE 41
0600 ESE (120) 35 39 ESE 47
0700 ESE (120) 45 51 ESE 52
0800 SE (130) 42 49 ESE 57
0900 SE (130) 42 49 SE 57
1000 SE (140) 37 41 SE 53

Table 4. Same as table 2 but for the Mount Desert Rock (MDRM1) buoy.


Forecasters at NWSFO GYX have known for years that the MRPCWN guidance for IOSN3 have always underforecast observed winds, in particular for significant wind events. This paper used 29 cases from the 1997-1998 season to verify wind forecasts for the IOSN3 buoy.

The results showed that forecasters could multiply the 12 to 24-hour MRPCWN guidance by 1 2 to determine expected sustained winds and to double the MRPCWN guidance to calculate expected (frequent) wind gusts for the second period of the coastal waters forecast. Peak storm winds would be, on average, about 13 percent higher than the hourly maximum wind gusts.

The cycle time of the guidance and the storm type made no difference in the results. However, the warm season may require larger multiplication factors. Additional case studies during the warm season are needed to determine if this is indeed the case.

During the research it was noted (but not reported here) that even a 6 to 12-hour forecast still required similar multiplication factors for shorter range (first period) forecasts.

This paper did not look into possible causes for the errors in the MRPCWN wind speed forecasts for the IOSN3 C-MAN. Possible reasons include equation development, changes in the models over the years, changes in observing equipment (or change in the height of the sensor elevation), the validity of observations during equation development, the length of record used in equation development, the sensor elevation not being representative of true surface (10 m) winds, or inherent errors in the NGM model to properly forecast a parameter(s) used in the equations for the MRPCWN forecast. Regardless of the reason (s), operational forecasters can use these results to drastically improve upon guidance, with a high degree of confidence, when forecasting significant wind events for the coastal waters of Maine and New Hampshire.


Bosart, L. F., 1975: New England Coastal Fronitogenesis. Quart. J. Roy. Meteor. Soc., 101,

_____, C. J. Vaudo, and J. H. Helsdon, Jr., 1972: Coastal Frontogenesis. J. of Appl Meteor.,
11, 1236-1258.

Coastal-Marine Automated Network (C-MAN) NWS Users Guide. National Data Buoy Center,
Oct. 1988. NOAA, DOC, 47 pp.

NWS Tallahassee, FL Internet Homepage. Interactive Marine Observations for the Isle of Shoals
Buoy. Http address: Awww.".

NWS Technical Procedures Bulletin No. 390 (Feb. 22, 1991). Coastal and Offshore Wind Guidance Along and Near the Conterminous United States Coast and Alaska Coast.