The final section of the NAEFS webcast will look at NAEFS applications in two case examples. Then I’ll offer a summary of the full web cast.

The final section of the NAEFS webcast will look at NAEFS applications in two case examples: a southern Plains case from August 2008 and a Northeast US case from February 2007. Then I’ll offer a summary of the full webcast.

The first example is a summer heat wave. The graphics are from NCEP’s Climate Prediction Center. Maximum temperatures and anomalies over the CONUS on 3 August are left to right in °F. A hot period in the south-central Plains of the CONUS peaked on 3 August, when temperatures exceeded 100°F over much of the central-to-southern Plains. These temperatures were greater than 10°F above normal. This case study will demonstrate use of NAEFS anomalies in the forecast.

Here are observed 500-hPa heights over North America at 18 UTC on 3 August 2008, contoured at 60 meter intervals, which we’ll compare to the large-scale flow forecasts from NAEFS. We see a strong ridge in the central US bounded by troughs on both coasts, a pattern favorable for heat waves in the center of the country.

We’ll start seven days before 3 August 2008, the longest lead time for which WFOs are responsible.

At 7 days, we look at the large scale; for example, synoptic scale flow at 500-hPa, to obtain situational awareness of the potential for high-impact events in the medium range. We also want to visualize extremes for the flow regime.

These graphics of 10th (upper right), 50th (lower left), and 90th percentiles (lower right) contour the extreme and median flow patterns at 500-hPa, and shade the climate anomalies in meters. Answer the question on the next slide, based on these graphics.

Now examine the 00 UTC 1 August run, with 10th, 50th, and 90th percentile graphics in the same locations. All forecasts are for 18 UTC 3 August 2008. Heights are contouring and coloring is the climate anomaly as indicated in the color bars.

Now, what do 2-m maximum temperature forecasts look like for the 6-hours ending 00 UTC 4 August?

From left to right, these are 1°x1° resolution graphics of the 10th, 50th, and 90th percentile forecasts for NAEFS 2-m maximum temperatures for the 6 hours ending 00 UTC 4 August 2008, in degrees F. Temperatures are contoured every 5 degrees, and anomalies are shaded as in the color bar below the graphics.

Now, we revisit observed CPC maximum temperature (left) and anomalies (right) for 3 August 2008. Recall that 100°F temperatures were common from KS to TX, with positive anomalies of 6-12°F.

Consider the forecasts versus observed 2-m temperatures and anomalies then continue to the question on the next slide.

The model temperature forecasts are too cool, but so is model climatology. Because the model analysis and not the station observations are used for bias correction, bias correction will not fix this kind of problem.

Model and observed temperatures respond similarly to the flow, however, so the forecast and observed anomalies are similar. As a result, a good deterministic forecast based on the NAEFS could use the most-likely, say median, anomaly forecast even if the temperatures are not well-forecast.

In this case, we examined NAEFS in an August 2008 heat wave over the central to southern Plains.

For 500-hPa heights:

1. At day 7 the heat wave wasn’t predicted in the NAEFS
2. A pattern emerged later (central ridge, high heights) favoring well-above normal temperatures by day 3

Day 3 NAEFS 2-m temperature forecasts showed:

1. Most of the area for the 10th, and all of the area for 50th and 90th %-ile was above normal
2. While forecast temperatures were considerably lower than verification, the 50th or median percentile anomalies verified best.

Lessons applying generally:

1. Know relationship between model climatology and station climatology. This will not be bias-corrected!
2. Use the NAEFS anomalies rather than temperatures for station forecasts!

Our next example is from the cold season, and shows the advantage of adding the CMC ensemble to the NCEP ensemble in a winter weather case from February 2007. The bias-corrected files were unavailable for this case, unfortunately.

Here are snow and ice plots for this storm, courtesy of the State College, PA WFO. On the left, we see 4-7” of snow fell in the DC area, part of an area of 4” or greater snows within the red contour. On the right, we see up to ¼” of glaze accumulated in parts of PA, WV, and MD.

Here are 6 day forecast spaghetti plots of 500-hPa 552-dm height contours from 00z 19 February 2007, valid 00z 24 February 2007. GEFS contours are red and CEFS contours blue. Combined, these represent NAEFS.

Here is the 996 hPa sea level pressure graphic at left and the 500-hPa 552-dm height graphic at right from the same run valid 00z 25 February 2007. At right, east- and west-most extent of CEFS ridgelines are marked with the blue zigzag double arrow, with highlighting of the eastmost and westmost members’ contours. The red GEFS ensemble forecasts are inside the CEFS envelope, as can be seen with the red zigzag double arrows and contour highlighting. The upstream CEFS troughs are located within the blue double arrow; the members at extreme locations highlighted. The GEFS trough axes are again mainly inside the CEFS forecast envelope. Verification of the 552-hPa contour (in dark green) for the ridge is near the middle of all NAEFS forecasts, though ridge curvature is sharper than most members; sharply curved ridges are difficult for NWP models to accurately predict. The upstream trough verification is within the NAEFS forecast envelope.

For the sea level diagram at left, GEFS contours are inside the CEFS, and CEFS forecast is more spread out. The verification for the 996-hPa contour is toward the middle of the NAEFS blue and red contour envelope, toward the west part of the GEFS contour envelope, and in the middle of the CEFS forecast envelope.

Note that the CEFS 500-hPa graphic shows *generally* tighter clustering of troughs near the verifying position than the GEFS, even though there are a few outliers beyond the westernmost GEFS members.

To summarize, we used spaghetti plots of sea level pressure and 500-hPa heights for the NAEFS and its CEFS and GEFS components to examine longer-range, synoptic scale forecasts for a cold season winter weather events over the Midwest and Northeast. In the medium range, we saw the CEFS improve the solution in the NAEFS. This was because some members of the CEFS showed a farther west position for their upper air ridge axes, better-suggesting possible cold-air damming east of the Appalachians than the GEFS.

Even though the CEFS improved the ensemble forecast in this case, there are also times where the GEFS may improve the NAEFS over the CEFS.

Now let’s summarize the full webcast, ending with this section.

We defined an ensemble as a set of forecasts valid at the same time from different forecast models, different initial conditions, or both.

We stated the NAEFS is constructed from the U.S. Global Ensemble Forecast System (GEFS) and the Canadian Meteorological Center’s Ensemble Forecast System (CEFS). We showed the NAEFS performs better as an EFS because it has double the membership of either of the EFSs that comprise it, and it uses multiple models. This gives better ensemble mean forecasts, better relationships between ensemble spread and ensemble mean skill, and better probabilistic forecasts.

Now let’s summarize the full webcast, ending with this section.

We defined an ensemble as a set of forecasts valid at the same time from different forecast models, different initial conditions, or both.

We stated the NAEFS is constructed from the U.S. Global Ensemble Forecast System (GEFS) and the Canadian Meteorological Center’s Ensemble Forecast System (CEFS). We showed the NAEFS performs better as an EFS because it has double the membership of either of the EFSs that comprise it, and it uses multiple models. This gives better ensemble mean forecasts, better relationships between ensemble spread and ensemble mean skill, and better probabilistic forecasts.

Also in section 2, we discussed use and interpretation of these products. The 10th and 90th percentiles describe the distribution extremes; standard deviation or spread the amount of uncertainty; and mean, median, and mode the middle or most frequent forecast of the distribution. Anomalies for the 10th and 90th percentiles show the relationship of the full forecast distribution to climatology. We also showed where graphical products and raw and bias-corrected data files can be obtained.

Finally, in this third section, we showed two case examples. First, we examined forecasts of large scale flow and air temperature for a heat wave. Forecast upper-air patterns were conducive to a heat wave within six days of the event, but the model air temperatures were too low.

However, forecast anomalies from model climatology compared favorably to observed anomalies from station climatology. Thus, forecasters can apply NAEFS anomalies to station climatology in their forecasts.

From the cold season case, adding the CEFS to the GEFS improved the large-scale, medium-range forecast of significant winter weather event in the northeastern U.S. Snowfall looked more probable when the CEFS's members were added to the NAEFS, and a significant snowfall verified.