Dec. 7, 2004: Revisions to the Weekend 6-10 day and 8-14 day automated blend forecasts

6-10 day and 8-14 day operational forecasts of 5- and 7-day mean 500mb heights and anomalies, surface temperature and precipitation are produced by a forecaster on week days (M-F), and through an automated consolidation of forecast tools on weekends (Sat, Sun).

The previous version of the automated forecast (GEN1) could consolidate only a limited number of forecast tools (about 10). The revised version (GEN2) can consolidate up to 50 tools. GEN2 also incorporates a new precipitation bias correction technique and probability scaling to bring the forecast probabilities more in line with realistic expectations. Also, the graphics have been improved to address a number of problems observed in the previous version, including incomplete contouring (particularly in the southwest U.S. precipitation) and the production of filled maps to improve readability.

In the near future, we hope to upgrade this process to allow objective selection of tool inputs based on previous skill, bias-correction of the D+11 precipitation, and input from the 6-10- & 8-14day probability of exceedence techniques, which we plan to implement this fiscal year.

Currently (Dec 2004) the GEN2 process uses 24 inputs for 6-10day temperature (Kleins, neural networks, CDC & analogs), 18 for 6-10day precipitation (neural networks, CDC, analogs & bias-corrected accumulated precipitation), 14 for 8-14day temperature (Kleins, neural networks, CDC & analogs) and 7 for 8-14day precipitation (neural networks, CDC & analogs). The differences result from the absence of D+11 ECMWF forecasts and of a precipitation bias correction for the D+11. Probability-only tools are converted to CDF (cumulative density function) values. The unweighted mean of the CDFs are bias-corrected in the case of the precipitation, then probabilities are determined, scaled & plotted. The scaled values are retained for verification. The unscaled values are also retained for QC.

Jul 18, 2002:We are soliciting comments on a proposal to release monthly and seasonal outlooks on the Thursday between 15-21 of each month. On average - this pushes the schedule back 3 days. Updated future release dates in 2002 include: Sep. 19 - Nov. 21 - Dec. 19. RELEASE DATES FOR AUGUST AND OCTOBER ARE UNCHANGED. The purpose is to ensure availability of model-based forecast tools which may improve our forecasts.

PLEASE DIRECT PHONE/EMAIL RESPONSES TO:
ROBERT LEFFLER 301-713-1970 x134
ROBERT.LEFFLER@NOAA.GOV
NO LATER THAN AUGUST 20 2002.

Beginning May 17, 2001: Base period means (BPM) for 1971-2000 will replace the 1961-1990 means in CPC's official 6-10-, 8-14-, 30-, and 90-day outlooks beginning May 17, 2001. Changing the means now will allow users to adjust to the impact the change will have on outlooks for winter 2001-2002 well in advance.

These means are intended for use in our operational outlooks only and are based on preliminary data, some of which may contain errors of generally small magnitude. NOAA's National Climatic Data Center (NCDC) will release a more accurate set of climate normals later this year. We recommend NCDC's normals be used for calculations requiring data corrected for station biases and other sources of error.

Oct 2, 2000: CPC will begin issuing operational 6-10 day and 8-14 day temperature and precipitation outlooks, daily, in a probabilistic format.

Operational ensembles of forecasts from the MRF model make this change possible. Computing technology that allows automated production and dissemination -on the one hand- but permits intervention by the forecaster -on the other hand- has made this feasible. With this change we are greatly increasing our service to users of our outlooks by:

- adding a new operational outlook for 8-14 days
- increasing the frequency of the 6-10 day outlook to daily
- increasing the information content of the 6-10 day outlook through the use of a probabilistic format

while maintaining accuracy of the forecasts.

The forecasts will be expressed in 3 categories Above normal, Near normal and Below normal, which are equally likely at any given location over thirty years, for a given calendar forecast period. The total probability will be shown. This means the 33.33% contour forms a boundary separating Near normal, on one side, from either Above normal or Below normal, on the other side. The 33.33% contour will be drawn thicker than the others.

When the 33.33% contour lies between Near normal and Below , it, and all the contours on the Below side of the line, will be colored blue.

When the 33.33% contour separates Near normal and Above, it, and all the contours on the Above side of the line, will be colored red.

Near normal contours will be colored green. To further clarify the map, the letters B, N and A will mark the highest values of the favored category on the map.

Because the climatological likelihood of each category is 33.33%, when the total probability of the A or B category exceeds 66.67%, the likelihood of the opposite category approaches zero, while the likelihood of extreme values increases. Regions where probabilities exceed 66.67% on the new outlook maps often correspond closely to the much above normal category on the old outlook maps.

In the Southwest and other climatologically dry regions, there will be a greater than 33.33% chance of no precipitation and occasionally even a normal (i.e. median) value of zero, especially during the dry seasons. In such cases, a forecast of Near normal is effectively a forecast of little or no precipitation.

The light blue dashed contours on the map are the climatological normal (30 year average) temperature and precipitation totals for the forecast period, expressed in degrees Fahrenheit and tenths of inches, respectively.

A single text bulletin (NMCPMDMRD FXUS06 KWBC) will accompany both the 6-10 day and 8-14 day outlooks. Two tables of U.S. State forecasts, one for the 6-10 day one for the 8-14 day outlook, will appear in the body of the bulletin. On Saturday and Sunday, the tables only, will appear in the bulletin. Likewise, on weekend days the forecasts will be fully automated and objective, while on weekdays forecasters may intervene in the content of the forecast maps. Also on weekdays, the bulletin will contain a prognostic discussion of the 6-10 and 8-14 day outlooks written by the forecaster.

Sample set of 6-10 day maps (in the new format).

June 1, 2000: Excessive Heat Outlooks for 3-7, 6-10 and 8-14 days will be implemented. The target audience for these technical outlooks are emergency managers and professional meteorologists. The outlooks are completely computer-generated (there is no forecaster intervention) and will be issued each day at 3:00 PM Eastern Time. The outlooks predict the likelihood that the daily mean heat index will be at least 85^oF for at least 3 days in 5 (for the 3-7 and 6-10 day outlooks) or 3 days in 7 (for the 8-14 day outlook). The outlooks will also predict the highest maximum heat index in the forecast period.

Beginning Feb 28, 2000, 6-10 day outlooks will include Alaska. Also, the nominal release time of the outlook will be changed from the current 3 PM Eastern, to 4 PM Eastern.