Wednesday, August 9, 2017

Exploring the long range convective forecast and TCF

Yesterday was the first full day of the Summer Experiment and things were in full swing at the TCF desk. One of the main goals of the experiment is to explore the idea of a long range forecast, i.e. adding outlooks beyond the main 4, 6, and 8 hour periods. For the purposes of our experiment, these prototype forecasts are issued by participants every hour along with the typical TCF.

A challenge noted almost immediately by participants was time. At the TCF desk, there is an hour in between the issuance of each 4, 6, and 8 hour forecast, during which time a scheduled collaboration chat occurs between AWC, CWSUs, and others, to determine the areas of concern for the short-term. On busy days, it can make that hour feel far too short. Adding on to that a long-range forecast each hour was found to be extremely difficult. In order to just get both the TCF and extended completed in time, participants noted that they didn't have much time to even look at new data sets. In order to add value, particularly to the long-range, it was suggested that perhaps another forecaster would be needed to focus only on that piece.

In later discussions, AWDE brought up the Extended Convective Forecast Plot (ECFP) currently available on the AWC webpage (, an automated plot using various long range models to provide convective forecasts out to 24 hours and beyond. Any long-range models contain inherent uncertainties the further out the forecast hour, and the question was posed whether having a person creating the long range would really add benefit all the time.
Example 36-hour forecast from the Extended Convective Forecast Plot (ECFP) on the AWC's website
Feedback from a traffic planner perspective was that TFMs don't create plans that far out in advance. Instead, they use it to gauge what they may have to deal with the following day. If the automated forecast could be improved, that would indeed be useful for these outlooks. However, realistically, TFMs only take a look at the beginning and end of each day, and would not be looking at a 24 or 36 hour forecast every hour. In light of this, it was suggested that a long range convective forecast be provided in the morning with the first TCF package, giving planners a look at the following morning, and then again in the evening, giving planners at look at the following evening. Additionally, adding a 36 hour forecast at these times was mentioned as well. This would not only reduce forecaster workload at the TCF desk, but also add value where it could really be used.

On the dataset side, forecasters were utilizing high resolution models like the NAM Nest, ARW, and others, along with the INSITE tool to generate their forecasts.
NAM Nest Composite Reflectivity forecast in N-AWIPS
 One comment was that, for the long range forecast, it was difficult to find models that actually went out far enough to be of use. Some upstream dataflow struggles caused AWC to lose some of the experimental long range datasets initially available. More evaluation of these will hopefully be completed throughout the rest of the experiment.

INSITE was able to be utilized to some extent for the 4, 6, and 8 hour TCF. It was noted that it is a very good dataset for a broad scale view of possible traffic flow impacts. However, for CWSUs or planners responsible for looking at particular routes in a small region, it may not be of as much help beyond determining that there is, in fact, going to be an impact.

For the long range forecast INSITE has potential. Having that broad scale view of potential impact may be useful in the 24 to 36 hour range. At the moment, though, INSITE is only generated out to 12 hours for each model. It was suggested that for the long range, it would need to be extended out to at least 24 hours to be of use. 

No comments:

Post a Comment