Thursday, August 10, 2017

Day 3 Recap

Day 3 of the 2017 Summer Experiment wrapped up on a positive note with much insightful discussion among participants. Check out some of the feedback collected below:

SFO desk

UKMET cloud cover - 13 UTC run, 19 UTC forecast
Forecasting ceiling and visibility around SFO has long been a challenge for aviation. Creating C&V model fields to try and predict the progression and cessation of fog and low stratus is even more difficult. However, several of the models being used at the SFO desk yesterday actually did fairly well.

Early model runs from the HRRR suggested a clearing time of 1800 UTC, while the high-res UKMET suggested a little later at 1900 UTC (right). Again, the UKMET showed skill in being a more realistic view of cloud cover (i.e. not smoothed). Compared to the satellite, it was relatively close.

GLAMP cloud ceiling height - 13Z run, 2000 UTC forecast
The GLAMP indicated lifting at 1900 UTC, but not fully clearing until 2000 UTC (below). With the fog and low stratus layer being particularly stubborn yesterday, this later clearing time ended up being the most accurate, and yet still a bit early. It was suggested that the GLAMP may have performed best in this case because the model takes into account winds, and winds yesterday were ~20mph from the west most of the afternoon, keeping fog inland longer into the day.


The C&V desk yesterday was focused east. AWC participants used the National blend to fill the Sky and CloudBasePrimary grids, while also populating the CloudBasePrimary using pop thunder to add a second layer of clouds and fill as broken in SkySecondary. This was reflected in the TAFs produced off the AWC grids

After this process, there was much discussion on the 'speckly' nature of many of the grids utilized.
AWC Sky grid

While speckles create a more realistic picture of clouds, they do not necessary help the TAF. In fact, it may hinder. With many speckles, the coverage and base of clouds will be very up and down in the TAF as the model progresses. In other words, there is no clear trend. Again, this is realistic, particularly with partial cloud cover. However, realistic isn't always the most useful for customers, particularly when the up down isn't particularly significant. Feedback from the AWDE perspective suggested that a 'perfect' grid isn't necessary.
AWC Sky grid - smoothed

 Instead, they would rather have something 'practically perfect', or smoothed, so that there is a clear trend in the TAF.

WFO participants also commented that the 'speckly' grids were not ideal for working at a local level. They are using the grids as a driver to create the TAFS, meaning they really need to the smoother fields to do so. During the experiment periods, they actually took the 'speckly' Sky grid and smoothed it out for their purposes before sending back, resulting in clear evidence of difference between a national and local perspective.

AWC CloudBasePrimary grid - the Georgia grid area was manipulated by the WFO participants, who smoothed the grids out in order to best work with the TAF format. The remainder of the grid was produced on an national scale by AWC.

TCF feedback was similar to that collected over the past few days. The workload was again a concern, as creating the long-term convective forecast along with the usual 4, 6, and 8 hour forecasts leaves little time for efficient data interrogation. It was suggested that maybe the extended focus only on high-impact areas, or areas where significant traffic constraints may occur, to free up forecaster time and provide value for planners.

On the model side, the ARWX painted a more realistic picture than the HREF as far as convective coverage, while participants fell back to the NAM and GFS for broad scale features. The ARWX was also noted to be a lot better at 24 hours and beyond, not showing a lot of confidence in coverage in the long term. No models had medium coverage for the long-term, which is realistic given the uncertainty in the models. Participants noted that the focus for these long-term convective forecasts should be on the high-confidence, low coverage.

No comments:

Post a Comment