Preliminary work testing different ICs in the RENCI/NCSU HUR-NC Model

Hi All,

I wanted to share some of the work I have been doing, comparing different options for TC initial conditions (ICs) in our HUR-NC WRF model. The objective of my work is to find out what the optimal IC is for TC prediction, and specifically to gauge whether a bogus vortex is a practical and useful solution. I am examining one strong case (Hurricane Ike, 2008) and one weak case (Tropical Storm Erika, 2009), in order to see how the bogus vortex affects the forecast in these different scenarios.

The model has a parent domain with 27-km grid spacing and two storm-following one-way inner nests with 9-km and 3-km grid length. HUR-NC did fairly well as an operational model, especially considering that it was using only GFS data for the model initialization. However, HUR-NC’s forecasts still had room for improvement, as compared with the TPC Best Track data.

I did some experiments where I re-ran HUR-NC for Hurricane Ike and Tropical Storm Erika, using the GFDL bogus vortex IC instead of the GFS IC. (Thanks to Bob Hart for the use of his method of “merging” the GFDL bogus vortex with the GFS boundaries.)

Here are close-ups of Hurricane Ike for the GFS-only and the merged (bogus vortex) model analyses on the parent grid:

Click images to enlarge, if desired.

GFS-Only SST and SLP

Merged SST and SLP

Potential Vorticity cross-section through GFS-only IC

Potential Vorticity cross-section through Merged IC

Notice the very robust structure of the GFDL bogus vortex as compared with the weaker, more amorphous GFS version of Ike.

The following is a graph of minimum SLP for Ike on the parent (27-km grid spacing) domain; it shows GFS-only, merged, and TPC Best Track SLP. Despite the merged run providing a more accurate initial intensity than the GFS-only experiment, both  model runs performed very similarly beyond a 24-hour lead time.

The GFS-only and merged model runs exhibited fairly similar track forecasts, due to the steering flow being so similar in both experiments.

In case you’re wondering whether the previous intensity errors would be improved/reduced on a higher resolution grid, below is what they look like on the 3-km domain. The bogus vortex still does not do well after 24 hours into the model run!

Obviously Ike is just one storm (although Erika results are even less encouraging, given the dissimilarity between Erika’s structure and the bogus vortext structure); however, it is clear that the bogus vortex is not the best solution for all tropical cyclone ICs. The bogus vortex does seem to improve the forecast during the early hours of the model run for storms that initialize with strong intensity, but seems to over-deepen more than the GFS-only storm beyond a 24 hour lead time.

Going forward, I will be testing out initial conditions generated by assimilating observations every six hours from surface stations, rawinsondes (including G-IV dropsondes), ACARS, satellite motion winds, and advisory TC position and MSLP (Courtesy of Ryan Torn, who is sharing his data kindly). These ICs are prepared for an ensemble model, but we will be using the one member that minimizes a cost function that measures the distance from the ensemble member position to the ensemble-mean position and the ensemble member’s minimum SLP to the advisory MSLP.

I am hopeful that this assimilated initial condition will add skill to HUR-NC. Once we get the process for integrating this IC into WRF, we will try to do some parallel runs with HUR-NC to see how the assimilated IC performs relative to the GFS-only and GFDL bogus vortex ICs.

This entry was posted in CSTAR, TC Improved Initial Conditions. Bookmark the permalink.

2 Responses to Preliminary work testing different ICs in the RENCI/NCSU HUR-NC Model

  1. Briana,

    Is there a good reference or introduction into initial conditions and NWP used for tropical cyclones? This topic is likely new to many of use and a good baseline reference would be helpful. Because of operational requirements we oftentimes depend upon guidance from the operational NAM/GFS even though we know that the appropriateness of the guidance is limited in these environments. It would also be helpful for forecasters to understand how the operational NAM/GFS initialize in these situations and how this may impact their handling of a TC. A good example might be the large wind associated with a TC approaching southern NC and the extent and intensity of the easterly and northeasterly winds ahead of it. We have to use guidance available to the WFOs such as the NAM/GFS for various grids and it would be helpful if we understood how the NAM/GFS initiailzed and handled these situations.


  2. Briana Gordon says:

    JB, that is a great question. I will look into this to see if there is anything like that out there and get back to you. If not, maybe we can work together to develop a helpful training guide on this topic.
    More soon,

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s