gdlewen
EF4
As part of a chase post-mortem, I was looking at the HRRR Simulated Reflectivity (using CoD graphic output). I understand that current observations, including radar data, are used to initialize models, and it is my further understanding that the Zero-Hour Forecast (F00) reflects the analysis of the current conditions.
What I don't understand is how the HRRR took the 23Z initialization, and turned it into this at 00Z (on 4/2):
The "primary cell" and its neighbor is more or less well-predicted, but the convection in SW Oklahoma has no representation in the F01 forecast. Plus there is an explosion of cells W and S of the primary cell, which disappear in the next forecast:
The F02 forecast still hasn't developed convection in SW OK, which is well ongoing by 01Z. Plus there is a new strong cell W of Tonkawa that really has nothing to do with the actual primary cell, which is kinda-maybe correctly modeled across the OK border in KS.
I'm not saying we should take the location of modeled cells literally (and I don't), but this does make me wonder what in the HRRR causes this behavior. It's a pretty big hour-to-hour divergence that it seems would be even more noticeable if the model wasn't reinitialized every hour. Just posing this as a general question....
 |  |
Composite Reflectivity for 4/1/2025 23Z based on KFDR, KINX, KTLX, KVNX data usage Py-Art. Little effort to exclude clutter beyond masking below 5dBz. (Probably should have added in KDDC and KICT.) | HRRR 23Z Run F00 Analysis Reflectivity |
What I don't understand is how the HRRR took the 23Z initialization, and turned it into this at 00Z (on 4/2):
 |  |
| Composite Reflectivity for 4/2/2025 00Z based on KFDR, KINX, KTLX, KVNX data. | HRRR 23Z Run F01 Predicted Reflectivity |
 |  |
| Composite Reflectivity for 4/2/2025 01Z based on KFDR, KINX, KTLX, KVNX data. | HRRR 23Z Run F02 Predicted Reflectivity |
The F02 forecast still hasn't developed convection in SW OK, which is well ongoing by 01Z. Plus there is a new strong cell W of Tonkawa that really has nothing to do with the actual primary cell, which is kinda-maybe correctly modeled across the OK border in KS.
I'm not saying we should take the location of modeled cells literally (and I don't), but this does make me wonder what in the HRRR causes this behavior. It's a pretty big hour-to-hour divergence that it seems would be even more noticeable if the model wasn't reinitialized every hour. Just posing this as a general question....
Last edited:
