Is it possible for WRF to generate a forecast using input data for a single time-slice only?

[beuraieon_a]

Hello, everyone. I want to run WRF in order to make a weather forecast for 3 or 5 days out into the future starting from a specific initialization time, just like how PAGASA runs the WRF to make these numerical weather prediction outputs (link: https://bagong.pagasa.dost.gov.ph/products-and-services/numerical-modelling-system). Is it possible for WRF to generate a model simulation of future atmospheric conditions for a specific simulation duration, given an input data for a single time-slice only?

Let's say I have a GFS 0.5-degree analysis data for a single time-slice only, at 00:00 UTC of 27 October 2023. Will it be possible for the WRF Model to accept input data from this single time-slice and use this in generating a model simulation initialized at or starting from 00:00 UTC 27 October 2023 and ending at, say, 00:00 UTC of 1 November 2023?

 

[beuraieon_a]

Hello, everyone. I want to run WRF in order to make a weather forecast for 3 or 5 days out into the future starting from a specific initialization time, just like how PAGASA runs the WRF to make these numerical weather prediction outputs (link: https://bagong.pagasa.dost.gov.ph/products-and-services/numerical-modelling-system). Is it possible for WRF to generate a model simulation of future atmospheric conditions for a specific simulation duration, given an input data for a single time-slice only?

Let's say I have a GFS 0.5-degree analysis data for a single time-slice only, at 00:00 UTC of 27 October 2023. Will it be possible for the WRF Model to accept input data from this single time-slice and use this in generating a model simulation initialized at or starting from 00:00 UTC 27 October 2023 and ending at, say, 00:00 UTC of 1 November 2023?

I hope that I will get any response from Ming Chen, kwerner and other moderators, admin and members who have knowledge or expertise on this. Thank you very much in advance.

 

[kwerner]

Hi,
It can sometimes take a few days for us to have time to respond to each inquiry, and we do not work on weekends. Thank you for your patience.

WRF is capable of running a forecast into the future; however, you must have input/boundary data for the entire time period, so you would need to obtain the forecast files. For e.g., if using GFS 0.25 degree input data, the files are in the format:

gfs.0p25.2019090412.f000.grib2

gfs.0p25.2019090412.f003.grib2

These are files initiated at 2019-09-04, 12 UTC, for the 000 time (i.e., 2019-09-04_00:00:00), and initiated at the same time, but representing the 3 hour forecast time (i.e., 2019-09-04_03:00:00).

 

[beuraieon_a]

Hello, thank you for your response.

WRF is capable of running a forecast into the future; however, you must have input/boundary data for the entire time period, so you would need to obtain the forecast files. For e.g., if using GFS 0.25 degree input data, the files are in the format:


gfs.0p25.2019090412.f000.grib2


gfs.0p25.2019090412.f003.grib2


These are files initiated at 2019-09-04, 12 UTC, for the 000 time (i.e., 2019-09-04_00:00:00), and initiated at the same time, but representing the 3 hour forecast time (i.e., 2019-09-04_03:00:00).

Just to check my understanding since I'm still confused. Does this imply that if I want WRF to run a forecast into the future, say starting from 00:00 UTC of 1 November 2023 (today) up to 00:00 UTC of 5 November 2023 (four days into the future), my input GFS 0.5-degree analysis data needs to have data up to 00:00 UTC of 5 November 2023, even though that date and time is still in the future?

 

[kwerner]

Yes, that is correct. The large-scale models (like GFS) run forecasts several hundred hours into the future, so they provide forecast files. I'm not certain whether the 0.5 degree files include this feature. I typically use GFS 0.25 degree input data files.

 

[beuraieon_a]

Hello, and thank you very much for this clarification. I actually have been using only the GFS 0.5-degree analysis datasets for previous years (no forecast data, only the analyses) as input to the WRF Model, thinking that WRF can generate model outputs using only analyses data as input. I intended to compare WRF model outputs for a past tropical cyclone versus the actual agency best track data using a 4-day set of GFS analyses data as input, and while editing "namelist.input", I immediately realized that I cannot extend the WRF model run beyond the period inclusive of my input data.