Refining WRF Integration for Local Urban Weather Modeling: Questions on Applicability and Computational Efficiency

J

JasonWok

New member
I'm currently exploring the integration of WRF with a physics-based building energy model, focusing on spatial resolutions at the room scale (3 meters around).

In WRF, I've configured the finest domain with a 100-meter resolution, comprising a grid of 50 * 50 points, and the time step is set to 0.6 seconds. Additionally, I've allocated 4 computing nodes for this task.

However, I have some questions:

1. Is it appropriate to employ WRF, typically used for mesoscale weather modeling, for calculating such fine-grained outdoor urban weather conditions?

2. If WRF is indeed suitable, considering each node handles only a portion of the grid (25 * 25 points), is there a method to accelerate the computation speed, given the current wall clock speed for a time step of 0.6 seconds is 0.2 seconds, which means it takes 4 month to calculate one year simulation?

Thank you.

Related links:
1. Domain decomposition of latitude and longitude array
2. Two-way online coupling WRF.exe with my.exe
3.
www.researchgate.net

(PDF) Integrating Urban Microclimate Prediction into Large Scale Building Energy Modeling

PDF | Code is available at: https://github.com/xixihaha1995/fortran_experiments The goal of this study is to enhance the accuracy of the large-scale... | Find, read and cite all the research you need on ResearchGate
www.researchgate.net www.researchgate.net
 
Please see my answers below:
(1) Is it appropriate to employ WRF, typically used for mesoscale weather modeling, for calculating such fine-grained outdoor urban weather conditions?
100-m resolution enables you to run WRF in LES mode. While it is theoretically fine to run at this resolution for urban weather simulation, I am concerned that you have to run over multi-nested domains. In addition, the missing of eddy structure in the parent domain of the 100-m domain cam damage the simulation to some extent.
(2) Adaptive time step could speed up the simulation. But I am not sure whether it works fine in your case, since the time step is always very small.
 
You must log in or register to reply here.
Back
Top