Numerical Weather Prediction at 200 m Local Resolution Based on an Unstructured Grid Global Model

• Published in: Earth and space science (Hoboken, N.J.) N.J.), 2022-10, Vol.9 (10), p.n/a
• Main Authors: Sze, Wai‐Pang, Tang, Sze‐Chai, Cheung, Chi‐Chiu, Tam, Chi‐Yung
• Format: Article
• Language: English
• Subjects: Boundary layers; Mountains; Nesting; Prediction models; Rain; Regions; Simulation; Water boundary; Weather forecasting; Wind speed
• ISSN: 2333-5084; 2333-5084
• DOI: 10.1029/2022EA002342

It is a great challenge to reach sub‐km resolution in numerical weather prediction models without using domain nesting techniques. The Model for Prediction Across Scales – Atmosphere (MPAS‐A) uses an unstructured grid to achieve smooth transition of resolution in a global model. However, a number of model physical parameterization schemes and methods to enhance computational efficiency assume resolution down to about 3 km only. We extended the MPAS‐A model to support resolution down to 200 m and carried out idealized and hindcast numerical experiments over a domain covering Hong Kong. The Shin‐Hong Planetary Boundary Layer scheme was integrated to parameterize partially resolved turbulence in a scale‐aware sense. The quality of static geographical data and land‐water boundary for a coastal urban locale are found to be vital to the forecast accuracy of ground temperature and humidity forecast. Resolving terrain to 200 m, the model is found to be able to simulate the sheltering effect of mountains as well as high wind speed at hill top comparable to station observation. Our results indicate that global model simulations with an unstructured grid can realistically reproduce local weather conditions, suitable for ultra‐high‐resolution predictions and studies. Plain Language Summary In daily weather prediction, physically based simulations that divide the Earth's surface into horizontal grid cells and the atmosphere above into vertical layers are run by supercomputers. It is difficult to use very small cells over the whole globe due to the tremendous computational resource required; the Model for prediction across scales – Atmosphere (MPAS‐A) model uses polygonal (dominantly hexagonal) cells varying in cell sizes to cover Earth's surface, achieving variable resolution with gradual resolution transition, unlike other gridding techniques that have an abrupt change of resolution that may create problems. However, only local 3 km mesh is officially provided by the MPAS‐A website. We extended the MPAS‐A model to support the smallest cell size down to 200 m over the city of Hong Kong to be used practically, solving problems in variable‐resolution mesh generation, handling partially resolved turbulence, and preparation of static data. We have run idealized and real case experiments and analyzed results to show that the accuracy of ground temperature, humidity, and wind prediction over the city is improved. It is also noteworthy that resolving terrain with ultra‐high‐r