Towards an operational data assimilation System: a detailed web-based framework for weather forecast

Abstract

Weather forecasts are of extreme importance nowadays. They help us daily to plan and schedule activities in different contexts of our society. This paper details an effi- cient web-based framework for operational weather forecasts. We not only present the different layers but also explain how to deploy them using cutting-edge tech- nologies. We build our forecast system on top of an in-house High-Performance- Computing cluster. The Weather Research Forecast model is the numerical model to employ during forecast steps; the imperfect model trajectories are adjusted ev- ery three hours according to real-noisy observations via sequential data assimila- tion. We build numerical grids by interpolating GFS-NOAA datasets onto regular grids. Observations are taken from METeorological Aerodrome Reports extracted from OGIMET. A three-dimensional variational data assimilation method is applied to digest observations. We use interactive Folium maps in Python to display model outputs for all variables. These maps are published in https://aml-cs.github.io/ wrf-baq-0.5km/ and consumed using the GitHub REST API.