Advertisement
Published in:
2021 | OriginalPaper | Chapter
Data-Driven Methods for Weather Forecast
Authors : Elias David Nino-Ruiz, Felipe J. Acevedo García
Published in: Computational Science – ICCS 2021
Publisher: Springer International Publishing
Activate our intelligent search to find suitable subject content or patents.
Select sections of text to find matching patents with Artificial Intelligence. powered by
Select sections of text to find additional relevant content using AI-assisted search. powered by
Abstract
In this paper, we propose efficient and practical data-driven methods for weather forecasts. We exploit the information brought by historical weather datasets to build machine-learning-based models. These models are employed to produce numerical forecasts, which can be improved by injecting additional data via data assimilation. Our approaches’ general idea is as follows: given a set of time snapshots of some dynamical system, we group the data by time across multiple days. These groups are employed to build first-order Markovian models that reproduce dynamics from time to time. Our numerical models’ precision can be improved via sequential data assimilation. Experimental tests are performed by using the National-Centers-for-Environmental-Prediction Department-of-Energy Reanalysis II dataset. The results reveal that numerical forecasts can be obtained within reasonable error magnitudes in the \(L_2\) norm sense, and even more, observations can improve forecasts by order of magnitudes, in some cases.