Atmospheric Data Assimilation by Self-configured Artificial Neural Networks

• Grant number: 16/14393-2
• Support Opportunities: Regular Research Grants
• Start date: August 01, 2017
• End date: July 31, 2019
• Field of knowledge: Physical Sciences and Mathematics - Geosciences - Meteorology
• Principal Investigator: Haroldo Fraga de Campos Velho
• Grantee: Haroldo Fraga de Campos Velho
• Host Institution: Instituto Nacional de Pesquisas Espaciais (INPE). Ministério da Ciência, Tecnologia e Inovação (Brasil). São José dos Campos , SP, Brazil
• Associated researchers: Haroldo Fraga de Campos Velho

Abstract

Data assimilation is the process by which short-forecast and observations are combined to obtain the representation of the initial state of the modeled system, e.g. is a technique to generate an initial condition to weather or climate forecasts. It is used as strategy to increase the accuracy and the range of forecast. This process is critical in operational Numerical Weather Prediction (NWP). However, with the evolution of the numerical models and numerical exponential increase in the number of observations, especially satellites, NWP become a challenge for the current meteorology. The data assimilation artificial neural networks (NN) approach presents a possible technique to solve this challenge. This project proposes to apply NN, multilayer perceptron, to two the three-dimensional global atmospheric models with analysis emulating a consistent analysis by NCEP (National Centers of Environmental Prediction), and run the models to obtain weather prediction. The models are from the Weather Forecast Center and Climate Studies (CPTEC)/ INPE and Florida State University (FSU) in the USA. This research began in the Laboratory for Computing and Applied Mathematics (LAC) / INPE, wich has shown great gain in computational performance to get the the initial conditions with similar quality from the emulated data assimilation techniques. (AU)