Tracking the model: Data assimilation by artificial neural network
| Autor: | Rosangela Cintra, Haroldo de Campos Velho, Steven Cocke |
|---|---|
| Rok vydání: | 2016 |
| Předmět: |
Optimization problem
010504 meteorology & atmospheric sciences Artificial neural network Computer science business.industry Process (computing) Kalman filter Atmospheric model 010502 geochemistry & geophysics Machine learning computer.software_genre Numerical weather prediction 01 natural sciences Data assimilation General Circulation Model Initial value problem Artificial intelligence business computer Algorithm Physics::Atmospheric and Oceanic Physics 0105 earth and related environmental sciences |
| Zdroj: | IJCNN |
| DOI: | 10.1109/ijcnn.2016.7727227 |
| Popis: | To generate reliable forecasts, we need good estimates of both the current system state and the model parameters. Numerical weather prediction (NWP) uses atmospheric general circulation models (AGCMs) to predict weather based on current weather conditions. The process of entering observation data into mathematical model to generate the accurate initial conditions is called data assimilation (DA). It combines observations, forecasting, and filtering step. The data assimilation process is performed by using artificial neural networks (NN) to obtain the initial condition to the atmospheric global model for the Florida State University (in USA. The NN is configured to emulate the analysis computed from the Local Ensemble Transform Kalman filter (LETKF) analysis. The method is tested employing synthetic observations. Multilayer Perceptron neural network is applied, with supervised training algorithm. An optimal configuration for the NN is obtained by solving an associated optimization problem. The data assimilation cycle is carried out at January, 2004. The results demonstrate the effectiveness of the NN technique for atmospheric data assimilation, with better computational performance and similar quality of LETKF analyses. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|