REFINEMENT OF ATMOSPHERIC TRANSPORT MODELS BASED ON SOURCE IDENTIFICATION AND MACHINE LEARNING METHODS
Section: PHYSICS OF THE TROPOSPHERE
Authors:
1.
Institute of Computational Mathematics and Mathematical Geophysics SB RAS
2.
Institute of Computational Mathematics and Mathematical Geophysics SB RAS
V.M. Matrosov Institute of System Dynamics and Control Theory SB RAS
V.M. Matrosov Institute of System Dynamics and Control Theory SB RAS
Type:
Сonference article
Pages:
from 526
to 529
Published:
17.08.2026
Subject area:
UDC 504.3
Language:
Russian
Keywords:
advection-diffusion-reaction model, trained parametric element, monitoring networks
Funding:
The development of a general algorithm for refining a mathematical model based on the sources identification and machine learning was carried out within the framework of the state assignment of the ICM&MG SB RAS FWNM-2025-0003, adaptation and testing for scenarios to the Baikal region were carried out within the framework of grant No. 075-15-2024-533 of the Ministry of Science and Higher Education of the Russian Federation for the implementation of a major scientific project in priority areas of scientific and technological development (the project "Fundamental research of the Baikal natural area based on a system of interrelated basic methods, models, neural networks and a digital platform for environmental monitoring of the environment").
Abstract:
A two-stage algorithm for refining the advection-diffusion-reaction model based on measurement data is investigated. Refinement is performed by adding a specially trained parametric element to the model equation. The first stage consists in identifying, based on measurement data, a general unsteady uncertainty function corresponding to the source function of the basic model. At the second stage, the identification results, including the state function and the values of the uncertainty function, are considered as a training sampling for a parametric element. The trained element is then integrated into the basic model. The developed algorithm for identifying the "true" parametric element based on measurement data from a regular model monitoring network has been tested on scenarios for the Baikal region.
A two-stage algorithm for refining the advection-diffusion-reaction model based on measurement data is investigated. Refinement is performed by adding a specially trained parametric element to the model equation. The first stage consists in identifying, based on measurement data, a general unsteady uncertainty function corresponding to the source function of the basic model. At the second stage, the identification results, including the state function and the values of the uncertainty function, are considered as a training sampling for a parametric element. The trained element is then integrated into the basic model. The developed algorithm for identifying the "true" parametric element based on measurement data from a regular model monitoring network has been tested on scenarios for the Baikal region.
Keywords:
advection-diffusion-reaction model, trained parametric element, monitoring networks
advection-diffusion-reaction model, trained parametric element, monitoring networks
1. Penenko A.V., Gochakov A.V., Antohin P.N. Algoritm usvoeniya dannyh na osnove operatora chuvstvitel'nosti dlya trehmernoy modeli perenosa i transformacii primesey v atmosfere. Optika atmosfery i okeana. 2024; 37(9):719–728. DOIhttps://doi.org/10.15372/AOO20240901.
2. Arcucci R., Zhu J., Hu S., Guo Y.-K. Deep Data Assimilation: Integrating Deep Learning with Data Assimilation. Applied Sciences. 2021; 11(3):1114. DOIhttps://doi.org/10.3390/app11031114.
