EGU2020-1130, updated on 10 Jan 2024
EGU General Assembly 2020
© Author(s) 2024. This work is distributed under
the Creative Commons Attribution 4.0 License.

A reproducible solar irradiance estimation process using Recurrent Neural Network

Amita Muralikrishna1,2, Rafael Santos1, and Luis Eduardo Vieira1
Amita Muralikrishna et al.
  • 1National Institute of Space Research, Applied Computing Course / LAC / CTE I, Sao Jose dos Campos, Brazil (
  • 2Federal Institute of Education, Science and Technology of Sao Paulo, Sao Jose dos Campos, Sao Paulo, Brazil (

The Sun have a constant action on Earth, interfering in different ways on life in our planet. The physical, chemical and biological processes that occur on Earth are directly influenced by the variation of solar irradiance, which is a function of the activity in the Sun’s different atmospheric layers and their rapid variation. Studying this relationship may require the availability of a large amount of collected data, without significant gaps that could be caused from many kinds of issues. In this work, we present a Recurrent Neural Network as an option for estimating the Total Solar Irradiance (TSI) and the Spectral Solar Irradiance (SSI) variability. Solar images collected on different wave components were preprocessed and used as the input parameters, and TSI and SSI data collected by instruments onboard of SORCE were used as reference of the results we expected to achieve. Complementary to this approach, we opted for developing a reproducible procedure, for which we chose a free programming language, in attempt to offer the same kind of results, with same accuracy, for future studies which would like to reproduce our procedure. To achieve this, reproducible notebooks will be generated with the intention of providing transparency in the data analysis process and allowing the process and the results to be validated, modified and optimized by those who would like to do it. This approach aims to obtain a good accuracy in estimating the TSI and SSI, allowing its reconstruction in gap scales and also the forecast of their values six hours ahead.

How to cite: Muralikrishna, A., Santos, R., and Vieira, L. E.: A reproducible solar irradiance estimation process using Recurrent Neural Network, EGU General Assembly 2020, Online, 4–8 May 2020, EGU2020-1130,, 2020.


Display file