[Paper Review] Short-term forecast of the total and spectral solar irradiance

L. E. A. Vieira, Thierry Dudok de Wit+1 more|arXiv (Cornell University)|Nov 21, 2011

Solar and Space Plasma Dynamics20 references3 citations

TL;DR

This paper presents a Layer-Recurrent Neural Network (LRN) model that forecasts total and spectral solar irradiance up to three days in advance using near-real-time solar magnetograms and intensity images. The model achieves sub-5% error for 24-hour forecasts in the 115–180 nm range, with performance degrading to ~20% error in the 180–310 nm band due to instrument degradation effects.

ABSTRACT

Among several heliophysical and geophysical quantities, the accurate evolution of the solar irradiance is fundamental to forecast the evolution of the neutral and ionized components of the Earth's atmosphere.We developed an artificial neural network model to compute the evolution of the solar irradiance in near-real time. The model is based on the assumption that that great part of the solar irradiance variability is due to the evolution of the structure of the solar magnetic field. We employ a Layer-Recurrent Network (LRN) to model the complex relationships between the evolution of the bipolar magnetic structures (input) and the solar irradiance (output). The evolution of the bipolar magnetic structures is obtained from near-real time solar disk magnetograms and intensity images. The magnetic structures are identify and classified according to the area of the solar disk covered. We constrained the model by comparing the output of the model and observations of the solar irradiance made by instruments onboard of SORCE spacecraft. Here we focus on two regions of the spectra that are covered by SORCE instruments. The generalization of the network is tested by dividing the data sets on two groups: the training set; and, the validation set. We have found that the model error is wavelength dependent. While the model error for 24-hour forecast in the band from 115 to 180 nm is lower than 5%, the model error can reach 20% in the band from 180 to 310 nm. The performance of the network reduces progressively with the increase of the forecast period, which limits significantly the maximum forecast period that we can achieve with the discussed architecture. The model proposed allows us to predict the total and spectral solar irradiance up to three days in advance.

Motivation & Objective

• To develop a near-real-time model for forecasting total and spectral solar irradiance based on solar surface magnetic activity.
• To address the challenge of limited continuous monitoring of solar spectral irradiance, especially as SORCE mission nears end-of-life.
• To improve space weather and atmospheric modeling by providing accurate, timely forecasts of solar irradiance variability.
• To reduce forecast error by classifying magnetic structures by their disk coverage and modeling their evolution over time.
• To enable operational forecasting with low latency and continuity, overcoming limitations of existing empirical models.

Proposed method

• Employing a Layer-Recurrent Network (LRN) to model the nonlinear relationship between evolving bipolar magnetic structures and solar irradiance output.
• Using near-real-time solar disk magnetograms and intensity images to extract and classify magnetic structures based on their fractional disk coverage.
• Classifying structures into four categories: small ephemeral regions (filling factor <16.7 ppm), intermediate (16.7–24.8 ppm), large active regions, and sunspots (umbrae and penumbrae).
• Constraining the neural network coefficients using observations from SORCE's SOLSTICE and XPS instruments for spectral irradiance in the 115–310 nm and 0.1–34 nm ranges.
• Dividing data into training and validation sets to test generalization and assess forecast accuracy across different time horizons.
• Applying feedback mechanisms in the LRN architecture to capture temporal dependencies in magnetic structure evolution.

Experimental results

Research questions

• RQ1 Can a neural network model accurately forecast total and spectral solar irradiance using only near-real-time solar surface magnetic field observations?
• RQ2 How does forecast accuracy vary across different spectral bands, particularly in the UV (115–310 nm) and extreme-UV (0.1–34 nm) ranges?
• RQ3 What is the impact of instrument degradation on forecast performance, especially in the 180–310 nm range?
• RQ4 How does forecast error scale with increasing prediction horizon (e.g., 12h, 24h, 48h, 72h)?
• RQ5 Can the model’s performance be improved by integrating a solar surface magnetic flux transport model to extend forecast capability?

Key findings

• The model successfully forecasts total and spectral solar irradiance up to three days in advance using real-time solar magnetic data.
• For 24-hour forecasts in the 115–180 nm band, the model error is below 5%, indicating high accuracy in the mid-UV range.
• In the 180–310 nm band, the forecast error increases to approximately 20%, likely due to degradation and reduced accuracy of MUV measurements.
• Forecast accuracy degrades progressively with longer prediction horizons, limiting the practical forecast window to about three days with the current architecture.
• The model's performance is significantly improved by classifying magnetic structures according to their disk coverage, enabling better representation of solar irradiance variability.
• The real-time forecast service is publicly available at http://www.lpc2e.cnrs-orleans.fr/~soteria, supporting operational space weather applications.