Modeling total solar irradiance from PMOD composite using feed-forward neural networks

Tebabal, A.; Damtie, B.; Nigussie, Melessew; Bires, A.; Yizengaw, E.

doi:10.1016/j.jastp.2015.10.007

Cited by 11 publications

(5 citation statements)

References 39 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…It is known that neural networks interpolate well within the input space, and therefore the network is expected to reproduce the data set that was used to train it with relatively good accuracy (Habarulema et al, 2007 ;Habarulema & McKinnell 2012 ;Tebabal et al, 2015). However, this model may not generalize well to new data that is outside the training set (Srivastava et al, 2014;Demuth & Beale, 2000)).…”

Section: Resultsmentioning

confidence: 99%

Temporal solar irradiance variability analysis using neural networks

2015

View full text Add to dashboard Cite

A feed-forward neural network which can account for nonlinear relationship was used to model total solar irradiance (TSI). A single layer feed-forward neural network with Levenberg-marquardt back-propagation algorithm have been implemented for modeling daily total solar irradiance from daily photometric sunspot index, and core-to-wing ratio of Mg II index data. In order to obtain the optimum neural network for TSI modeling, the root mean square error (RMSE) and mean absolute error (MAE) have been taken into account. The modeled and measured TSI have the correlation coefficient of about R=0.97. The neural networks (NNs) model output indicates that reconstructed TSI from solar proxies (photometric sunspot index and Mg II) can explain 94% of the variance of TSI. This modeled TSI using NNs further strengthens the view that surface magnetism indeed plays a dominant role in modulating solar irradiance.

show abstract

Section: Resultsmentioning

confidence: 99%

Temporal solar irradiance variability analysis using neural networks

2015

View full text Add to dashboard Cite

show abstract

“…Purely empirical models use indices of solar activity which are regressed to direct TSI measurements in order to reconstruct TSI variations (e.g. Hudson et al 1982;Schatten et al 1985;Foukal & Lean 1988;Chapman et al 1996Chapman et al , 2013Preminger et al 2002;Wang et al 2005;Wang & Lean 2021;Steinhilber et al 2012;Delaygue & Bard 2011;Zhao & Han 2012;Coddington et al 2016;Tebabal et al 2015;Privalsky 2018;Mauceri et al 2019;Chatzistergos et al 2020a). Some empirical models also include machine learning extrapolations (e.g.…”

Section: Irradiance Models and Reconstructionsmentioning

confidence: 99%

Long-term changes in solar activity and irradiance

Chatzistergos¹,

Krivova²,

Yeo³

2023

Preprint

View full text Add to dashboard Cite

The Sun is the main energy source to Earth, and understanding its variability is of direct relevance to climate studies. Measurements of total solar irradiance (TSI) exist since 1978, but this is too short compared to climate-relevant time scales. Coming from a number of different instruments, these measurements require a cross-calibration, which is not straightforward, and thus several composite records have been created. All of them suggest a marginally decreasing trend since 1996. Most composites also feature a weak decrease over the entire period of observations, which is also seen in observations of the solar surface magnetic field and is further supported by Ca ii K data. Some inconsistencies, however, remain and overall the magnitude and even the presence of the long-term trend remain uncertain. Different models have been developed, which are used to understand the irradiance variability over the satellite period and to extend the records of solar irradiance back in time. Differing in their methodologies, all models require proxies of solar magnetic activity as input. The most widely used proxies are sunspot records and cosmogenic isotope data on centennial and millennial time scale, respectively. None of this, however, offers a sufficiently good, independent description of the long-term evolution of faculae and network responsible for solar brightening. This leads to significant uncertainties in the amplitude of the long-term changes in solar irradiance. Here we review recent efforts to improve irradiance reconstructions on time scales longer than the solar cycle and to reduce the existing uncertainty in the magnitude of the long-term variability. In particular, we highlight the potential of using 3D magnetohydrodynamical simulations of the solar atmosphere as input to more physical irradiance models and of historical full-disc Ca ii K observations encrypting direct facular information back to 1892.

show abstract

“…The extension of the component library in this direction would be computationally costly, and therefore it is beyond the scope of this study. Different types of (implicit) nonlinearities can be accounted for by using neural networks approach (see, e.g., Tebabal et al (2015)).…”

Section: Restrictionsmentioning

confidence: 99%

Method of frequency dependent correlations: investigating the variability of total solar irradiance

Pelt

Käpylä

Olspert

2017

A&A

View full text Add to dashboard Cite

Context. This paper contributes to the field of modeling and hindcasting of the total solar irradiance (TSI) based on different proxy data that extend further back in time than the TSI that is measured from satellites. Aims. We introduce a simple method to analyze persistent frequency-dependent correlations (FDCs) between the time series and use these correlations to hindcast missing historical TSI values. We try to avoid arbitrary choices of the free parameters of the model by computing them using an optimization procedure. The method can be regarded as a general tool for pairs of data sets, where correlating and anticorrelating components can be separated into non-overlapping regions in frequency domain. Methods. Our method is based on low-pass and band-pass filtering with a Gaussian transfer function combined with de-trending and computation of envelope curves. Results. We find a major controversy between the historical proxies and satellite-measured targets: a large variance is detected between the low-frequency parts of targets, while the low-frequency proxy behavior of different measurement series is consistent with high precision. We also show that even though the rotational signal is not strongly manifested in the targets and proxies, it becomes clearly visible in FDC spectrum. A significant part of the variability can be explained by a very simple model consisting of two components: the original proxy describing blanketing by sunspots, and the low-pass-filtered curve describing the overall activity level. The models with the full library of the different building blocks can be applied to hindcasting with a high level of confidence, R c ≈ 0.90. The usefulness of these models is limited by the major target controversy. Conclusions. The application of the new method to solar data allows us to obtain important insights into the different TSI modeling procedures and their capabilities for hindcasting based on the directly observed time intervals.

show abstract

Modeling total solar irradiance from PMOD composite using feed-forward neural networks

Cited by 11 publications

References 39 publications

Temporal solar irradiance variability analysis using neural networks

Temporal solar irradiance variability analysis using neural networks

Long-term changes in solar activity and irradiance

Method of frequency dependent correlations: investigating the variability of total solar irradiance

Contact Info

Product

Resources

About