Solar Cycle Signal in Climate and Artificial Neural Networks Forecasting

Abstract

Natural climate variability is partially attributed to solar radiative forcing. The purpose of this study is to contribute to a better understanding of the influence of solar variability on the Earth’s climate system. The object of this work is the estimation of the variation of multiple climatic parameters (temperature, zonal wind, relative and specific humidity, sensible and latent surface heat flux, cloud cover and precipitable water) in response to solar cycle forcing. An additional goal is to estimate the response of the climate system’s parameters to short-term solar variability in multiple forecasting horizons and to evaluate the behavior of the climate system in shorter time scales. The solar cycle is represented by the 10.7 cm solar flux, a measurement collected by terrestrial radio telescopes, and is provided by NOAA/NCEI/STP, whereas the climatic data are provided by the NCEP/NCAR reanalysis 1 project. The adopted methodology includes the development of a linear regression statistical model in order to calculate the climatic parameters’ feedback to the 11-year solar cycle on a monthly scale. Artificial Neural Networks (ANNs) have been employed to forecast the solar indicator time series for up to 6 months in advance. The climate system’s response is further forecasted using the ANN’s estimated values and the regression equations. The results show that the variation of the climatic parameters can be partially attributed to solar variability. The solar-induced variation of each of the selected parameters, averaged globally, was of an order of magnitude of 10−1–10−3, and the corresponding correlation coefficients (Pearson’s r) were relatively low (−0.5–0.5). Statistically significant areas with relatively high solar cycle signals were found at multiple pressure levels and geographical areas, which can be attributed to various mechanisms.