Context. The variable Sun is the most likely candidate for the natural forcing of past climate changes on time scales of 50 to 1000 years. Evidence for this understanding is that the terrestrial climate correlates positively with the solar activity. During the past 10 000 years, the Sun has experienced the substantial variations in activity and there have been numerous attempts to reconstruct solar irradiance. While there is general agreement on how solar forcing varied during the last several hundred years -all reconstructions are proportional to the solar activity -there is scientific controversy on the magnitude of solar forcing. Aims. We present a reconstruction of the total and spectral solar irradiance covering 130 nm-10 μm from 1610 to the present with an annual resolution and for the Holocene with a 22-year resolution. Methods. We assume that the minimum state of the quiet Sun in time corresponds to the observed quietest area on the present Sun. Then we use available long-term proxies of the solar activity, which are 10 Be isotope concentrations in ice cores and 22-year smoothed neutron monitor data, to interpolate between the present quiet Sun and the minimum state of the quiet Sun. This determines the longterm trend in the solar variability, which is then superposed with the 11-year activity cycle calculated from the sunspot number. The time-dependent solar spectral irradiance from about 7000 BC to the present is then derived using a state-of-the-art radiation code. Results. We derive a total and spectral solar irradiance that was substantially lower during the Maunder minimum than the one observed today. The difference is remarkably larger than other estimations published in the recent literature. The magnitude of the solar UV variability, which indirectly affects the climate, is also found to exceed previous estimates. We discuss in detail the assumptions that lead us to this conclusion.

Context. The solar irradiance is known to change on time scales of minutes to decades, and it is suspected that its substantial fluctuations are partially responsible for climate variations. Aims. We are developing a solar atmosphere code that allows the physical modeling of the entire solar spectrum composed of quiet Sun and active regions. This code is a tool for modeling the variability of the solar irradiance and understanding its influence on Earth. Methods. We exploit further development of the radiative transfer code COSI that now incorporates the calculation of molecular lines. We validated COSI under the conditions of local thermodynamic equilibrium (LTE) against the synthetic spectra calculated with the ATLAS code. The synthetic solar spectra were also calculated in non-local thermodynamic equilibrium (NLTE) and compared to the available measured spectra. In doing so we have defined the main problems of the modeling, e.g., the lack of opacity in the UV part of the spectrum and the inconsistency in the calculations of the visible continuum level, and we describe a solution to these problems. Results. The improved version of COSI allows us to reach good agreement between the calculated and observed solar spectra as measured by SOLSTICE and SIM onboard the SORCE satellite and ATLAS 3 mission operated from the Space Shuttle. We find that NLTE effects are very important for the modeling of the solar spectrum even in the visual part of the spectrum and for its variability over the entire solar spectrum. In addition to the strong effect on the UV part of the spectrum, NLTE effects influence the concentration of the negative ion of hydrogen, which results in a significant change of the visible continuum level and the irradiance variability.