engleski

Testing a link between cosmic rays and cloudiness over daily timescales

Much debate still remains regarding a hypothesized link between the solar-modulated cosmic ray (CR) flux and the Earth’s cloud properties influencing the climate. Recently, it was shown that numerous long-term studies of satellite-based cloud observations are limited by non-trivial disadvantages, such as: satellite data intercalibration issues, view-angle biases, and the influences of factors on cloud cover like ENSO and volcanic eruptions interfering with the analyses. Consequently, the reported studies have failed to present compelling evidence of a CR–cloud link. The satellite-data limitations can be resolved by focusing on short-term (daily) timescales using Forbush decrease events and epoch-superpositional (composite) methods. Unfortunately, these studies have also arrived at a range of conflicting conclusions. It may be the case, that for the short-timescale studies, a hypothesized CR induced signal in clouds may be drowned in the meteorological noise, and noise may even be mixed with the (likely far smaller) hypothesized signal. Using extensive Monte Carlo simulation techniques and two most widely used satellite cloud datasets (ISCCP and MODIS), we quantitatively demonstrate how the high noise levels present in composites of small sample sizes, or for overly isolated sample areas, may predominately account for the inconsistent results obtained. Furthermore, we find that assumptions made by classical statistical tests (like the Student’s T-test) are frequently violated by both the restricted samples and methods frequently employed in the literature (such as normalization to an averaging period). We conclude that such tests should be avoided, in favor of MC simulations, which offer a far more robust method of assessing significance and enabled us to correctly assess the significance of some recent short-term studies purporting to identify evidence of a CR–cloud link.

cloud cover; cosmic rays; Forbush decreases; Monte Carlo; statistics

nije evidentirano