A Meta‐Analysis of Studies Attributing Significance to Solar Irradiance

• Published in: Earth and space science (Hoboken, N.J.) Vol. 10; no. 1
• Main Authors: Vachula, Richard S., Cheung, Anson H.
• Format: Journal Article
• Language: English
• Published: Hoboken John Wiley & Sons, Inc 01-01-2023; American Geophysical Union (AGU)
• Subjects: Age; Hypotheses; Hypothesis testing; Literature reviews; Meta-analysis; Paleoclimate; Radiation; Statistical analysis; Time series
• ISSN: 2333-5084; 2333-5084
• DOI: 10.1029/2022EA002466

The relationship between solar irradiance and climate is greatly debated. This inferred relationship is often characterized via the statistical analysis of paleoclimate data. REDFIT is a commonly used statistical tool that overcomes uneven sampling to identify significant periodicities of variability in proxy data. We critically examine the use of REDFIT to identify solar signals in these data. By conducting a literature review, we show the REDFIT significance thresholds used by researchers to analyze paleoclimate data vary considerably. As there is some subjectivity and practicality involved in any statistical analysis, some variability is to be expected. However, we observe that the bulk of the significance thresholds used in the literature are less stringent than the critical false‐alarm level outlined by REDFIT's creators. We reexamine periodicities deemed “significant” in a published data set to show that using this more stringent, more objective critical false‐alarm threshold likely eliminates the previously inferred significance of solar signals in proxy data. Likewise, we address a lack of consideration of age model uncertainty on REDFIT's reliability in identifying solar periodicities. Overall, we show that the relationship between solar irradiance and climate, as identified by REDFIT analyses, may not be as robust as previous work might suggest.
Plain Language Summary
We know the sun has cycles but it is sort of unclear whether or not they cause Earth's climate to change. Much research has identified these cycles in natural climate archives (like marine sediment cores, tree rings, and stalagmites) with a statistical program called REDFIT. But, it can be difficult to determine the appropriate cut‐off with which REDFIT identifies “significant” cycles. In other words, different researchers have used a variety of significance thresholds, so these analyses have not been as strictly objective or consistent as we might hope. This is a problem because it means the solar cycles identified might just be noise in the data. Also, insufficient attention has been paid to the large age uncertainties of these climate archives and what their impact on the solar cycles identified by REDFIT analyses might be. We show that age uncertainty can change whether or not a solar cycle is deemed significant in a REDFIT analysis of a paleoclimate data set. Overall, we showcase an inconsistency of methods in the paleoclimate literature. We should probably be more transparent and conservative about how we use this statistical program.
Key Points
Solar‐climate relationships are identified in proxy data using REDFIT, but studies do not use the most conservative significance thresholds
Age model uncertainty is an overlooked complication for the identification of solar‐climate relationships
The link between solar irradiance and climate may not be as robust as previous REDFIT‐based analyses might imply