Abstract
An amendment to this paper has been published and can be accessed via a link at the top of the paper.
Introduction
This post summarizes a significant Author Correction published in Scientific Reports regarding the causal link between solar activity and ocean warming. A coding error involving an incorrect file path led researchers to use the wrong solar dataset in their original study. After re-running the analysis with the correct data, the detected causal link is significantly weaker and occurs over shorter timescales than previously reported, though the authors maintain that a partial causal relationship still exists.
Methodology: Quantifying Causality
To move beyond simple correlation, the authors utilize the Liang–Kleeman information flow method. This tool from dynamical systems theory quantifies the transfer of information from one variable to another—effectively measuring causality. To see how this relationship changes over time and across different cycles, they employed:
- Multi-scale analysis: Breaking the data down into different time frequencies (e.g., 10-year cycles vs. 40-year cycles).
- Sliding window approach: Analyzing the data in 22-year "chunks" to see how causality fluctuates over the decades.
- Wavelet coherence: A method to see where two time series "sync up" in both time and frequency.
The Data Correction
The correction centers on a data error. Due to an incorrect path setting in their code, the authors originally used an unintended solar dataset. The analysis was re-done using the dataset specified in their "Data availability" section (TSI data) alongside global Ocean Heat Content (OHC) records.
Key Insights and Recalibration
The most striking aspect of this correction is how much a single data input error changed the quantitative results, even if the qualitative "story" remained similar. The "Information Flow" value dropped by an order of magnitude, highlighting how sensitive causality detection algorithms can be to the specific version or source of geophysical data used.
Shrinking Timescales
The original paper suggested the sun influenced the ocean on 20-to-40-year cycles. The corrected data suggests this influence is much more "short-term," primarily occurring on 10-to-25-year timescales.
Geographic Recalibration
Originally, the authors saw significant solar influence in the Atlantic. The corrected analysis removed the Atlantic from the list of significantly affected areas, leaving the South Indian Ocean, western tropical Pacific, and East Australia as the primary regions where solar causality is detectable at a 68% confidence level.
Key Results
The re-analysis led to several major downward revisions of the sun's measurable impact on ocean heat:
- Global Information Flow: The measured flow from TSI to global ocean heat dropped from 0.068 to 0.0051.
- Confidence Levels: The confidence that the sun is a driver of global mean ocean heat dropped from 57% to 52%, meaning the causality is not statistically significant on multi-decadal scales.
- Sliding Window Significance: In the original study, a 22-year sliding window showed significant causality at a 68% confidence level. In the corrected version, it is not significant even at that low threshold.
- PC Correlation: Wavelet analysis now shows that solar irradiation correlates mainly with the first Principal Component (PC1) of ocean heat on inter-annual scales, with nearly no significant correlation to other major modes (PC2 and PC3).
Conclusion
The primary limitation revealed here is the robustness of causality detection. When the "signal" of solar influence is as small as 0.0051, it becomes very difficult to distinguish from background noise or other forcing factors (like greenhouse gases). The authors conclude that while TSI is still a "part cause" of ENSO and global warming, its measurable footprint in the ocean is smaller and faster-acting than their initial model suggested.
Summary updated