Why rivers ignored the barometer: the secret was in the thermometer

Original: "Redefining climate index with sea surface temperature of key area can better express atmospheric physical activity"
arXiv:1709.07146 · 2017-09-21 · CC BY-SA 4.0 · 1 min
Scientists found a way to predict river floods simply by measuring the temperature difference between different parts of the ocean.

For a hundred years, climatologists measured the ocean's pulse with a barometer, but they should have used a thermometer. Indices like El Niño are built on the atmospheric pressure difference between two points, but rivers barely respond to that. The authors of the new study simply subtracted the average temperature of one ocean region from the average temperature of another, and the correlation with river runoff jumped to 0.8–0.9. For the Bangui station on the Ubangi River, the correlation reached 0.91, which is almost a perfect match.

The key is that earlier data were "smoothed" by standardization, which ate away the main signal. But the raw temperature difference preserves the physical meaning.

The most surprising part: the strength of the ocean–river link changes every six months, exactly mirroring how Earth moves closer to and farther from the Sun. This suggests that climate is driven not only by what happens on Earth but also by the planet's orbital motion. If this holds up, river floods could be predicted years in advance, knowing only the distance from Earth to the Sun.

The correlation between sea surface temperature near the Aleutian Islands and the runoff of the Indus River at Darband station reaches 0.911, comparable to the correlation between air temperature and the boiling point of water in a single kettle.

r_{xy} = \frac{\sum_{i=1}^{n}(x_i - \bar{x})(y_i - \bar{y})}{\sqrt{\sum_{i=1}^{n}(x_i - \bar{x})^2} \sqrt{\sum_{i=1}^{n}(y_i - \bar{y})^2}}
A measure of the linear relationship between two time series, used to assess the correlation between climate indices and runoff.
r_k = \sqrt{\frac{t_{\alpha}^2}{n - 2 + t_{\alpha}^2}}
Threshold value for testing the significance of a correlation at a given confidence level.
Concepts
sea surface temperature sea surface temperature anomaly Southern Oscillation index ocean-atmosphere coupling El Niño index Teleconnection Southern Oscillation ENSO climate ERSST NOAA CPC Indian Ocean Dipole relative ONI OISST v2.1 Niño 4 atmosphere natural variability Satellite layers La Niña annual cycle