Statistical Evaluations of Variations in Dairy Cows’ Milk Yields as a Precursor of Earthquakes
Summary
WIVVI summary (paraphrased): Researchers examined one year of daily dairy-cow milk-yield records to test whether production varied systematically before ten nearby earthquakes of different magnitudes. Cross-correlation analysis identified a negative association approximately three weeks before the earthquakes, followed by a positive association about one and a half weeks beforehand that gradually approached zero near the event dates. The researchers also compared milk-yield changes with very-low-frequency and low-frequency radio-propagation measurements interpreted as indicators of lower-ionospheric disturbances. The reported milk-yield decreases occurred earlier than those radio anomalies. These results suggest a possible temporally ordered association, but the small number of earthquakes and exploratory observational design do not establish causation, a biological sensing mechanism, or reliable earthquake forecasting.
Source
Animals
Keywords
dairy cows, milk yield, milk production, earthquake precursors, earthquake occurrence, longitudinal monitoring, time-series analysis, cross-correlation, very low frequency, low frequency, ionospheric perturbations, animal physiology
Source Type
research article
Key Findings
Cross-correlation analysis of one year of daily milk-yield records and ten earthquakes found a statistically significant negative association approximately three weeks before the earthquake dates. A positive association appeared roughly one and a half weeks beforehand and diminished as the earthquake dates approached. Very-low-frequency and low-frequency radio-propagation anomalies were reported before all ten earthquakes and occurred after the milk-yield decreases. The temporal pattern is notable, but it does not demonstrate that the earthquakes caused the production changes or that milk yield can reliably predict future earthquakes.
Limitations
The analysis covered only one year and ten earthquakes, providing limited statistical power and preventing robust assessment across different years, farms, herds, regions, seasons, and seismic settings. Milk production is affected by lactation stage, pregnancy, health, feed, temperature, humidity, husbandry, farm operations, and other stressors that may produce temporal patterns unrelated to earthquakes. Cross-correlation across multiple time lags can identify apparently significant relationships by chance unless testing choices, multiplicity, autocorrelation, and model validation are rigorously controlled. The study did not use a prospectively declared alarm rule or an independent validation dataset and therefore could not establish sensitivity, specificity, false-alarm frequency, or practical forecasting performance. The ordering of milk-yield and ionospheric changes does not demonstrate that they shared a cause, and the proposed involvement of ultralow-frequency magnetic fields remains hypothetical.
Methods Summary
Researchers analyzed one year of routinely recorded daily milk yields from dairy cows and selected ten earthquakes using magnitude and distance criteria intended to identify events potentially relevant to the farm. Expected production patterns and known influences on milk yield were considered when deriving daily variations. Cross-correlation analysis was used to examine statistical relationships between milk-yield changes and earthquake occurrence at different lead times. The timing of the milk-yield results was then compared with very-low-frequency and low-frequency radio-propagation measurements used as indicators of possible lower-ionospheric disturbances. The study was observational and retrospective rather than a prospective forecasting trial.