Towards a method for forecasting earthquakes in Iceland using changes in groundwater chemistry
Summary
This study extended groundwater-chemistry monitoring at a northern Iceland artesian well through 2023 and tested whether criteria derived from the authors’ 2014 study would have identified later earthquakes. Of three magnitude-5-or-greater earthquakes during 2014–2023, the strongest event could have been identified four to six months in advance, whereas the other two were missed. Depending on the chosen forecast window and anomaly definition, estimated sensitivity was 20–32%, false-positive rate 1–3%, and positive predictive value 62–85%. Because the assessment was performed retrospectively at one site, the authors state that the method requires prospective validation before it can be considered an operational forecasting tool.
Source
Communications Earth & Environment
Keywords
groundwater chemistry, earthquake forecasting, hydrogeochemical anomolies, stable isotopes, time-series analysis, Iceland
Source Type
research article
Key Findings
Applying the predefined four-to-six-month window to the 2014–2023 extension, groundwater-chemistry anomalies preceded the strongest later earthquake, magnitude 6.0 in 2020, but did not identify the magnitude 5.0 event in 2018 or the magnitude 5.2 event in 2022. Across the full five-earthquake record, at least three events followed identified anomalies. The retrospective forecasting analysis estimated sensitivity of 20–32%, meaning that approximately 68–80% of qualifying earthquakes would have been missed under the tested criteria. The result supports further site-specific prospective testing but does not establish dependable earthquake forecasting.
Limitations
The method was evaluated retrospectively and remains specific to one artesian well and one Icelandic region. Only five qualifying earthquake sequences occurred across the full record, producing wide uncertainty around performance estimates. The later analysis used criteria derived from the same monitoring program, and it was not a blinded or preregistered real-time forecasting trial. Its low sensitivity means most qualifying earthquakes were not identified. Independent replication at other sites is still needed. No ULF magnetic measurements were analyzed, so the study does not test groundwater–ULF co-occurrence or physical coupling.
Methods Summary
Researchers extended weekly sampling from the Hafralækur artesian well to cover October 2008 through 2023. Stable hydrogen and oxygen isotopes were measured by cavity ring-down spectroscopy, and dissolved elements were measured by ICP-OES. Derived isotope parameters were used to distinguish groundwater mixing from water–rock interaction. An anomaly was defined using deviations of at least two standard deviations from the cumulative mean. A binomial test and retrospective calculations of sensitivity, false-positive rate, and positive predictive value compared four-, five-, and six-month forecast windows with magnitude-5-or-greater earthquake sequences affecting northern Iceland.