Febriani et al. (2014) report changes in ultra low frequency (ULF: 0.001–5 Hz) geomagnetic field data a few weeks before the 2 September 2009 Tasikmalaya earthquake (M7.5, hypocentral depth 57 km) from a ground-based sensor at Pelabuhan Ratu, West Java, Indonesia, 135 km from the epicentre. This was the largest and, according to the authors, the only earthquake preceded by anomalous magnetic changes of 12 M > 5 earthquakes that occurred offshore south of Java from 1 September 2008 to 31 October 2010.

Febriani et al. (2014) suggest that the magnetic changes they reported may have been induced by an alleged preparatory phase of the earthquake. The idea that electromagnetic precursors may appear before earthquakes is based on the hypothesis that earthquakes have a preparatory phase. That is, the earthquake initiates in a preparation zone (whose size depends on the magnitude of the earthquake) where physical phenomena lead to the subsequent shock and to the possible appearance of precursory signals (see, e.g. Dobrovolsky et al., 1979). However, many researchers disagree that earthquakes have a preparatory phase (see, e.g. Geller, 1997; Kagan, 1997). According to them earthquakes appear to be chaotic, scale-invariant phenomena controlled by the local physical properties of the fault whose geometry and frictional characteristics determine the starting and stopping of the rupture. Therefore, any small shock may grow into a stronger earthquake, and how big the quake will become is determined by how it is stopped, and not by how it starts (Johnston, 2015). Therefore, the notion of a preparatory phase of earthquakes appears to have no physical basis.

There are many papers (see the references section in Masci, 2010, 2011a, 2013) in which the authors report pre-earthquake changes in ULF magnetic field data suggesting a possible relationship between the changes they identified and the impending earthquake. Conversely, recent reports (see e.g. Campbell, 2009; Thomas et al., 2009a, b; Masci, 2010, 2011a, b, 2012, 2013; Masci and De Luca, 2013; Masci and Thomas, 2013a, b, 2015) have shown that many of these pre-earthquake changes are, indeed, global-scale variations driven by the frequent disturbances in the geomagnetic field, or they are generated by instrumental malfunction. These papers have cast serious doubt on the idea that ULF magnetic anomalies are convincing and always recurring phenomena preceding large earthquakes. Therefore, at present ULF magnetic disturbances cannot be considered a promising candidate for developing earthquake prediction capabilities. We note that Febriani et al. (2014) ignore the findings of the recent reports in which it has been shown that many ULF magnetic changes reported to occur before earthquakes are not precursors. They, in fact, refer to these invalid precursors as support of the search for precursory signatures of earthquake in ULF magnetic data (see Table S1 in the Supplement). In support of their findings, they also refer to the Hattori's empirical relationship between the earthquake magnitude and the distance from the earthquake epicentre of the ULF station where the pre-earthquake anomaly has been detected (see Febriani et al. (2014), Fig. 10). In Fig. S1 of the Supplement, we show this relationship where we have highlighted with red dots alleged ULF magnetic precursory changes that have been proven invalid. In Table S2 we report the papers in which these alleged precursors have been denied. Note that the empirical relationship shown in Fig. S1 is taken from Febriani et al. (2014) and is not derived from undisputed precursors. Thus, we conclude that Febriani et al. (2014) were motivated to search for precursory signals in magnetic data by reports of false precursors of earthquake.

Febriani et al. (2014) analyse nighttime (16:00–21:00 UT) geomagnetic field data in the frequency range 10 ± 3 mHz. They calculate the ratio between the spectral intensity of the vertical and each horizontal magnetic field components, i.e. the so-called spectral density ratio. According to Febriani et al. (2014), the magnetic data analysed are very disturbed by artificial noise even during nighttime. Thus, before performing the spectral analysis based on wavelet transform, they remove the intense transient signals. Then, they use the minimum energy method in an attempt to further reduce the noise. More precisely, for each day, they divide 4 h (16:30–20:30 UT) of magnetic data in eight 30 min intervals. Data before 16:30 UT and after 20:30 UT are excluded due to the edge effect of the wavelet transform. Then, the energy of the geomagnetic field vertical component Z (the component usually more disturbed by artificial noise) is calculated in each 30 min interval. Finally, the spectral density ratio is calculated in the interval where Z shows the minimum energy. Febriani et al. (2014) investigate the scaling proprieties of the geomagnetic field components by means of detrended fluctuation analysis (DFA) as well. DFA is a well-established method to extract quantitative time dynamic in time series. The DFA α exponent can be considered as an indicator of the roughness of the time series: the higher α is, the smoother the time series (Peng et al., 1995). α may be related to the fractal dimension D by the relationship D = 3 - α.

In Fig. 1 we show the spectral density ratio SZ/SY (where Y is the east–west component of the geomagnetic field) and the DFA α exponent of the Z component, as reported by Febriani et al. (2014, Fig. 9) 30 days before and after 2 September 2009. According to them, a magnetic anomaly is identified when the exponent α and the ratio SZ/SY exceed the threshold value of α‾ - 2σα and SZ/SY‾ + 2σSZ/SY, respectively. Mean values and the corresponding σ are calculated over the 2-month period in Fig. 1. Based on their definition of an anomaly, Febriani et al. (2014) report to have found anomalous changes prior to the Tasikmalaya earthquake. More specifically, a few weeks before the earthquake, they note a decrease of the exponent α, which corresponds to an increase of ratio SZ/SY (see shaded areas in Fig. 1). Febriani et al. (2014) maintain that the decrease of α in correspondence with the increase of the spectral density ratio identifies a precursory signature of the Tasikmalaya earthquake in magnetic data. No changes in SZ/SY and α are shown coincident with the earthquake when the primary energy is released.

We disagree with Febriani et al. (2014). First, there is no physical reason that magnetic anomalies, whatever their origin might be, are identified when the exponent α and the spectral SZ/SY exceed the threshold values they assumed. Then, their method for checking the geomagnetic conditions by means of the Dst index is not rigorous. We agree that geomagnetic activity should be a key parameter in interpreting observed pre-earthquake ULF magnetic changes (see Balasis and Mandea, 2007). ULF disturbances from the ionosphere and magnetosphere, indeed, may lead researchers to interpret erroneously the origin of magnetic anomalies they identified (see, e.g. Masci, 2010, 2011a). The 3 h global geomagnetic index Kp and the daily sum ΣKp are usually used as representative of the geomagnetic activity over planetary scales (Menvielle and Berthelier, 1991). Conversely, the Dst index that Febriani et al. (2014) use for checking the geomagnetic conditions is designed to monitor the strength of the equatorial electrojet, and it is usually used as indicator of the geomagnetic storm level and ring current intensification (Mayaud, 1980).

As expected, in Fig. 1 we note many decreases of α in correspondence with increases in the spectral density ratio. This inverse correspondence may be explained by taking into account that the spectral density ratio, the DFA α exponent, and the fractal dimension D of the ULF geomagnetic field are sensitive to global trends in geomagnetic activity (see Masci, 2010, 2011a; Wanliss et al., 2014). Namely, when the geomagnetic activity decreases, the reduction of the geomagnetic field horizontal component is usually larger than the reduction of the vertical component. Therefore the spectral density ratio increases. At the same time, the decrease of the geomagnetic activity indicates that the magnetosphere evolves toward a lower degree of organization (see, e.g. Balasis et al., 2009). Thus, the fractal dimension of the geomagnetic field increases, while the DFA α exponent decreases. On the contrary, an increase of the geomagnetic activity induces a decrease of the spectral density ratio (because the increase in the geomagnetic field horizontal components is larger than the increase of the vertical component) and a decrease of the fractal dimension and an increase of α (because the magnetosphere evolves towards a higher degree of organization). Thus, we expect to find an inverse correspondence between ΣKp and the spectral density ratio and the fractal dimension of the geomagnetic field, and a direct correspondence between ΣKp and the α exponent. However, due to global averaging used to calculate Kp, this correspondence is not expected always or everywhere. In this perspective, recent papers (see Masci, 2010, 2011a, 2013, and other papers reported in Tables S1 and S2) have demonstrated that many pre-earthquake ULF magnetic changes hypothesized to be seismogenic are, instead, part of global geomagnetic activity changes. In Fig. 1 we have used the same approach adopted in these papers by comparing the exponent α and the ratio SZ/SY reported by Febriani et al. (2014) with the ΣKp index. In Fig. 1a, as expected, we note a close correspondence between α and ΣKp, both before and after the earthquake. A close inverse correspondence can be also seen in Fig. 1b between ΣKp and the ratio SZ/SY calculated without the minimum energy method. However, we would like to point out that we should not expect to always find this correspondence, since (i) as stated by Febriani et al. (2014) the high environmental noise in the geomagnetic field components was not attenuated enough after removing intense transient signals; (ii) several gaps are present in α and SZ/SY time series; (iii) SZ/SY shows many inexplicable zero values; (iv) α and SZ/SY are calculated from local magnetic data, whereas, as already mentioned above, ΣKp is representative of daily averaged geomagnetic disturbances on a planetary scale. Contrary to Fig. 1b, however, in Fig. 1c we see a lower correspondence between SZ/SY calculated applying the minimum energy method and ΣKp. The lower correspondence may be explained considering that for each day Febriani et al. (2014) calculate the spectral density ratio, using the minimum energy method, in one of the eight 30 min intervals between 16:30 and 20:30 UT. Since ΣKp is representative of global daily averaged geomagnetic disturbance, by reducing the period of analysis, it is likely that the correspondence between geomagnetic data and ΣKp becomes less noticeable. Thus, the high dispersion of SZ/SY values in Fig. 1c may be due to the short time interval (30 min) used in the spectral analysis, as well as because the SZ/SY time series consists of values that are calculated in different 30 min intervals.

We have reviewed the findings of Febriani et al. (2014) that show pre-earthquake changes in magnetic field record before the M7.5 Tasikmalaya earthquake occurred on 2 September 2009 south of Java. We have shown that the changes they reported in the DFA α exponent of the geomagnetic field vertical component and the spectral density ratio SZ/SY are closely related to the geomagnetic ΣKp index and are unlikely to be of seismogenic origin. Thus, we conclude that the pre-earthquake magnetic changes reported by Febriani et al. (2014) are an effect of the global geomagnetic activity.