One of the major challenges of seismology is to predict earthquakes. However, despite applying modern space and/or ground based technologies, a reliable method for short-term earthquake forecasting with reliable warnings several hours or days in advance is currently not possible and remains limited to only a few minutes before the quake happens. During the past decade, earthquake precursors, like ionospheric perturbations, electromagnetic phenomena, ground heating and composition changes of soil-efflux (radon, CO₂, etc.), have been scientifically investigated. However, they are far from being successfully applicable, because these precursors are highly regionally variable depending not only on the large-scale tectonic main stress regime, but also on the small-scale tectonic regime [1,2,3].

Abnormal animal behaviors have been reported previously, but create problems in monitoring and reliability [4,5,6,7]. However, long-term in situ data collection, statistically significant validation of these precursors in focus regions over a long period of time and plausible scenarios explaining the evolution of such behaviors are still missing.

The situation is different with red wood ants (RWA; Formica rufa-group (Hymenoptera: Formicidae). A particular advantage of RWA is their high sensitivity to environmental changes. Forced by some million years of evolutionary selection, they may have developed anticipatory mechanisms [8,9]. Besides an extremely strong temperature sensitivity (0.25 K) [10], they have chemo-receptors for the detection of CO₂-concentrations [11,12] and an electromagnetic field sensitivity [13,14,15]. Recent research in the West Eifel (West Germany), combined with geostatistical analyses, have demonstrated the correlation of soil gas anomalies, especially helium anomalies, as outstanding fault zone tracers and spatial distribution of red wood ant (RWA) mounds along tectonically active, gas-permeable faults (cf. Figure 1) [16,17].

At the reference location Oberehe (cf. Figure 2(a)), a site with small soil/sediment covers, a direct degassing of geogenic gases into the atmosphere occurs. The linear alignment of the RWA mounds directly corresponds to that of the helium anomalies. The Hough transform (a well-established algorithm for identifying linear structures in sets of points [18]) applied on the spatial distribution of RWA mounds (cf. Figure 2(b)) in the West Eifel (1,140 km²) and the directions of fault zones published in [19,20] (cf. Figure 2(c)) shows that the centers of the modes, i.e., the local maxima in the histograms, denote the preferential alignment directions, while their widths indicate the corresponding directional variations. In the RWA histogram, the local maxima are separated by vertical black lines. For each mode, the average value is indicated by a solid vertical red line and the ±1 standard deviation interval surrounded by two dashed vertical red lines. More specifically, the azimuthal directions in which the RWA mound positions are aligned correspond to those of the main fault directions. The comparison of both the spatial RWA mound distribution with the significantly increased gas anomalies confirms that the RWA colony is situated in a tectonically sheared region. Although the reference site is surrounded by various forest stands, no further RWA mounds were found in those forest stands during the area-wide mapping. The statistical analyses show that the spatial distribution of RWA mounds maps tectonically active, gas-permeable strike-slip faults (cf. Figure 2) and that RWA mounds can be used as biological indicators of tectonically active, gas-permeable strike-slip faults [16,17,21]. This is especially useful when information about the active tectonic regime is incomplete or the resolution by technical means is insufficient [16,17]. These faults may be potential earthquake areas and are simultaneously information channels deeply reaching into the crust. Recent research shows that, among others, electromagnetic field changes emitted from the pre-focal area several days before the upcoming earthquake are discussed as one trigger mechanism for bioanomalies prior to large earthquakes [22,23,24,25]. Also mid-term to short-lived thermal infrared anomalies from the rock surface in the 8–12 μm region along thrust or strike-slip faults prior to major earthquakes are reported [26,27,28]. Consequently, an integrated approach in understanding geophysical phenomena requires consideration also of the interaction between the abiotic and biotic environment.

For a detailed and statistically well-based study, it is necessary to investigate whether changes in the ants’ standard daily routine can be correlated with local seismic events. The seismically active study area at the Neuwied Basin (Eifel, Germany) is located in an area with a complex tectonic history. Commencing during the Neogene and persisting during the Quaternary, the uplift of the Rhenohercynian Zone affects the dynamics of the study area. The currently NW-SE striking main stress direction opens pathways for geogenic gases and potential magmas [29,30,31]. Concurrently, a conjugate dextral wrench fault system (striking in 100°–110° WNW-ESE representing the extensional regime of the recent compressional stress field and a conjugated shear system trending in 140°–150° NNW-SSE) in combination with a clockwise block rotation and relatively small kinematic WSW-movements exists [29,32,33,34]. At the same time, Variscan faults as part of a conjugated shear system are reactivated [35,36]. This situation has led to a complex tectonic break clod. Due to the ongoing extension of the Cenozoic rift system in the western part of the Eurasian plate, geotectonic processes still influence the active regional tectonics and lead to a large number of low-magnitude seismic events [37,38]. In the study area, an average of about 100 earthquakes per year with magnitudes up to M 3.9 occur located on different tectonic fault regimes (strike-slip faults and/or normal or thrust faults) [29,39,40].

For three years, from 2009 to 2012, we have monitored two RWA mounds (Formica rufa-group; F. pratensis and F. polyctena, 24/7 by high-resolution cameras. Both mounds are located at two different fault regimes (on a WNW-ESE and on a NNW-SSE trending fault, respectively) approximately 22 km apart. The location of the two cameras is shown in Figure 3. We refer to the cameras as AntCams. During the observation period, ten magnitude > 2.0 earthquakes in a 40 km radius occurred (cf. Table 1).

Early results of the manually analyzed long-term observations show that ants have a well-identifiable standard daily routine that can be best characterized by an M-shaped curve: activity rises from dawn until approximately 11 A.M. Activity peaks can be observed at midday and in the late afternoon (cf. Figure 8(a)). During the evening, the activity is reduced again. This approach is contradictory to the observation documented so far [49]. It shows that it is possible to obtain an objective and testable null hypothesis and detailed data by a long-term, 24/7 in situ monitoring of the ants’ behavior using suitable technical equipment and experimental setup. A previous study [49] merely considers a serendipitous event. Additionally, no further information is provided on whether the acquired very short (30 s) single video streams have been objectively analyzed by an automated routine.

The video observations and the times of occurrence of local earthquakes indicate changes in the ants’ behavior hours before the earthquake took place: the nocturnal rest phase and daily activity were suppressed, and standard daily routine did not resume until the next day (cf. Figure 8(b)). The results show that it is essential to monitor the ants’ behavior before the earthquake strikes a region. The eight minutes observation period in the run-up of the Landers M 7.4 earthquake in the Mojave Desert is too short and cannot produce statistically significant results, as it cannot be determined for such a short period of time whether the observed ants’ behavior is normal or abnormal. To define the “normal” state, it would have been necessary in that scenario to observe the ants’ behavior over longer periods of time during earthquake-free intervals in order to obtain a statistically relevant reference for observations during earthquakes. Furthermore, our observations suggest that a documentation of the aftershock activity alone is insufficient and that the conclusion that ants do not show any reaction on earthquakes, as it was made in the literature so far, is not justified based on the limited amount of data examined in previous studies [49]. Another prerequisite is to have detailed information on the active tectonic regime in the corresponding region and an understanding of geotectonic processes for interpretation of geo-biological processes, i.e., RWA as bioindicators for the identification of active tectonic systems. Observations some 100 km away from the epicenter should be interpreted very carefully, since tectonic fault types and systems can change strongly within this distance. Additionally, it has not been examined so far, e.g., by geochemical analyses, if Messor pergandei has a close relation to tectonic fault systems.

One of the major challenges of seismology is to determine whether fault zones generate any indication of impending earthquakes. The evolutionary mechanism of exaptation, i.e., a vibration triggered early warning response or animals’ perception of these stimuli to P-waves, is reported in [5]. Precursors, as electromagnetic phenomena or the change of the gas content and concentration, might be sensed by the ants due to their electromagnetic sensitivity and ability to detect CO₂ [16,17,18]. The objective of this methodological approach was to examine if ants show reactions prior to earthquakes. The specific mechanisms for how and why they show such reactions need to be researched in in situ experiments and in close cooperation with biologists.

In the night from September 19 to 20, a mismatch between the manual and the automatic analyses is apparent (Figure 7(b)). Analyses of the corresponding video material showed that the algorithm works correctly and clearly detects and counts the ants’ movements. A possible explanation is the change in brightness of the material on top of the mound at night that does not play a role for the manual analyses, but for the automated ones. It is necessary to further analyze this issue in long-term studies.

Although we have monitored two different species (F. pratensis and F. polyctena), their daily activity patterns do not show systematic differences at days without any earthquake event and can be considered the same. According to the correlation coefficients (R between 0.82 and 0.97), the statistical correlation can be described as strong. This highly similar behavior of the two different species, as presented in Figure 9, is contradictory to previous studies and the common tenet [50,51].

External influence factors, like predators (woodpeckers, blackbirds, etc.) or occasionally “anting” of birds to banish parasites with formic acid, which only last for some seconds to some minutes at maximum, create only a slight increase of the peak activity, but do not have any influence on the ants’ daily routine. In practice, such influences can be avoided by fencing the mound.

The external influence of larger animals, like badgers or boars, does not show raised activity peaks, unless a part of the mound is affected by digging (badger) or destruction (woodpecker, black bird etc.). Then, the ants’ activity rises to protect the mound and the population. This can be documented by a peak activity (e.g., reconstruction of the mound), which amounts to only a small fraction of the daily routine and is negligible, for the daily routines have to be considered across the whole day. Nevertheless, this extraordinary situation is documented (cf. Figure 10).

During the winter season, the ants’ activity is—as it is well known—very small to non-existent. However, we have observed ants to be visible on top of the mound when an earthquake with a magnitude of about 2 occurs, despite snow cover and freezing temperatures and, moreover, with delayed activity compared to the spring-autumn seasons. The question is whether the ants’ activity will rise to a higher level for earthquakes with higher magnitudes.

The automatically analyzed activity index (“purple curve” in Figure 11) shows that our image analysis system works independently of given manual results. Furthermore, a comparison of the manually inspected video observations (green curve) and the “raw” activity indices, A_d(t) and A_n(t) (prior to the mapping to the manually inferred categories C(t) based on the polynomial regression functions) (blue curve), shows that both activity curves show the same trends in the daily ants’ routine.

According to our first analyses, clearly anomalous ants behavior may occur for the local earthquake with a magnitude > 2 (Figure 8(b)). At lower magnitudes (cf. Table 3), an anomalous behavior may possibly be present as well (Figure 12). The question is: how RWA would react in regions with higher magnitude (M ≥ 4) earthquakes and if there is a habituation effect with regard to lower magnitudes. Possibly, the ants show reactions only for higher magnitudes. A comparison of the earth tides and the earthquake events suggest a possible correlation for the given time interval. An accumulation of earthquakes (June 09 (day 161); June 12–14 (days 163–165) and June 26–29 (days 177–180)) occurs only when vertical displacements are positive and at their maximum. Also, nuptial flights occur only when vertical displacements are positive at their maximum in the given time interval (June 22–30 (days 174–181)). It is necessary to further analyze these relationships in long-term studies.

Possibly, the ants’ behavior may also depend on the specific fault type. Strike-slip faults create deep-reaching tectonic fractures that might be gas-permeable and show mineralizations. Earthquakes occurring on these faults would trigger gas anomalies prior to the event that can propagate over a certain distance from the future quake epicenter, whereas along normal or reverse faults, due to their physical characteristics, such anomalies would be smaller.

RWA bioanomalies (suppression of the nocturnal rest phase and daily activity, continuation of the standard daily routine not before the next day) have been recorded and analyzed for several earthquakes with magnitudes of up to about 3. The automated image analyses routine has been shown to provide a valuable tool to objectively identify and classify the ants’ activity on top of the mounds and an examination of its correlation with earthquakes. The results show that the automated system works and can be utilized in practice.

According to our first analyses, clearly anomalous ant behavior occurs for the local earthquake on September 10, 2009, with a magnitude of 3.2. At lower magnitudes, an anomalous behavior may possibly be present as well. The specific mechanisms for how and why they show such reactions need to be researched in in situ experiments and in close cooperation with biologists. An open question is whether RWA would react in the same way on earthquakes with higher magnitudes (M > 4), which have not occurred during our period of observation. This will be one topic of future research. In the next step, the analyses routine will be integrated into the AntCams to transmit the activity indices online and in real-time. After a test phase of about half of a year, we plan to install our monitoring system in tectonically more active regions.

Although the investigation and results presented here are promising, they are only a first step towards a completely new research complex. Long-term studies have to show whether confounding factors and climatic influences can clearly be distinguished. However, our early results suggest that it makes sense to consolidate and extend the research to determine a pattern for exceptional activity situations. These studies are not provided for earthquake prediction, but are an important step towards the understanding of geobiological processes.