A review on remotely sensed land surface temperature anomaly as an earthquake precursor

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

Contents lists available at ScienceDirect

Int J Appl Earth Obs Geoinformation journal homepage: www.elsevier.com/locate/jag

Review article

A review on remotely sensed land surface temperature anomaly as an earthquake precursor

MARK

⁎

Anshuman Bhardwaja,b, Shaktiman Singhb,c, Lydia Samb,c, P.K. Joshid, , Akanksha Bhardwaje, F. Javier Martín-Torresa,f, Rajesh Kumarb a

Division of Space Technology, Department of Computer Science, Electrical and Space Engineering, Luleå University of Technology, Luleå, Sweden Department of Environmental Science, Sharda University, Greater Noida, India c Institut für Kartographie, Technische Universität Dresden, Germany d School of Environmental Sciences, Jawaharlal Nehru University, New Delhi, 110 067, India e Banaras Hindu University, Varanasi, India f Instituto Andaluz de Ciencias de la Tierra (CSIC-UGR), Armilla, Granada, Spain b

A R T I C L E I N F O

A B S T R A C T

Keywords: Earthquake Land surface temperature (LST) LST anomaly Precursor Remote sensing

The low predictability of earthquakes and the high uncertainty associated with their forecasts make earthquakes one of the worst natural calamities, capable of causing instant loss of life and property. Here, we discuss the studies reporting the observed anomalies in the satellite-derived Land Surface Temperature (LST) before an earthquake. We compile the conclusions of these studies and evaluate the use of remotely sensed LST anomalies as precursors of earthquakes. The arrival times and the amplitudes of the anomalies vary widely, thus making it difficult to consider them as universal markers to issue earthquake warnings. Based on the randomness in the observations of these precursors, we support employing a global-scale monitoring system to detect statistically robust anomalous geophysical signals prior to earthquakes before considering them as definite precursors.

1. Background Earthquakes are one of the most sudden, most difficult to predict, and therefore most destructive natural phenomena. During the last few decades of the 20th and the first decade of 21st century (1968–2008), nearly 18,807 earthquakes of magnitudes > 4.5 have been reported by the Prompt Assessment of Global Earthquakes for Response (PAGER) system developed by the United States Geological Survey (USGS) (Marano et al., 2010). The associated casualty count has risen sharply in recent decades due to the rapidly increasing global population, with the staggering count of ∼7,00,000 casualties in the first decade of the 21st century that may increase to a predicted count of 2.57 ± 0.64 million in the latter half of the 21st century (Holzer and Savage, 2013). The year 2015 alone saw several deadly earthquakes across the globe (Fig. 1), motivating a streamlining of precursory studies (Bhardwaj et al., 2017; Daneshvar and Freund, 2016). 1.1. Prediction and precursors Earthquakes are the result of surging tectonic stress and are extremely difficult to predict (Geller et al., 1997) due to a lack of distinct

⁎

statistical patterns required to model future occurrences (Console et al., 2002). However, the degree of the unpredictability of earthquakes is highly debated amongst seismologists, with several publications optimistically highlighting the usefulness and need for the continuous research on earthquake precursors to improve earthquake predictions (Wyss, 1997). The confidence in the science of short-term earthquake prediction received a boost in 1975, when a warning was issued in Haicheng, China hours before a major M7.4 earthquake, saving many lives (Cicerone et al., 2009). However, it soon received a setback due to the failure to predict the 1976 M7.8 earthquake in Tangshan that caused massive devastation (Lomnitz, 1994). Verma and Bansal (2012) have given many accounts of the successes and failures of earthquake predictions, sparking optimism and pessimism, amongst researchers, in turn. Nevertheless, the truth as of now is that the available seismological techniques have limited use for precisely forecasting the times, locations and strengths of earthquakes (Saraf et al., 2009). This indicates a need to strengthen forecasts using various proxies, called earthquake precursors. Several researchers have defined the term earthquake precursor. Ishibashi (1988) proposed that earthquake precursors fell within two main categories: physical and tectonic. He defined a physical precursor

Corresponding author. E-mail addresses: [email protected], [email protected] (P.K. Joshi).

http://dx.doi.org/10.1016/j.jag.2017.08.002 Received 16 September 2016; Received in revised form 31 July 2017; Accepted 2 August 2017 0303-2434/ © 2017 Elsevier B.V. All rights reserved.

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

Fig. 1. This figure is generated using the United States Geological Survey (USGS) earthquake database for the period 1 January 2015 to 26 November 2015, highlighting 135 earthquakes of more than M6 (http://earthquake.usgs.gov/earthquakes/search/). The most destructive of these earthquakes are shown in the inset, falling within the criteria of USGS Prompt Assessment of Global Earthquakes for Response (PAGER) alert levels orange and red. The epicenters of these earthquakes were: (1) Nepal (28.147°N, 84.708°E), (2) Chile (31.57°S, 71.654°W), (3) Afghanistan (36.441°N, 70.717°E). More information on the PAGER levels can be found at http://earthquake.usgs.gov/research/pager/.

synchronous) provide an excellent opportunity for detailed and organized research on the proposed atmospheric and surficial precursors (Table 1) in different wavelengths and with sufficient temporal resolutions (Alvan et al., 2014; Bhardwaj et al., 2017; Freund, 2013; Piroddi and Ranieri 2012). In recent years, an increasing volume of literature has emerged on reporting the probable atmospheric, ionospheric, and surficial precursors for several large magnitude earthquakes (e.g., Bhardwaj et al., 2017; Daneshvar et al., 2014; Daneshvar and Freund, 2016; Ganguly, 2016; Petrini et al., 2012; Piroddi et al., 2014; Piroddi and Ranieri, 2012; Pisa et al., 2011) and remote sensing has played a crucial role in making such observations. The present review focuses on the detection of Land Surface Temperature (LST) anomalies using satellite observations before major earthquakes. The contemporary space-borne infrared (IR) sensors monitor the global thermal regime at varying spatial, spectral, and temporal resolutions (e.g., Terra/Aqua Moderate Resolution Imaging Spectro-radiometer (MODIS), National Oceanic and Atmospheric Administration-Advanced Very High Resolution Radiometer (NOAA-AVHRR) and Landsat). The repeat period for observations varies from 12 h for the sun-synchronous satellites to 30 min for geostationary satellites, thus ensuring an optimal global survey to detect changes in LST connected to an imminent earthquake (Tronin, 2006). The next section briefly describes the proposed probable physical basis behind this LST anomaly.

Table 1 Well-studied earthquake precursors. Precursor (anomaly)

Example reference

Seismicity Electric field Magnetic field Gas/aerosol emissions Ionospheric Water level changes Air temperature Land surface temperature Surface deformations Unusual animal behaviour

Ihmle and Jordan (1994) Dobrovolsky et al. (1989) Balasis and Mandea (2007) Guo and Wang (2008) Pulinets (2004) Martinelli (2000) Alvan et al. (2014) Tramutoli et al. (2005) Borghi et al. (2009) Logan (1977)

to be a direct or indirect indicator of the beginning or development of an irreversible rupture-generating physical process within the preparation zone before the onset of an earthquake. However, a tectonic precursor is defined as the manifestation of a tectonic movement outside of the preparation zone of an imminent earthquake, pertaining to a sequence of particular local tectonism (Ishibashi, 1988). The International Association of Seismology and Physics of the Earth’s Interior (IASPEI) defines both of these precursors together as any quantifiable change in regular environmental observations, as a preparatory mechanism before the main seismic event (Wyss, 1991). Hayakawa et al. (2000) refer to these precursors as earthquake sequences leading to the main earthquake shock and continuing for some time after it. Yao (2007) identified earthquake precursors as unusual changes in the entire physical and chemical environmental regimes at a regional scale. Cicerone et al. (2009) described precursors as a wide variety of physical phenomena preceding at least several earthquakes. A list of significantly studied precursors is given in Table 1. The Earth observation satellites (both geostationary and sun-

1.2. Proposed physical basis of LST anomalies Enhanced thermal infrared (TIR) emissions from the earth’s surface preceding an earthquake, which are often perceivable by remote sensors, can be called a thermal anomaly (Freund et al., 2005). Several of the studies mentioned in Fig. 2 have tried to determine the physical basis behind the thermal anomaly that causes fluctuations in the LST. These studies discuss the theoretical and experimental results, although 159

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

Fig 2. A compilation of some of the most cited postulates explaining the physical basis behind the LST anomaly.

Table 2 Hypotheses describing physical basis behind thermal anomaly. Hypothesis

Arguments

1. Earth degassing mechanism

Opening of microcracks/conduits due to a gradually increasing stress built-up; release of multiple greenhouse gases (e.g., CO2, N2O, O3, H2, He, CH4, H2S, Rn, and water vapor); increase in lower atmosphere and land surface temperatures through a localized greenhouse effect Changing groundwater levels alter soil moisture and thus change soil physical properties and temperature Frictional heating on fault surfaces as a result of seismicity Released Rn causes air ionization subsequently forming clusters of air ions along with the condensation of water vapor molecules; released large heat of condensation leads to changes in the air humidity and near-surface air temperature; temperature difference gets transferred to surface Stress-induced dissociation of the peroxy bonds (PHPs); activation of p-holes; mid-infrared (MIR) luminescence observed by thermal remote sensors

2. Groundwater anomaly 3. Frictional heating 4. Seismo-ionosphere coupling

5. p-hole activation mechanism

during the past two decades (Table 3). In the following subsections we try to organize and discuss these studies under various themes for a clearer understanding of their significance. To the best of our knowledge, we have tried to identify and discuss some of the most significant, highly cited, and peer-reviewed studies within this research domain.

the universal applicability of any one of these proposed mechanisms in isolation is debatable. Detailed discussions of these processes, their critiques and applicability can be found in several review papers (e.g., Cicerone et al., 2009; Freund, 2011; Pulinets et al., 2006; Qiang et al., 1991; Roeloffs, 1988; Saraf et al., 2009; Surkov et al., 2006; Tramutoli et al., 2001; Verma and Bansal, 2012). In Fig. 2, we present a simplified compilation of a few of the most cited hypotheses to explain the physical basis behind the pre-earthquake LST anomaly. Table 2 further illustrates the arguments for each of these hypotheses. More consideration is needed to first establish the presence of preearthquake LST anomalies with high statistical confidence before their probable mechanisms can be investigated, as there are points of uncertainty associated with the anomalies themselves, for example: (1) What is the range of these LST anomalies? (2) How confidently can they be detected via remote sensing? and (3) How reliable is the LST as an earthquake precursor? We try to analyze these aspects in view of the published studies in the coming sections. We focus on those articles specifically discussing the LST anomaly, while articles discussing TIR anomalies as a whole (the combination of both surface and atmospheric anomalies) are outside of the scope of the present review.

2.1. Pioneer works suggesting the usability of remote sensing Tronin and his coworkers have reported initial satellite-based surface thermal anomaly observations before earthquakes (Tronin, 1996, 1999, 2000a, 2000b; Tronin et al., 2002, 2004). Tronin (2000a) discussed several minor earthquakes of magnitudes 3.6-5.5 between August 1998 and February 1999 in northeastern China. A thorough twophased LST anomaly analysis was performed during periods of higher seismicity using 94 AVHRR images. Standard deviation was used to identify the anomaly and was reported to be ± 3 °C 6–24 days before the event. This work was significant as it further confirmed the idea that the presence of LST anomalies is associated with lineaments and faults that are integral parts of the earth’s seismicity. In the same year, Tronin (2000b) extended this work by adding an analysis of the Kobe and Kanto earthquakes in Japan (Table 3). However, the magnitudes of these earthquakes were larger than those of the China test sites. Tronin (2000b) reported that the Japanese LST anomaly was more localized, had a smaller precursor time (∼1 week) and had a higher amplitude (6 °C). Another highlight was that the magnitudes and distributions of the Chinese LST anomalies reported were different from those noticed

2. Literature survey of satellite-derived LST anomaly We have identified and compiled several significant LST precursory studies that use remote sensing observations and were published as research articles in peer-reviewed journals and conference proceedings 160

161 Muzaffarabad (Pakistan) Iran Western China Abruzzo (Italy) Gujarat (India) Gujarat (India) Vrancea (Romania), Yamnotri (India)‘ Bam, Zarand, Borujerd (all in Iran)

8 October 2005 June 2002–June 2006 November 2001-May 2008 6 April 2009

26 26 27 26

(σ) reflects that the numerical values represent standard deviation.

Saravan (Iran) Abruzzo (Italy)

16 April 2013 6 April 2009

a

Italy

20 May 2012, 29 May 2012

January 2001 January 2001 October 2004, 22 July 2007 December 2003, 22 February 2005, and 31 March 2006

6.1–7.9

Kalat (Pakistan), Izmit (Turkey), Bhuj (India), Hindukush (Afghanistan), Boumerdes (Algeria), Bam (Iran) Bhuj (India) Boumerdes (Algeria) Izmit (Turkey) Muzaffarabad (Pakistan) Bam, Zarand (both in Iran) Izmit (Turkey), Hector Mine (USA), Bhuj (India), Kunlun (Tibet), Colima (Mexico), Boumerdes (Algeria)

7.7 6.3

5.8, 5.9

7.6 7.6 5.9, 5.1 6.1–6.6

7.6 5.8–6.6 7.3–9 5.8

7.9 6.8 7.8 7.8 6.6, 6.4 6.8–7.9

3.6–5.5 4.9–6.9 6.9 4.1–7.8 5.9 7.7 6.3–7.5

North-east China Kobe (Japan), Kanto (Japan), North-east China Irpinia (Italy) North-east China, Kanto (Japan) Athens (Greece) Gujarat (India) Kamchatka peninsula (Russia)

August 1998-February 1999 16 January 1995, 21 December 1996, 29 January 1999 23 November 1980 December 1994-February 1999 7 September 1999 26 January 2001 24 June 1983, 8 June 1993, 21 June 1996, 8 October 2001 4 March 1990, 17 August 1999, 26 January 2001, 25 March 2002, 21 May 2003, 26 December 2003 26 January 2001 21 May 2003 17 August 1999 8 October 2005 26 December 2003, 22 February 2005 17 August 1999, 16 October 1999, 26 January 2001, 14 November 2001, 21 January 2003, 21 May 2003

Magnitude

Study area

Earthquake Date

2–10

∼2 days-1 week before 3 days before ∼1 week before 8 days before Not specified ∼6 days before 4–6 days before

6

∼3 weeks before

Not specified Not specified

Not specified ±4 5–10 3–12

∼2 weeks before ∼9 days before 7–8 days before 1–20 days before

4 days before 15 days before

4–8 2–13 ±1 Not specified

7 days before 1–10 days before ∼2 weeks after 20 days before

5–7 5–10 0.5–5 (σ) 3 5–10 ±4

±3 ± 3–6 Not specified ± 3–6 0.5–5.5 (σ)a 3–4 ± 2–10

Anomaly (°C)

6–24 days before ∼1–3 weeks before Not specified 6–24 days before ∼19 days before ∼5 days before 4–7 days before

Precursor appearance

NOAA-AVHRR NOAA-AVHRR NOAA-AVHRR Terra MODIS NOAA-AVHRR Landsat NOAA-AVHRR Terra MODIS Aqua MODIS Terra MODIS NOAA-AVHRR Terra MODIS Terra MODIS Aqua MODIS Terra MODIS Terra MODIS NOAA-AVHRR Terra MODIS Aqua MODIS MMERRA Terra MODIS Aqua MODIS MODIS MeteosatSEVIRI

NOAA-AVHRR

NOAA-AVHRR NOAA-AVHRR NOAA-AVHRR NOAA-AVHRR NOAA-AVHRR Terra MODIS NOAA-AVHRR

Satellite/Sensor

Akhoondzadeh (2014) Lisi et al. (2015)

Singh et al. (2010) Blackett et al. (2011) Rawat et al. (2011) Saradjian and Akhoondzadeh (2011) Qin et al. (2012)

Panda et al. (2007) Saraf et al. (2008) Ma et al. (2010) Pergola et al. (2010)

Saraf and Choudhury (2005a, 2005b, 2005c) Saraf and Choudhury (2005a) Saraf and Choudhury (2005b) Tramutoli et al. (2005) Chen et al. (2006) Choudhury et al. (2006) Ouzounov et al. (2006)

Tronin (2000a) Tronin (2000b) Tramutoli et al. (2001) Tronin et al. (2002) Filizzola et al. (2004) Ouzounov and Freund (2004) Tronin et al. (2004)

Reference

Table 3 A list of significant LST precursory studies using remote sensing and published as complete research articles in peer reviewed journals and conference proceedings. This table was compiled using keyword searches in Scopus and Web of Science. The articles mentioned here were published between 2000 and 26 November 2015.

A. Bhardwaj et al.

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

earthquake-related LST anomalies by studying the M ∼ 5.9 Athens earthquake of 7 September 1999 where they reported a standard deviation of 0.5–5.5 °C in the LST. The premise of their work was based on the use of a geostationary remote sensing platform to reduce the noise (increasing S/N) by accounting for the natural/observational variations of the polar orbit platforms. Importantly, this paper also suggested the existence of pre-seismic TIR anomalies and encouraged further studies in this direction. Similarly, Tramutoli et al. (2005) assessed the RAT’s performance for the M7.8 Izmit earthquake on 17 August 1999, and reported standard deviation values similar to those reported by Filizzola et al. (2004). In addition to AVHRR, they also utilized a geostationary Meteosat platform to mask the observational variations. They presented the natural and observational factors contributing to the noise in the LST anomaly measurements in tabular form. Both Filizzola et al. (2004) and Tramutoli et al. (2005) observed the first occurrence of the precursor within the range given by Tronin (2000a), i.e., ∼1–3 weeks before the earthquake. However, a major limitation of this approach for LST anomaly detection was highlighted by Blackett et al. (2011) when they demonstrated that the anomalous values in the case of the Gujarat (India) 2001 earthquake were actually positive biases resulting from gaps in the MODIS LST dataset due to cloud cover and the subsequent mosaicking of neighboring orbital data.

by Qiang et al. (1999), which was possibly influenced by the Kuroshio Sea’s current fluctuation. Tronin et al. (2002) expanded the observation periods for both study sites, covering a larger number of seismic events with a wider range of magnitudes. Although the extent of the anomaly and the time of the precursor appearance were the same as those reported by Tronin (2000b), Tronin et al. (2002) also established that the thermal anomalies were undetectable for the earthquakes with depths of more than 60 km. Tronin et al. (2004) analyzed a combination of air temperature, LST, and well observation measurements in the Kamchatka peninsula to propose a litho-atmo-ionospheric coupling model to describe related seismic processes. This was the first holistic analysis to incorporate ancillary field data with remote sensing observations. Earlier studies by Tronin and coworkers had focused on relatively flat terrain, which was suitable for remote sensing observations and inferences. However, Tronin et al. (2004) focused on the Kamchatka peninsula in eastern Russia, which is known for its high seismicity, volcanism, geothermal anomalies, complex terrain, and inclement weather conditions, all of which make it unsuitable for remote sensing. The reported time of appearance and amplitude of the anomaly were similar to those suggested by the abovementioned studies and are shown in Table 3 (Qiang et al., 1999; Tronin, 1996, 1999, 2000a, 2000b; Tronin et al., 2002).

2.3. NOAA-AVHRR follow-up works covering several earthquakes 2.2. Pioneer works introducing statistical robustness in LST anomaly measurements

Encouraged by the outcomes of the abovementioned studies, several other researchers have tried to use the AVHRR LST product to observe the LST anomaly before earthquakes at locations across the globe. Saraf and Choudhury (2005a, 2005b, 2005c) investigated the LST anomaly for 6 earthquakes (M6.1-7.9) occurring between March 1990 and December 2003. To assist their analyses on the days with significant cloud cover, they used data from the Special Sensor Microwave Imager (SSM/ I) on board the Defense Meteorological Satellite Program (DMSP). They did not explore the statistical robustness of the anomaly detection mechanism and simply relied on the visual interpretation of the temporal LST maps. However, the importance of the Saraf and Choudhury (2005)’s study is its wider spatial coverage (Afghanistan, Algeria, India, Iran, Pakistan, Turkey). Similar analysis and detailed accounts of the Gujarat (India) and Boumerdes (Algeria) earthquakes can be found in Saraf and Choudhury (2005a, 2005b). Saraf et al. (2008) and others studied earthquakes in Iran using AVHRR data. In this series, Choudhury et al. (2006) investigated the Bam and Zarand earthquakes, with magnitudes of 6.6 and 6.4, respectively. They reported 5–10 °C LST anomalies ∼6 days before the earthquake events. Yet again, the statistical significance of the observed anomalies was not the focus of this research. However, Choudhury et al. (2006) discussed the LST anomalies in relation to the observed air temperature measurements, thus discussing the TIR anomaly in its entirety. Saraf et al. (2008) focused on ten different M5.8-6.6 earthquakes in Iran, which occurred during the period of June 2002–June 2006. Although the anomaly detection was not determined statistically but rather was based on the visual characterization of LST maps, the paper’s largest observed LST anomaly had an amplitude of 13 °C. Rawat et al. (2011) performed similar analysis for the M5.9 Vrancea (Romania) earthquake on 27 October 2004 and the M5.1 Yamnotri (India) earthquake on 22 July 2007. These anomalies were in the range of 5–10 °C appearing ∼1 week before the earthquakes. However, the key point of their research was using outgoing longwave radiation (OLR) anomalies as a proxy for and in association with the LST anomaly. OLR is dependent on temperature and humidity. Thus, any earthquake-induced TIR or humidity anomalies lead to anomalous OLR values. All of these studies used the NOAA-AVHRR data to perform the LST anomaly precursory analysis. The presence of MODIS onboard the National Aeronautics and Space Administration (NASA) Terra satellite provided data with better spatial and spectral resolutions to perform such studies. In recent years, the LST measurements derived from MODIS data have

Tramutoli et al. (2001) tried to solve the problem of separating the natural and observational noise signals (which are due to variations in the time of each revisit, seasons, view angles, air density, and co-registration errors) from the TIR signal anomalies associated with increased seismic activity using a Robust Estimator of TIR Anomalies (called RETIRA by Filizzola et al., 2004). A derivative of this estimator utilizing LST is seen in Eq. (1) (Filizzola et al., 2004). This automatically provides a correction for the effects of atmospheric water vapor content.

αLST(x , y, t ) =

(ΔLST(x , y, t ) − ΔLST(x , y) ) σΔLST(x , y)

(1)

αLST(x,y,t) refers to the LST anomaly on time t at geographical coordinates x and y. ΔLST(x,y,t) is the difference between the current LST at location x, y and its spatial average at a given time t. ‹ΔLST(x,y)› and σΔLST(x,y) are the time average and standard deviation values of ΔLST(x,y,t), respectively. Both, ‹ΔLST(x,y)› and σΔLST(x,y) and σΔLST(x,y) serve as constants and are computed for each location (x,y) using several years of historical satellite-derived LST records acquired during similar observational conditions (around the same day for all years as the day of the earthquake event in the test year). Thus, these constants are represented by two reference images showing the normal LST and its variability at each location for observational conditions as similar as possible to those of the image at hand. RETIRA was an implementation of one of the future applications of the Robust AVHRR Techniques (RAT) suggested by Tramutoli (1998). Tramutoli et al. (2001) examined the effectiveness of RETIRA in finding TIR anomalies in case of the M6.9, Irpinia earthquake on 23 November 1980. Their main contribution was to bring more statistical robustness to the anomaly measurements, in addition to minimizing the natural/ observational noises. They further suggested an important problem for future research, i.e., the need to account for the presence of snow/cloud cover, the variations of water vapor concentrations and the satellite view angles in order to improve the algorithm. Using a completely different approach, they supported the findings of the earthquake-induced LST anomaly studies by Tronin and his coworkers, and suggested a need for future investigations in this area. Filizzola et al. (2004) further assessed the applicability of RETIRA for the investigation of 162

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

dense vegetation cover. Although this aspect of their research is only weakly described and is based on a visual analysis of a single AVHRRderived Normalised Difference Vegetation Index (NDVI) map, we believe that Panda et al. (2007) highlighted a valid concern of LST masking due to the local terrain conditions. In the following years, several more papers tried to effectively utilize MODIS data to observe the LST precursor. Ma et al. (2010) studied the tectonic activities (M7.3-9) in western China between November 2001 and May 2008. Their approach towards studying the LST anomaly was considerably different from those in previous studies. First, they tried to define the various components associated with the LST. Eqs. (4) and (5) represent the components of the satellite-derived LST.

proven to be extremely useful as direct inputs for anomaly models. Subsection 2.4 discusses the studies based on these comparatively new thermal data. 2.4. Advent of MODIS in LST anomaly investigations The MODIS sensor flies onboard the Terra and Aqua satellites, providing nearly complete daily global coverage in 36 spectral bands from 0.415 to 14.235 μm and with varying spatial resolutions of 250 m (2 bands), 500 m (5 bands), and 1000 m (29 bands). Ouzounov and Freund (2004) were among the initial researchers who used the MODIS LST product exclusively when investigating the mid-IR emissions prior to an earthquake. The earthquake they studied was an extremely destructive M7.7 earthquake in Gujarat (India) on 26 January 2001. They reported a thermal anomaly of 3–4 °C, which appeared ∼5 days before the event. It is worth mentioning here that Saraf and Choudhury (2005) used AVHRR data and reported the start of this anomaly 12 days prior to the Gujarat earthquake, with a thermal range of 5–7 °C. The reported magnitudes of the LST anomaly were similar in both studies, but the origin time of the precursor varied. This can be attributed to the entirely different approaches used by these studies. Ouzounov and Freund (2004) used the LST derived from the MODIS bands 31 and 32 with 10.78–12.27 μm bandwidths. Another significant aspect of their research was to provide an additional laboratory-based proof of the concept of p-hole activation (Freund, 2002, 2003, 2007; Freund et al., 2005) as a reason for such an LST anomaly. In a similar series of studies, Ouzounov et al. (2006) performed a more detailed analysis of 6 major earthquakes (M ≥ 6.8) between 1999 and 2003. They utilized multisensor data (Landsat, NOAA-AVHRR, Terra MODIS, Aqua MODIS) to observe the pre-earthquake LST anomaly and used geosynchronous weather satellite images (GOES-10) to better interpret the results. They then tried to establish the spatial relationship between anomalies and the geological faults. They defined the LST anomaly as the difference between the spatial daily root mean square (RMS) values and the mean value in a given area surrounding the epicentre (Eq. (2)). This was another statistical approach to identify the LST anomaly and is different from that proposed by Filizzola et al. (2004). d αLSTdE = LSTRMS − LSTRMS

(4)

ΔT = THigh + TMid + TLow

(5)

Here, TO is the stable component of the LST (TLST), Ta is the annually changing component denoting the normal annual variation field due to the temporal solar activities (and not the tectonic activities), and ΔT represents the LST anomaly associated with the tectonic activities. Short-term variations in ΔT are usually linked to external factors (climatic parameters, atmosphere, sunspots, monsoons, vegetation, human activity) while the long-term variations can be attributed to the Earth’s internal activity (tectonic activity) and thus can serve as an earthquake precursor. THigh, TMid, and TLow are the three constituents of the LST anomaly (ΔT), corresponding to high (0–0.36 years), medium (0.36–1.4 years), and low frequencies (1.4–4 years), respectively. Ma et al. (2010) used wavelet analysis to obtain the frequency components of ΔT. Like many past studies, they observed the LST anomaly fields with respect to the locations of geological faults within the study area, but they did so in a more detailed manner. However, it is worth noting that when considering the coincidence of faults and anomalies, the differences between the surface expression of a fault and the seismic activities deep underground are never considered. The fault usually dips instead of going straight down, potentially causing the focus of a quake to be far from the surface fault. In these cases, any geological coincidence becomes less likely. In that year and those following it, two more studies investigated one of the most destructive and most studied earthquakes in recent decades, the M7.6 Gujarat earthquake on 26 January, 2001. Singh et al. (2010) studied the LST anomaly in relation to the Measurements of Pollution in the Troposphere (MOPITT)-derived CO emissions prior to this earthquake, which indicated a strong litho-atmo-ionospheric coupling. Blackett et al. (2011)’s observations for the same earthquake, however, produced surprising results that were contrary to all of the previous studies. They stated that there was a lack of any credible evidence of an anomalous LST precursor and further recommended care when identifying seismic thermal anomalies. They analysed six continuous years (2001–2006) of daily MODIS LST data for the existence of LST anomalies using the predefined analytical techniques: (1) the LST difference between the year of earthquake and any other year without an earthquake (Ouzounov and Freund, 2004), (2) the LST difference between each year (2001–2006) and the mean LST for all the years, and (3) the RAT modified for MODIS (Pergola et al., 2010). Using these methods, Blackett et al. (2011) reported an anomaly in the earthquake year; they also showed that the anomaly fell within the inter-annual anomaly envelope. Their conclusive remarks were (1) identifying potential thermal anomalies requires ‘baseline’ conditions as a prerequisite: the larger the number of years used in the analysis, the greater the robustness of the approach; (2) data gaps owing to cloud cover and swathrelated issues significantly affect anomaly detection; and (3) there is an absence of any robust evidence for the existence of an LST anomaly prior to the 2001 Gujarat earthquake, which suggests the necessity of a cautious approach to LST identification until further statistically robust proofs are reported. These recommendations made by Blackett et al. (2011) seem to be valid for their experiment (i.e., for cases involving a

(2)

Here, αLSTEd refers to the LST anomaly on a given day d in the year of earthquake E without incorporating inter-annual variability. LSTdRMS is the RMS value of LST on day d and LST RMS is the mean RMS value of LST for the entire time interval of analysis within a particular year. The next step in this process is to account for the inter-annual variability in the LST. This can be calculated using Eq. (3).

αLSTdIA = αLSTdE − αLSTdP

TLST = TO + Ta + ΔT

(3)

αLSTIA d

Here, refers to the LST anomaly on a given day d in the year of earthquake E after incorporating the inter-annual variability. αLSTPd refers to the LST anomaly on the same day d in the previous year(s) P without incorporating the inter-annual variability. Both, αLSTEd and αLSTPd are calculated for their respective years using Eq. (2). Using this analysis, Ouzounov et al. (2006) reported the first occurrence of the LST anomaly within the week prior to the earthquakes and within a range of ± 4 °C. Panda et al. (2007) reported a 4–8 °C anomaly 7 days before the M7.6 Muzaffarabad earthquake in 2005. Their interpretations were visual rather than statistical, and they correlated the LST observations with the air temperature data from two meteorological stations (Islamabad and Srinagar). The lack of statistical robustness in this research and several others (Choudhury et al., 2006; Saraf et al., 2008) makes them vulnerable to scientific scrutiny and requires the use of a statistical approach to reinvestigate these earthquakes to validate the reported anomalies. However, the point of the Panda et al. (2007) study was to observe the masking of the LST anomaly due to the presence of 163

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

is refined using the adjacent pixels in the third and fifth channels, in order to improve the consistency between the spatial LST distributions over lakes and the surrounding land. As a convention, the adjacent clear-sky pixels at a 99% confidence (66% or higher for inland water pixels) in the third and fifth channels prompt the pixels in fourth channel to be treated as clear-sky pixels for the LST retrieval (Wan, 2006). However, this entire procedure is tricky, and the estimated error values for the LST product are given only for clear-sky conditions. The effect of cloud contamination is not considered in the error estimation. Clouds play a vital role in controlling the radiative balance of the Earth and therefore require precise estimations in remote sensing products. The estimation of cloudiness in the case of high, thin cirrus clouds is even more challenging and the MODIS cloud mask product has been reported to underestimate cirrus cloud cover (Di Rosa et al., 2009). This discussion proves that the utmost care is needed in interpreting the results of anomaly detection algorithms, as errors in cloud cover estimations are capable of creating anomalous results.

multiyear approach and accounting for data gaps), although they disagree with most of the published literature. Saradjian and Akhoondzadeh (2011) studied three earthquakes of M > 6 in Iran and reported the appearance of anomalies 1–20 days before the earthquakes, within the range of 3–12 °C. The highlight of their research was the techniques (interquartile, wavelet transform, and Kalman filter) they used for the LST anomaly detection. The interquartile method helped to construct the higher and lower limits (envelope) for the LST data; the values outside of this envelope were called anomalous. The wavelet transform was sensitive to sudden changes in the LST. The Kalman filter method performed best in cases with the most extreme and erratic variations of the LST. While exploring these new anomaly detection techniques, Akhoondzadeh (2014) further studied the M7.7 Saravan earthquake on 16 April, 2013, using an integrative approach involving an Artificial Neural Network (ANN) and a Particle Swarm Optimization (PSO). PSO algorithms have been reported to solve a wide range of optimization problems. ANN models, although efficient, may become trapped in local minima and need backwards propagation algorithms to be trained. Akhoondzadeh (2014) used a PSO to overcome the ANN stagnation, and this combination of ANN and PSO methods for anomaly detection was highly efficient. Akhoondzadeh (2014) was also able to resolve the ambiguity between the ionospheric TEC anomalies induced by seismic activity and those fluctuations due to solar activity using a wavelet transform. Akhoondzadeh (2014) compared these results with several statistical methods, such as the mean (e.g., Blackett et al., 2011; Qin et al., 2012), median, wavelet (e.g., Akhoondzadeh, 2012; Saradjian and Akhoondzadeh, 2011), Kalman filter (e.g., Akhoondzadeh, 2012; Saradjian and Akhoondzadeh, 2011), Auto-Regressive Integrated Moving Average (ARIMA), Support Vector Machine (SVM) (e.g., Akhoondzadeh, 2013a, 2013b), and Genetic Algorithm (GA) (e.g., Akhoondzadeh, 2013a) methods. The study reported that the PSOtrained ANN was another appropriate method to identify the anomaly in an LST time-series. Qin et al. (2012) used a completely different dataset to study the LST anomaly: a Modern Era Retrospective-Analysis for Research and Applications (MERRA) reanalysis dataset, which uses MODIS observations extensively. They studied two Italian earthquakes of M ∼ 6 occurring in May 2012 and reported the appearance of a 6 °C LST anomaly ∼3 weeks before the earthquakes. They relied on three statistical parameters for the inter-annual LST comparisons: the mean (μ), standard deviation (σ), and maximum. To account for the contribution of climatic factors towards the thermal anomaly, they selected the earthquake-related anomaly candidates with temperature values beyond μ+1.5σ or maxima. The obvious advantages of using reanalysis data were that the authors were able to increase the observational period to include the past 33 years (1979–2011) and had an hourly temporal observation resolution. However, one aspect that was missing here was that Qin et al. (2012) did not consider the occurrence of similar tectonic activities in the study area during the observational period throughout the 33 years. Using the USGS database (http:// earthquake.usgs.gov/earthquakes/search/), we identified several occurrences of such events in the study years. Leaving out the years with increased seismic activities would have improved the robustness of this analysis. Blackett et al. (2011) provided a new perspective on MODIS-based anomaly detection when they demonstrated that the anomalous values in the case of the 2001 Gujarat (India) earthquake were actually the positive biases resulting from gaps in the MODIS LST data due to cloud cover and the subsequent mosaicking of neighbouring orbital data. This conclusion also prompts the need for an assessment of the accuracy or reliability of the LST products in the case of cloud cover. The MODIS LST product is generated using the MODIS cloud mask product (MOD35_L2) that utilizes band 22 (4–11 μm) to determine the cloudiness of a pixel (Wan, 2006). The fourth channel in band 22 is noisy and tends to overestimate the cloudiness. Therefore, the cloud mask product

2.5. Effectiveness of statistical approaches While discussing these anomalous signals, a separate consideration of the significance and effectiveness of the adopted statistical/mathematical approaches in detecting the actual anomalies is needed. One major challenge is to distinguish between the natural variations of local climatic variability (temperature, solar flux, humidity, cloudiness) and the genuine anomalous precursory LST values. The RETIRA given in Eq. (1) is a modified version of the RAT approach (Tramutoli, 1998) and appears to be the most widely used method in the literature for LST anomaly detection. This method relies on spatial and temporal averages, as well as standard deviations of LST values, and has been found to negate the presence of highly variable contributions from atmosphere (water vapour, transmittance), land (emissivity, topography), and observational conditions (solar and satellite zenith angles) (Tramutoli et al., 2001). Tramutoli et al. (2001) also mentioned the intrinsic exportability of this method to different geographical settings and remote sensors. A look at Eq. (1) clearly highlights two extremely important considerations when using RETIRA: (1) the extents of spatiotemporal boundaries and (2) the addition of as many annual records as possible to increase the comparable time-series for anomaly isolation. The dependence of RETIRA on its spatiotemporal domain makes it vulnerable to subjectivity. The RETIRA results can show wide variations in their amplitudes and frequencies, depending on the defined spatial domain and the number of image pixels within that domain. Similarly, the observational window for anomaly detection in the temporal domain is also prone to bias. If this observational window is increased to several months, it can affect the anomaly detection during seasonal transitions. Similarly, too small of an observational window can miss any anomalous signals which occur weeks before an earthquake. There is not a conclusive convention for deciding these spatiotemporal boundaries, which compromises the robustness of RETIRA. The work of several researchers (e.g., Blackett et al., 2011; Bhardwaj et al., 2017) involving multiple years of data in their analyses is also a significant step towards distinguishing the anomalous LST values from the background values. This elongation of multiyear time-series is equally valid for other anomaly detection approaches which rely on inter-annual variability, such as the one given by Ouzounov et al. (2006) in Eqs. (2) and (3). Eqs. (2) and (3) also highlight the use of the RMS values, instead of the means and standard deviations, to detect the anomalies separately on intra-annual and inter-annual scales. A notable point here is that, statistically, an average measures the central tendency of a given dataset, while the RMS depends on the sinusoidal nature of the observations. However, the reported anomalies in the literature presented within this review display extremely random natures and do not necessarily have sinusoidal distributions.

164

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

Alvan, H.V., Mansor, S., Omar, H., Azad, F.H., 2014. Precursory signals associated with the 2010 M8.8 Bio-Bio earthquake (Chile) and the 2010 M7.2 Baja California earthquake (Mexico). Arabian J. Geosci. 7 (11), 4889–4897. Balasis, G., Mandea, M., 2007. Can electromagnetic disturbances related to the recent great earthquakes be detected by satellite magnetometers? Tectonophysics 431, 173–195. Bhardwaj, A., Singh, S., Sam, L., Bhardwaj, A., Martin-Torres, F.J., Singh, A., Kumar, R., 2017. MODIS-based estimates of strong snow surface temperature anomaly related to high altitude earthquakes of 2015. Remote Sens. Environ. 188, 1–8. Blackett, M., Wooster, M.J., Malamud, B.D., 2011. Exploring land surface temperature earthquake precursors: a focus on the Gujarat (India) earthquake of 2001. Geophys. Res. Lett. 38, L15303. Borghi, A., Aoudia, A., Riva, R.E.M., Barzaghi, R., 2009. GPS monitoring and earthquake prediction: a success story towards a useful integration. Tectonophysics 465, 177–189. Chen, S.Y., Ma, J., Liu, L.Q., Liu, P., 2006. An enhancive phenomenon of thermal infrared radiation prior to Pakistan earthquake. Prog. Nat. Sci. 16 (11), 1487–1490. Choudhury, S., Dasgupta, S., Saraf, A.K., Panda, S., 2006. Remote sensing observations of pre-earthquake thermal anomalies in Iran. Int. J. Remote Sens. 27 (20), 4381–4396. Cicerone, R.D., Ebel, J.E., Britton, J., 2009. A systematic compilation of earthquake precursors. Tectonophysics 476, 371–396. Console, R., Pantosti, D., D’Addezio, G., 2002. Probabilistic approach to earthquake prediction. Ann. Geophys. 45 (6), 723–731. Daneshvar, M.R.M., Freund, F.T., 2016. Remote sensing of atmospheric and ionospheric signals prior to the Mw 8.3 Illapel earthquake, Chile 2015. Pure Appl. Geophys. 174 (1), 11–45. Daneshvar, M.R.M., Tavousi, T., Khosravi, M., 2014. Synoptic detection of the short-term atmospheric precursors prior to a major earthquake in the Middle East, North Saravan M 7.8 earthquake, SE Iran. Air Qual. Atmos. Health 7 (1), 29–39. Di Rosa, D., Notarnicola, C., Posa, F., 2009. Cross-comparison and validation of MODIS aqua cloud mask by using CLOUDSAT and CALIPSO datasets. Geoscience and Remote Sensing Symposium, 2009 IEEE International, IGARSS 2009 Vol. 3 pp. III-1019. Dobrovolsky, I.P., Gershenzon, N.I., Gokhberg, M.B., 1989. Theory of electrokinetic effects occurring at the final stage in the preparation of a tectonic earthquake. Phys. Earth Planet. Inter. 57, 144–156. Filizzola, C., Pergola, N., Pietrapertosa, C., Tramutoli, V., 2004. Robust satellite techniques for seismically active areas monitoring: a sensitivity analysis on September 7, 1999 Athens’s earthquake. Phys. Chem. Earth. 29 (4), 517–527. Freund, F., Keefner, J., Mellon, J.J., Post, R., Takeuchi, A., Lau, B.W.S., La, A., Ouzounov, D., 2005. Enhanced mid-infrared emission from igneous rocks under stress. Geophys. Res. Abstr. 7, 09568. Freund, F., 2002. Charge generation and propagation in igneous rocks. J. Geodyn. 33, 543–570. Freund, F., 2003. Rocks that crackle and sparkle and glow: strange pre-earthquake phenomena. J. Sci. Explor. 17 (1), 37–71. Freund, F., 2007. Pre-earthquake signals—part II: flow of battery currents in the crust. Nat. Hazards Earth Syst. Sci. 7, 543–548. Freund, F., 2011. Pre-earthquake signals: underlying physical processes. J. Asian Earth Sci. 41, 383–400. Freund, F.T., 2013. Earthquake forewarning-a multidisciplinary challenge from the ground up to space. Acta Geophys. 61 (4), 775–807. Ganguly, N.D., 2016. Atmospheric changes observed during April 2015 Nepal earthquake. J. Atmos. Solar Terrestrial Phys. 140, 16–22. Geller, R.J., Jackson, D.D., Kagan, Y.Y., Mulargia, F., 1997. Enhanced: earthquakes cannot be predicted. Science 275 (5306), 1616–1620. Guo, G., Wang, B., 2008. Cloud anomaly before Iran earthquake. Int. J. Remote Sens. 29 (7), 1921–1928. Hayakawa, M., Molchanov, O.A., Kodama, O.A., Tanaka, T., Igarashi, T., 2000. On a possibility to monitor seismic activity using satellites. Adv. Space Res. 26 (6), 993–996. Holzer, T.L., Savage, J.C., 2013. Global earthquake fatalities and population. Earthq. Spectra 29 (1), 155–175. Ihmle, P.F., Jordan, T.H., 1994. Teleseismic search for slow precursors to large earthquakes. Science 266, 1547–1551. Ishibashi, K., 1988. Two categories of earthquake precursors, physical and tectonic, and their roles in intermediate-term earthquake prediction. Pure Appl. Geophys. 126 (2), 687–700. Lisi, M., Filizzola, C., Genzano, N., Paciello, R., Pergola, N., Tramutoli, V., 2015. Reducing atmospheric noise in RST analysis of TIR satellite radiances for earthquakes prone areas satellite monitoring. Phys. Chem. Earth. 85 (-86), 87–97. Logan, J.M., 1977. Animal behavior and earthquake prediction. Nature 265, 404–405. Lomnitz, C., 1994. Fundamentals of Earthquake Prediction. John Wiley & Sons, New York, pp. 326. Ma, J., Chen, S., Hu, X., Liu, P., Liu, L., 2010. Spatial-temporal variation of the land surface temperature field and present-day tectonic activity. Geosci. Front. 1 (1), 57–67. Marano, K.D., Wald, D.J., Allen, T.I., 2010. Global earthquake casualties due to secondary effects: a quantitative analysis for improving rapid loss analyses. Nat. Hazards 52, 319–328. Martinelli, G., 2000. Contributions to a history of earthquake prediction research. Seism. Res. Lett. 71, 583–603. Ouzounov, D., Freund, F., 2004. Mid-infrared emission prior to strong earthquakes analyzed by remote sensing data. Adv. Space Res. 33, 268–273. Ouzounov, D., Bryant, N., Logan, T., Pulinets, S., Taylor, P., 2006. Satellite thermal IR phenomena associated with some of the major earthquakes in 1999–2003. Phys. Chem. Earth 31 (4), 154–163.

3. Conclusion The amplitudes of the observed anomalies reach a maximum of up to 13 °C. However, there is no definite correlation between the amplitude of an anomaly and the magnitude of the associated earthquake. The average initial appearance of these signals varies from ∼1 to 3 weeks prior to an earthquake. Although the statistical and methodological robustness displayed by most of these papers (Table 3) indicate our ability to definitively detect anomalies, the use of two different analysis techniques for the same earthquake often leads to differences in the amplitudes and appearance times detected. Thus, the timing of the precursor’s appearance and the amplitudes of the measured anomalous signals vary widely, making it difficult to consider them universally robust reasons for issuing earthquake warnings. Another aspect missing from the reviews of this topic is a statement on any visible relationship between the different geological settings and their corresponding pre-earthquake LST anomalies. Here, we tried to find a pattern between the LST anomalies and their geological settings, but we could not observe any definite correlation. Even in the same geological settings, the anomalies can show different values, and in different geological settings, they can display similar values (Table 3). Based on the randomness in the observations of these precursors, we support employing a global-scale monitoring system to detect statistically robust anomalous geophysical signals prior to an earthquake before considering them to be a definite earthquake precursor. Here, we focused on the LST anomaly as a precursor. However, we conclude that a holistic approach, where researchers from different research backgrounds explore several of the other precursors mentioned in Table 1 in conjunction with each other, is needed to establish a workflow for searching for a possibility to enhance the predictability of major earthquakes. The planned future remote sensing missions can provide surface and atmospheric remote sensing data in better spatial, spectral, and radiometric resolutions, and we can hope that the next decades will see some advancements in observing these signals more clearly. For example, Landsat 9 will be equipped with the Thermal Infrared Sensor 2 (TIRS-2) that will provide the LST and will act as an upgraded version of the Landsat 8 TIRS instrument by solving the known stray light and reliability issues with the TIRS. The Sentinel-4 mission with its geostationary orbit can provide extensive and continuous trace gas concentrations and aerosols measurements in the atmosphere and can be helpful in approving or disapproving the atmospheric precursor hypotheses. The similar atmospheric observations will be provided by Sentinel-5 and Sentinel-5P with a sun-synchronous orbit. Nevertheless, the validity of earthquake precursor research is still debatable, as the presence of such signals is encouraging but the arbitrariness of the observations is dispiriting and does not support the possibility of earthquake prediction in the near future. Acknowledgements Authors are thankful to the anonymous reviewers and the editorial board for providing valuable comments to enhance the quality of the manuscript. PKJ is thankful to DST-PURSE of Jawaharlal Nehru University for research support. LS and SS acknowledge German Academic Exchange Service (DAAD) for their PhD scholarships. References Akhoondzadeh, M., 2012. Anomalous TEC variations associated with the powerful Tohoku earthquake of 11 March 2011. Nat. Hazards Earth Syst. Sci. 12, 1453–1462. Akhoondzadeh, M., 2013a. Genetic algorithm for TEC seismo-ionospheric anomalies detection around the time of the Solomon (Mw = 8.0) earthquake of 06 February 2013. Adv. Space Res. 52, 581–590. Akhoondzadeh, M., 2013b. Support vector machines for TEC seismoionospheric anomalies detection. Ann. Geophys. 31, 173–186. Akhoondzadeh, M., 2014. Thermal and TEC anomalies detection using an intelligent hybrid system around the time of the Saravan, Iran, (Mw = 7.7) earthquake of 16 April 2013. Adv. Space Res. 53, 647–655.

165

Int J Appl  Earth Obs Geoinformation 63 (2017) 158–166

A. Bhardwaj et al.

Saraf, A.K., Rawat, V., Banerjee, P., Choudhury, S., Panda, S.K., Dasgupta, S., Das, J.D., 2008. Satellite detection of earthquake thermal precursors in Iran. Nat. Hazards 47, 119–135. Saraf, A.K., Rawat, V., Choudhury, S., Dasgupta, S., Das, J., 2009. Advances in understanding of the mechanism for generation of earthquake thermal precursors detected by satellites. Int. J. Appl. Earth Obs. Geoinf. 11, 373–379. Singh, R.P., Kumar, J.S., Zlotnicki, J., Kafatos, M., 2010. Satellite detection of carbon monoxide emission prior to the Gujarat earthquake of 26 January 2001. Appl. Geochem. 25 (4), 580–585. Surkov, V.V., Pokhotelov, O.A., Parrot, M., Hayakawa, M., 2006. On the origin of stable IR anomalies detected by satellites above seismo-active regions. Phys. Chem. Earth. 31, 164–171. Tramutoli, V., Di Bello, G., Pergola, N., Piscitelli, S., 2001. Robust satellite techniques for remote sensing of seismically active areas. Annali Di Geofisica 44 (2), 295–312. Tramutoli, V., Cuomob, V., Filizzolab, C., Pergolab, N., Pietrapertosa, C., 2005. Assessing the potential of thermal infrared satellite surveys for monitoring seismically active areas: the case of Kocaeli (Izmit) earthquake, August 17, 1999. Remote Sens. Environ. 96, 409–426. Tramutoli, V., 1998. Robust AVHRR Techniques (RAT) for environmental monitoring: theory and applications. In EUROPTO conference on remote sensing for geology, land management, and cultural heritage III, Barcelona, Spain. Proc. SPIE 3496, 101–113. Tronin, A.A., Hayakawa, M., Molchanov, O.A., 2002. Thermal IR satellite data application for earthquake research in Japan and China. J. Geodyn. 33, 519–534. Tronin, A.A., Biagi, P.F., Molchanov, O.A., Khatkevich, Y.M., Gordeev, E.I., 2004. Temperature variations related to earthquakes from simultaneous observation at the ground stations and by satellites in Kamchatka area. Phys. Chem. Earth. 29 (4), 501–506. Tronin, A.A., 1996. Satellite thermal survey—a new tool for the study of seismoactive regions. Int. J. Remote Sens. 17, 1439–1455. Tronin, A.A., 1999. Satellite thermal survey application for earthquake prediction. In: Hayakawa, M. (Ed.), Atmospheric and Ionospheric Electromagnetic Phenomena Associated with Earthquakes. Terrapub, Tokyo, pp. 717–746. Tronin, A.A., 2000a. Thermal IR satellite sensor data application for earthquake research in China. Int. J. Remote Sens. 21 (16), 3169–3177. Tronin, A.A., 2000b. Thermal satellite data for earthquake research. Geoscience and Remote Sensing Symposium, 2000. Proceedings. IGARSS 2000. IEEE 2000 International 6 2703–2705. Tronin, A.A., 2006. Remote sensing and earthquakes: a review. Phys. Chem. Earth 31, 138–142. Verma, M., Bansal, B.K., 2012. Earthquake precursory studies in India: scenario and future perspectives. J. Asian Earth Sci. 54–55, 1–8. Wan, Z., 2006. MODIS Land Surface Temperature Products Users’ Guide. Institute for Computational Earth System Science, University of California, Santa Barbara, CA. Evaluation of Proposed Earthquake Precursors (Conducted 1989/1990 by the IASPEI Sub Commission on Earthquake Prediction). In: Wyss, M. (Ed.), American Geophysical Publication pp. 94. Wyss, M., 1997. Cannot earthquakes be predicted. Science 278 (5337), 487–490. Yao, Q.L., 2007. Field effect and regional conversion as the mechanism of natural hazard chains. Meteorol. Disaster Reduct. Res. 30 (3), 31–36.

Panda, S.K., Choudhury, S., Saraf, A.K., Das, J.D., 2007. MODIS land surface temperature data detects thermal anomaly preceding 8 October 2005 Kashmir earthquake. Int. J. Remote Sens. 28 (20), 4587–4596. Pergola, N., Aliano, C., Coviello, I., Filizzola, C., Genzano, N., Lacava, T., Lisi, M., Mazzeo, G., Tramutoli, V., 2010. Using RST approach and EOS-MODIS radiances for monitoring seismically active regions: a study on the 6 April 2009 Abruzzo earthquake. Nat. Hazards Earth Syst. Sci. 10 (2), 239–249. Petrini, R., Italiano, F., Riggio, A., Slejko, F.F., Santulin, M., Buccianti, A., Bonfanti, P., Slejko, D., 2012. Coupling geochemical and geophysical signatures to constrain strain changes along thrust faults. Bollettino di Geofisica Teorica ed Applicata 53 (1), 113–134. Piroddi, L., Ranieri, G., 2012. Night thermal gradient: a new potential tool for earthquake precursors studies. An application to the seismic area of L’Aquila (Central Italy). Sel. Top. Appl. Earth Observations Remote Sens. 5 (1), 307–312. Piroddi, L., Ranieri, G., Freund, F., Trogu, A., 2014. Geology, tectonics and topography underlined by L'Aquila earthquake TIR precursors. Geophys. J. Int. 197 (3), 1532–1536. Pisa, D., Parrot, M., Santolik, O., 2011. Ionospheric density variations recorded before the Mw 8. 8 earthquake in Chile. J. Geophys. Res.: Space Phys. 116 (A08309). http://dx. doi.org/10.1029/2011JA016611. Pulinets, S.A., Ouzounov, D., Karelin, A.V., Boyarchuk, K.A., Pokhmelnykh, L.A., 2006. The physical nature of thermal anomalies observed before strong earthquakes. Phys. Chem. Earth. 31, 143–153. Pulinets, S.A., 2004. Ionospheric precursors of earthquakes: recent advances in theory and practical applications Terrestrial. Atmos. Oceanic Sci. 15 (3), 413–435. Qiang, Z., Xiu-Deng, X., Chang-Gong, D., 1991. Thermal infrared anomaly—precursor of impending earthquakes. Chin. Sci. Bull. 36 (4), 319–323. Qiang, Z.J., Dian, C.G., Li, L.Z., 1999. Satellite thermal infrared precursors of two moderate-strong earthquakes in Japan and impending earthquake prediction. In: Hayakawa, M. (Ed.), Atmospheric and Ionospheric Electromagnetic Phenomena Associated with Earthquakes. TERRAPUB, Tokyo, pp. 747–750. Qin, K., Wu, L.X., De Santis, A., Cianchini, G., 2012. Preliminary analysis of surface temperature anomalies that preceded the two major Emilia 2012 earthquakes (Italy). Ann. Geophys. 55 (4), 823–828. Rawat, V., Saraf, A.K., Das, J., Sharma, K., Shujat, Y., 2011. Anomalous land surface temperature and outgoing long-wave radiation observations prior to earthquakes in India and Romania. Nat. Hazards 59 (1), 33–46. Roeloffs, E.A., 1988. Hydrologic precursors to earthquakes: a review. Pure Appl. Geophys. 126, 177–209. Saradjian, M.R., Akhoondzadeh, M., 2011. Thermal anomalies detection before strong earthquakes (M > 6.0) using interquartile, wavelet and Kalman filter methods. Nat. Hazards Earth Syst. Sci. 11 (4), 1099–1108. Saraf, A.K., Choudhury, S., 2005a. NOAA-AVHRR detects thermal anomaly associated with 26 January, 2001 Bhuj Earthquake, Gujarat, India. Int. J. Remote Sens. 26 (6), 1065–1073. Saraf, A.K., Choudhury, S., 2005b. Satellite detects surface thermal anomalies associated with the Algerian earthquakes of May 2003. Int. J. Remote Sens. 26 (13), 2705–2713. Saraf, A.K., Choudhury, S., 2005c. Thermal remote sensing technique in the study of preearthquake thermal anomalies. J. Ind. Geophys. Union 9 (3), 197–207.

166