1 Introduction

Earthquakes occur frequently in the countries surrounding the Bay of Bengal. The northern part of the Bay of Bengal, particularly the Andaman region (7°–22°N and 88°–100°E), experienced 348 earthquakes between 1900 and 1980 (Bapat 1982). The Sunda Subduction Zone (SSZ), an elongated zone from Java to the Myanmar coast, is capable of triggering large tsunamigenic earthquakes (Satake and Atwater 2007). The 2004 Indian Ocean earthquake and tsunami (2004 IOT) has already challenged the perceived level of regional risk, rupturing around 1300 km of seafloor from Sumatra to the Andaman Island (Satake and Atwater 2007; Satake et al. 2006). The Bay of Bengal region has experienced several large historic earthquakes including the 19 August 1868 (Newcomb and McCann 1987), the 31 December 1881 (Ortiz and Bilham 2003) and the 26 June 1941 events all of which originated within the Andaman Trench (Fig. 1). Notably, the 2015 Nepal earthquake originating near the main frontal thrust between the subducting India plate and the overriding Eurasia plate to the north caused approximately 8200 deaths in Nepal, India, China and Bangladesh (Fan and Shearer 2015). Thus, it is a region where large earthquakes can occur. It is difficult to gain a sophisticated understanding of the regional earthquake hazard because there is a fragmented historical record of past events. In order to quantify the hazard, it is necessary to have fundamental data about the occurrence, magnitude, impact and effects of past earthquakes. This research seeks to extend our understanding of the regional earthquake hazard by establishing a new catalogue of past earthquakes over as a long a time period as possible.

Fig. 1
figure 1

The distribution of 562 earthquakes between 810BC and 2012, showing four clusters (shown by black dashed lines) of earthquake: west coast of Sumatra, Andaman and Nicobar Island, Bangladesh and adjoining regions and NW coast of India. Key features include the AT = Andaman Trench and SSZ = Sunda Subduction Zone

Earthquake hazard assessment may be enhanced by reviewing written and geological records of past events (Ambraseys 1971; Dominey-Howes 2002; Heidarzadeh et al. 2008). The historical record of earthquakes is one of the most important components of the seismic risk analysis of an area (Ambraseys 1980; Glade et al. 2001) because significant evidence of earthquakes and fault information has not been discovered and collected in many parts of the world (Ambraseys 1980). However, reports of earthquakes between BC800 and 1900 occurring in India have very limited information associated with them (e.g. just epicentre location and magnitude) that restricts their usefulness to stakeholders (Martin and Szeliga 2010).

For example, Bangladesh developed its policies to mitigate the effects of earthquake based on events occurring only after 1885, neglecting the potential for large earthquakes from adjacent active faults in NE India, Myanmar and the Bay of Bengal region (Choudhury 2005). Therefore, more detailed archival research is needed to understand the earthquake hazard of the northern Bay of Bengal and adjacent regions (Fig. 1). Interestingly, Satake and Atwater (2007) noted that for the Pacific Ocean region, there was a misalignment between the size of earthquakes actually observed and the size of those inferred from geological records with the former being smaller. Consequently, it is reasonable to assume that the earthquake (and tsunami risk) in the northern Bay of Bengal region is similarly higher than that would be expected from existing records of past events. To understand natural hazards, it is very important to use long-term historical hazard data and geological records that are longer than short-term instrumental data (Ambraseys 1980; Nott 2003). If this is not done in the northern Bay of Bengal region, governmental authorities may significantly underestimate risk. This has important implications for disaster risk reduction (DRR) policy and practice.

Regional earthquake catalogues exist for many areas of the world including Greece (Papazachos and Papazachou 1997); Turkey (Ambraseys and Finkel 1995); Thailand (Nutalaya et al. 1985); Indonesia (Soetardjo et al. 1985); India (Bapat et al. 1983; Chandra 1977; Oldham 1883; Rao and Rao 1984); and China (Lee et al. 1976). From an analysis of the available documents listed in Table 1, it is apparent that in most situations, the authors do not provide baseline data about each earthquake event they list. Instead, they only observe the magnitude and date of occurrence where that information is available. In most situations, only one source record is cited for each unique earthquake event. Further, they only cite sources derived from the modern research literature.

Table 1 List of formally published data sets of earthquake events documented as occurring in the NE Indian Ocean and Bay of Bengal region

The earthquake catalogue of Milne (1911) includes earthquakes affecting Bangladesh, India and Myanmar from AD7 to 1899. Oldham (1883) provided details of earthquakes in India and Bangladesh for the period between the years 893 and 1869. Earthquake catalogues that focused on Southeast Asia (Nutalaya et al. 1985) and the SSZ (Soetardjo et al. 1985) report the extent of earthquake effects as far afield as India and Bangladesh. In the last 50 years, several authors (Bapat et al. 1983; Chandra 1977; Iyengar et al. 1999; Rao and Rao 1984; Singh and Shankar 1992) have attempted to collect, collate and analyse earthquake data from India. Ambraseys (Ambraseys 2000, 2004; Ambraseys and Bilham 2003; Ambraseys and Douglas 2004) and Bilham (Bilham 2004; Bilham and England 2001) have analysed earthquake hazards in India and Bangladesh. Akhter (2010) has specifically analysed historical and instrumentally recorded earthquakes that affected Dhaka. However, to the best of our knowledge, no attempt has been made to synthesise all available data sources and records of past earthquake events in order to explore the earthquake hazard in Bangladesh. In this paper, we undertake this task.

2 Methods

2.1 Process of new earthquake catalogue development

The new earthquake catalogue shown in Table 2 utilised global online earthquake databases (e.g. the National Geophysical Data Center (NGDC 2012) and Geoscience Australia (GA), (Geoscience Australia, 2011), the International Seismological Centre (ISC 1964–1990), the Incorporated Research Institutions for Seismology (IRIS 2011), regional catalogues (i.e. Oldham 1883), annals, chronicles, diaries, letters, travellers and their accounts, previously published earthquake catalogues (Table 1), journal articles in the peer-reviewed literature, books, reports, newspaper articles, local historical books, institutional and administrative memoirs and the historical archive from the India Office Records of the British Library and Royal Society, London.

Table 2 A catalogue of earthquakes between 810BC and 2012 for the Bay of Bengal

Global earthquake databases (i.e. NGDC, NTL, GA, ISC, IRIS) incorporate data from both historical and instrumental records of earthquake. While we focus on identifying events that have affected the northern Bay of Bengal and Bangladesh, we have broadened the study to include the area located between latitude 15°S to 28°N (i.e. Sunda Subduction Zone) and longitude 70°E to 106°E (i.e. Kutch–Saurashtra) to ensure the comprehensive coverage of the study area and adjacent seismically active regions (Fig. 1). Consequently, the study area includes Bangladesh, India, Myanmar, Sri Lanka, the Andaman and Nicobar Islands, Thailand, Malaysia and NW Indonesia.

A portion of China lies within the above-mentioned latitude and longitude boundaries, but we chose not to include earthquake events whose epicentres were located in China. The reason for this is that the emphasis of this work is on earthquakes occurring along the coastline of the NE Indian Ocean, Bangladesh and adjacent regions. The selection of the study region is mainly dependent on three factors: (1) the lead researcher has personal research interests in Bangladesh and adjacent regions, (2) no up to date earthquake catalogue is available for Bangladesh and adjacent regions, and (3) there is a need to understand earthquake hazards and associated risks in Bangladesh by considering all earthquakes from adjacent and distant active seismic sources. The locations of earthquakes with epicentres within 300 km of the coast were included as well as all data from marine sources (i.e. the Bay of Bengal and the NE Indian Ocean). Setting these boundaries ensured that all earthquake events occurring in Bangladesh and adjacent regions (i.e. the Dauki Fault and the Sylhet Fault) were included in the catalogue.

In developing the catalogue, we took great care to identify all records of past earthquakes as far back in time as possible. This is primarily because limited historical work on pre-seventeenth century earthquakes has been completed and because such data can provide useful insights into longer-term hazard useful for stakeholders. Since we emphasise recording historic and prehistoric events, we have documented all felt intensity earthquake data available from archival sources. Detailed data sets from instrumentally recorded periods were readily accessible via NGDC, NTL, GA, ISC, IRIS sources. However, to construct a geography of earthquake hazards in Bangladesh, we have selected an earthquake magnitude above 4 as the inclusion criteria for the instrumental period. The reason for selecting magnitude 4 as below level of threshold for earthquake events inclusion in the catalogue is that the people may feel and concern from this level earthquake event. Subsequently, the government of Bangladesh may take mitigate measures reducing the effects of earthquakes. Geological (Rajendran et al. 2004) and archaeological records of past earthquakes (Banerji 1923; Chaudhury 1964) are limited, but, where available, have been incorporated in the new catalogue because previously, such sources were not considered (Table 1).

2.2 Documentary sources

‘Inscribed text’ is the central feature of artefacts we call ‘documents’ (Scott 1990). There are also non-text-based documents that include photographs, film, video, television, displays, drawings and pictures, and graphical presentations (Mason 2002). Individuals, groups and organisations routinely produce documents throughout everyday life. While these documents are not produced for research purposes, they nonetheless provide evidence of events and possessions in the lived world (Payne and Payne 2004).

Historical archival documents include but are not limited to myths, inscriptions, annals, chronicles, diaries, letters, epistolaries, travellers and their accounts, antiquarian historiography, local histories, archive of institutional and administrative memories, newspapers, historical earthquake cartography, iconographic sources (paintings, drawings, relief carvings and other artistic depictions), photographs and films and ethnographic and anthropological sources (Guidoboni and Ebel 2009; Hay 2005).

Ensuring the quality of documentary sources is important for enhancing the scientific value of the evidence (Hay 2005). Scott (1990) recommends the use of authenticity, credibility, representativeness and meaning in order to evaluate the quality of a document. In regard to the idea of authenticity, Scott (1990) refers to whether unquestionable sources were used and whether the evidence can be considered as genuine. Raising questions about the credibility of a source document is important in helping to determine whether the evidence provided has been in anyway distorted and is free from errors (Scott 1990). While drawing conclusions from the source documents, it is also important to consider whether the evidence represents the subject under investigation. Lastly, evidence must be ‘clear and comprehensive’ in relationship to its meaning (Scott 1990).

To maximise authenticity, credibility, representativeness and quality of the secondary data of earthquake events, different strategies are used in this research. For example, for each individual earthquake event, we have attempted to search for multiple sources and have cross-referenced different source documents. None of the sources were considered on face value (Hay 2005). In order to verify the descriptions given in historical documents, we undertook content analysis and the results were checked against the context of descriptions, effects of events and locations of occurrence where available. This was necessary because the sources we used were not originally generated for the purpose of geographical analysis (Baker 1997).

2.3 Published palaeo-earthquake records

The short time period covered by instrumental records can only provide a partial geography of hazards and associated risks (deMenocal 2001). For instance, based on recent instrumental data the Sumatra–Andaman region was not thought to have a high tsunami risk (Berninghausen 1966; Jankaew et al. 2008). However, many researchers (Monecke et al. 2008; Satake and Atwater 2007) began to report geological evidence of tsunamis in the region after the 2004 Indian Ocean tsunami. Therefore, it is essential to turn to geological evidence that spans much longer time series (Nott 2003). Palaeo-records can act as powerful sources of additional evidence that extends the records of past hazard events improving risk assessment (Dominey-Howes 2002). We have used geological evidence of earthquake events that are available in published sources to ascertain the source locations of these events in Bangladesh.

2.4 Recording of earthquake magnitudes

While collecting and collating earthquake magnitudes from secondary sources, we faced several challenges relating to the use of different magnitude scales and the lack of any magnitude scale for some events. On occasion, a source provides Modified Mercalli Intensity (MMI) values for the effects of an earthquake. However, when neither an MMI nor magnitude scale was provided, we have attempted to collect and review the effects of the earthquakes if available. After careful analysis of the effects, this has been converted to the MMI scale and a new MMI value is provided in Table 2.

2.5 Cross-checking and validation of data

A total of 168 sources were consulted to develop the new earthquake catalogue. With regard to those 168 sources, we identify and discuss issues arising such as differences in magnitude, inconsistencies in citations and variations in the date of occurrence. Accordingly, it was necessary to employ a rigorous system for cross-checking and data validation. This is achieved by cross-checking multiple sources and performing a content analysis. This process verifies the date of occurrence of an event, explores inconsistencies in citations, identifies problems relating to magnitudes of earthquakes and evaluates likely effects.

Every effort has been made to carefully and faithfully record details from the original document sources. However, as is understood for this type of research, no matter how careful the researcher, the possibility exists for transcription errors during compilation and we ourselves may have made errors and such errors will remain in the static catalogue we generate.

3 Results

A total of 562 earthquakes (Table 2) have been identified in the study region between 810BC and 2012. Of these events, one earthquake (event no. 5) was identified from archaeological sources, four earthquakes (event no. 1, 2, 4 and 8) were identified from geological sources, 218 (no. 3, 6, 7, 9–223) from historical records and the remaining 339 earthquakes (event no. 224–562) were identified from instrumental sources. Of the 562 earthquakes, magnitude data are not available for 63 events. Of the remaining 499 earthquakes, 253 events are recorded as having a magnitude of less than 6. The number of earthquakes with a magnitude ranging 3–3.9, 4–4.9 and 5–5.9 is 14, 93 and 146, respectively (Table 2). A total of 246 earthquakes are reported to have occurred with a magnitude of over 5.9. Of these, 14 events are reported to have had a magnitude over 7.9. These events occurred in 825–835, 1663, 1762, 1797, 1833, 1861, 1897, 1903, 1928, 1950, 1969, 2004, 2005 and 2007. The catalogue comprises the event number, the event date, earthquake locations (including coordinates and geographical location names), earthquake parameters (magnitude, intensity and focal depth), earthquake effects (deaths, injuries, damage in million US $ and house destroyed) and information about the sources used to identify that event. Last, any relevant notes about the event are listed in a column marked ‘comments’ (Table 2).

3.1 Problems identified in historical earthquake data

While collecting and collating data from the source documents, inconsistencies and discrepancies were observed in the data. For example, for the 23 August 1936 event, the NGDC (2012) reported nine deaths. However, Hamzah et al. (2000) reported 91 deaths. We note that there were differences in dating for the same event as well as incorrect and/or incomplete citations. These issues are detailed in the following sections.

3.2 Incorrect and incomplete citations

Numerous incorrect and incomplete citations in sources were found while recording the data. For example, for the 13 August 1821 event (no. 62), Oldham (1883) refers to the Asiatic Journal (1822b) volume 13. When referencing material for the 8 July 1834 (no. 84) and the 21 July 1834 (no. 85) events, Oldham (1883) refers to the Asiatic Journal (1834), volume 13. This raises questions regarding volume 13. So, volume 13 was published in 1822 and 1834? Rao and Rao (1984) identified the occurrence of an earthquake on 31 March 1843 (no. 102) and referenced Oldham (1883) as the original reference. However, after careful checking of Oldham (1883), we confirm that no event was recorded on that date. Another example of inconsistent referencing is found in relation to the 1861 Kolkata earthquake. For the 16 February 1861 (no. 153) Kolkata earthquake, Oldham (1883) referred to the ‘Journal of Asiatic Society of Bengal’ and ‘Friends of India’, but he failed to provide a year, volume and issue number for either source.

3.3 Differences in dating between sources for the same event

For a number of earthquakes, we observed that the event was given incorrect or different dates between the various sources. Oldham (1883) mentioned an earthquake that occurred on the 16 February 1861 (no. 153) in Kolkata. Later, Chandra (1977) and Nandy (1994) cited the date as the 18 February 1861, despite using Oldham (1883) as their original source. We have checked the original reference (Oldham 1883), where the event date is recorded as 16 February 1861. Rao and Rao (1984) indicated an earthquake on the 8 November 1863 providing the original reference of Oldham (1883). However, after careful checking we confirm that Oldham (1883) recorded the earthquake as occurring on the 18 November 1863 (no. 160).

3.4 Differences in magnitude and intensity of earthquakes

On some occasions, earthquake data sources do not include any magnitude scale. However, where magnitude has recorded, we also observed differences in the reported magnitude value between sources. For example, in our earthquake catalogue, some data have been collected from two Indian earthquake catalogues, Chandra (1977) and Rao and Rao (1984). Neither of these catalogues provides specific earthquake magnitudes for their events. While they listed magnitude scales such as M L, M b and M s that are generally indicative of magnitude scales of their catalogues, they did not attribute these scales to individual events. When a specific magnitude scale (Table 2) was available for the events recorded from source documents, we have incorporated it in the new catalogue by providing superscript numerical values suggesting the source code for original references (Table 2).

For the 21 July 1956 earthquake (no. 340), Chandra (1977) provided a magnitude of 6.1, but Rao and Rao (1984) provided 7. Again, for the 2 March 1964 event (no. 364), Hamzah et al. (2000) provided a magnitude of 5.2, but the NGDC (2012) provided a magnitude of 7, which indicates a substantially larger event. For the 9 December 1751 earthquake (no. 32), both Rao and Rao (1984) and Chandra (1977) cite Kelkar (1968) as their source for the intensity value. However, Rao and Rao (1984) indicate an intensity value of V, while Chandra (1977) lists an intensity value of VI. This means that they incorrectly copied from the source documents or came to different conclusions based on the same sources. The same thing occurs for the 31 October 1757 earthquake (no. 35). Quoting Kelkar (1968), Chandra (1977) provided an intensity value of V for this event, while Rao and Rao (1984) gave an intensity value of IV. Similarly for the 9 February 1823 earthquake (no. 66), quoting Oldham (1883) as the original reference, Chandra (1977) provided an intensity of value of VI, but Rao and Rao (1984) provided an intensity value of VII. Again for the 18 August 1828 event (no. 75), Rao and Rao (1984) provided an intensity value of V, but Chandra (1977) provided an intensity value of VII, even though they both quoted Oldham (1883).

We can only report eight earthquake events prior to 1500 (Table 2). A total of 47 events were reported prior to 1800 (Fig. 2a). Since 1800, the number of reported earthquakes increases greatly. This may be attributed to the introduction of instrumental methods of recording around 1900 (Bolt 1930). A total of 352 earthquake events are reported to have occurred between 1900 and 2012. Between 1900 and 1949, the NE Indian Ocean region experienced 110 earthquakes (Fig. 2b), while a further 242 earthquakes occurred in the same region between 1950 and 2012 (Fig. 2b).

Fig. 2
figure 2

Distribution of earthquakes in the NE Indian Ocean and adjacent coast through time. a By century, b by decade between 1901 and 2012. Few earthquakes were reported before 1800. The number of earthquakes reported greatly increased from the eighteenth century onwards

4 Discussion and conclusions

This research documented and analysed geological, archaeological and documentary records of earthquakes in order to construct a complete history of past events of magnitude 4 and above—a first step towards improved earthquake risk reduction. The catalogue presented here (all earthquakes in the historic and prehistoric periods; and a magnitude of more than 4 in the instrumentally recorded period) supersedes all the previous catalogues that document earthquakes in the northern Bay of Bengal and Bangladesh. That said, such a document would likely miss some events due to the unavailability of complete information sets and the time constraint of the research. However, a more accurate reconstruction of the historical time series of earthquakes that has affected Bangladesh was hindered by a variety of problems in data. The types of data problems that we faced while collecting, collating and analysing historical data included incorrect and incomplete citations, incomplete records, inaccurate and contradictory description of the dates of occurrence, only one source for some events, under and over estimated magnitude and deaths associated with the events, non-stationary trends, missing data and misidentified phenomena.

Of some concern, we observe that former tsunami catalogues have variously described individual events as an earthquake, a tsunami or a cyclone. After a careful review of historical earthquakes in India provided by Oldham (1883) and the Bangladesh Meteorological Department (2011), this research revealed that on at least six occasions, earthquakes and tropical cyclones occurred simultaneously making classification of the events problematic. These events occurred on the 26 May 1618, 11 October 1737, 19 October 1800, 10 December 1807, 9 June 1869 and 2 September 1895. Oldham (1883) reports that on the 26 May 1618 a severe tropical cyclone and earthquake occurred simultaneously resulting in 2000 deaths and the loss of 60 vessels at Mumbai. However, no other descriptions of earthquake effects were mentioned. Later, Milne (1911), Chandra (1977) and Rao and Rao (1984) referred to Oldham (1883) and added this event to their respective earthquake catalogues, but did not add further descriptions of the effects (other than quoting the original reference). The event is also listed in other earthquake catalogues (e.g. Bapat et al. 1983; Chaudbury 1965; Milne 1911; NGDC 2012). This earthquake probably also caused great destruction to the Delta of the Indus River (Chaudbury 1965). Although major regional tropical cyclone catalogues (i.e. Jayanthi and Sarma 1986; Murty and El-Sabh 1992) failed to include this event, we have added it to both this earthquake catalogue and our previously published tropical cyclone catalogue (Alam and Dominey-Howes 2015). However, we recommend undertaking further research to validate the event.

The event of 1737 was reported by Murty and Rafiq (1991) as a tsunami that penetrated up the Hooghly River following an earthquake near Kolkata. Milne (1911) who referred to Oldham (1883) also listed this event as an earthquake. That said, we note in reference to Oldham (1883) that this event was reported as relating to an earthquake and a tropical cyclone (but not a tsunami) occurring simultaneously. Bilham (1994) decided that the event was just a tropical cyclone based on field study and reference to documents contained in the British Library.

For the 19 October 1800 event, Oldham (1883) referred to the Asiatic Annual Registrar of 1801 published in 1802 (Anon 1802), mentioning a severe earthquake lasting a minute that occurred during a violent tropical cyclone in Andhra Pradesh. Both Chandra (1977) and Rao and Rao (1984) listed this event as an earthquake and cited Oldham (1883) and Kelkar (1968). The original reference, Anon (1802) stated:

On the 19th about 10 minute after 4 A.M., the wind blew a hurricane, when we felt a severe shock of an earthquake; which kept the earth in continual agitation for nearly a minute. It shook down many houses, but I believe no person was killed in consequence.

The Indian state of Andhra Pradesh where the event occurred is seismically active and prone to tropical cyclones (Dube et al. 2009; Murty and El-Sabh 1992). From the descriptions of the original reference (Anon 1802), the occurrence and effects of a tropical cyclone and an earthquake are clear. Therefore, the event has been included in both our tropical cyclone (Alam and Dominey-Howes 2015) and earthquake catalogues.

For the 10 December 1807 event, Oldham (1883) referred to the Asiatic Annual Registrar of 1808 published in 1811 and described a sever tropical cyclone with accompanying heavy seas at Chennai during which three (earthquake) shocks occurred. No other effects of the earthquake were mentioned. Later, Chandra (1977) and Rao and Rao (1984) referred to Oldham (1883) and added this event in their respective earthquake catalogues without further validation and descriptions of the effects. Anon (1811) who is the author of ‘the Annual Registrar 1808’ mostly described the changes in the direction of wind, the formation of tropical depression and the occurrence of a tropical cyclone in Madras between 9 and 10 December. For example, Anon (1811, p. 128) stated:

When it had completely veered round to the southward it suddenly burst into a hurricane, the like of which was never remembered at Madras.

The canal forced its banks and overflowed the country as far as the government bridge on one side, and beyond the powder mills on the other, where the water was three feet deep. Boats were carried away, and several were found at the burying ground, and one near the government gardens.

The sea rose much beyond its usual height, bringing some of the Massoulah boats within a few yards of the Custom House, and destroying others to the amount, we understood of about 40. …The mud houses of the natives are in most places swept away, and with them many of their wretched inhabitants. It is apprehended many thousands have perished. Dead bodies of men, women and children, were found lying in every direction when the storm abated.

From the accounts given by Anon (1811), the event was definitely a severe tropical cyclone. Jayanthi and Sarma (1986) in their historical tropical cyclone catalogue of India included this event as a tropical cyclone and did not mention any earthquake. We cannot obtain further original source documents for this event. To be consistent with the primary and secondary sources, we included the event in both our earthquake and tropical cyclone catalogues (Alam and Dominey-Howes 2015). However, we would recommend further validation of this event as to whether it should be included as a simultaneously occurring earthquake and tropical cyclone.

For the 9 June 1869 event, Oldham (1883) referred to the Times of India describing an earthquake shock during the occurrence of a tropical cyclone. However, no detailed information about the source was provided, limiting validation of the original source. However, Oldham’s (1883) description about the occurrence of a tropical cyclone and an earthquake is clear, mentioning slight shock of an earthquake during a tropical cyclone in Kolkata (formerly ‘Calcutta’). Chandra (1977) listed the event in the earthquake catalogue referring to Oldham (1883) as a source document and provided a value of V on MMI scale for Kolkata. We have added the event in both the earthquake and tropical cyclone catalogues according to Oldham’s (1883) description. We suggest that it is necessary to examine further sources to determine whether both a tropical cyclone and an earthquake actually did occur simultaneously.

The coincidence of an earthquake and tropical cyclone was reported to have occurred on the 3 September 1895 and struck Khulna in Bangladesh (BMD 2011). For this event, the Bangladesh Meteorological Department (2011) notes that the entire region was devastated by a simultaneous strong earthquake and tropical cyclone. No other catalogues of earthquakes for that region include any event occurring on this date. Additionally, the region only has a limited historic record of earthquakes. The Bangladesh Meteorological Department is responsible mainly for recording meteorological events and consequently may not accurately record earthquakes. Therefore, due to a lack of references for this event, we have only included it in the tropical cyclone catalogue. However, we would recommend further research to determine whether any earthquake occurred on that date.

Although we have undertaken rigorous validation of processes of our earthquake data through cross-checking and repeated document analysis, the catalogue still may not be error free. The reason for this is that it is a hand-entered catalogue. It should be borne in mind that MMI values are often skewed to higher values by locations with higher shaking susceptibility, certain types of soil texture or locations with buildings that respond to a particular ground shaking frequency in disastrous ways (Ambraseys and Douglas 2004). As such, the MMI that we have provided in Table 2 should be treated with caution for the purpose of reuse and further research to more accurately refine these MMIs is required but was beyond the scope of this research. If errors are identified in our new catalogue, we would be grateful if researchers could notify the lead author.

The data in the earthquake catalogue may be useful to formulate the earthquake risk reduction strategies in Bangladesh and adjacent region. We are undertaking further research and analysis of the earthquake data contained in the catalogue to understand frequency and intensity and to reconstruct spatial risk pattern in Bangladesh. The results of the analyses are forthcoming in different outlets.