Disaster declarations associated with bushfires, floods and storms in New South Wales, Australia between 2004 and 2014

Australia regularly experiences disasters triggered by natural hazards and New South Wales (NSW) the most populous State is no exception. To date, no publically available spatial and temporal analyses of disaster declarations triggered by hazards (specifically, bushfires, floods and storms) in NSW have been undertaken and no studies have explored the relationship between disaster occurrence and socio-economic disadvantage. We source, collate and analyse data about bushfire, flood and storm disaster declarations between 2004 and 2014. Floods resulted in the most frequent type of disaster declaration. The greatest number of disaster declarations occurred in 2012–2013. Whilst no significant Spearman’s correlation exists between bushfire, flood and storm disaster declarations and the strength of the El Niño/Southern Oscillation (ENSO) phase, we observe that bushfire disaster declarations were much more common during El Niño, and flood disaster declarations were five times more common during La Niña phases. We identify a spatial cluster or ‘hot spot’ of disaster declarations in the northeast of the State that is also spatially coincident with 43% of the most socio-economically disadvantaged Local Government Areas in NSW. The results have implications for disaster risk management in the State.

Emergency and Rescue Management Act 1989 No 165 (SERM Act), ensures NSW has a system to cope with disasters. At the heart of the National Risk Assessment Guidelines that the NSW government and its emergency managers use to plan for disasters is the requirement for fundamental data on the types, frequencies, distributions and impacts of hazards 11 . Given the variety of hazards that may affect NSW, we note that there are no publically available spatial and temporal analyses of disaster declarations. This contrasts with other jurisdictions around the world such as in the United States where others have conducted similar spatial and temporal analyses of hazards and their accompanying disasters using publically available datasets [12][13][14][15] . Further, in NSW, no studies have explored the relationships, if any, between socio-economic disadvantage and the spatial occurrence of hazards that result in disaster declarations. This omission stands as an obstacle for supporting effective disaster risk management planning and policy.
Given the omissions just identified, the aims of this study are to: (1) source, collate, analyze and map information about disaster declarations in NSW between 2004 and 2014 focusing on the commonly recurring and sudden onset types of bushfires, floods and storms mirroring similar studies undertaken in the United States; (2) use the results of (1) to identify patterns and trends in spatial and temporal disaster declarations using spatial techniques and statistics; (3) examine for any obvious relationships between the frequency and distribution in space and time of the disaster declarations and the intensity of the El Niño/Southern Oscillation (ENSO) climate phenomena using appropriate statistical techniques; (4) determine whether the occurrence of disaster declarations are spatially clustered in NSW; (5) explore whether there are any associations between socio-economic disadvantage and 'hot spot' clusters of disaster declarations; and (6) to use the results of (1) to (5) to consider the implications for future disaster risk reduction planning and management by relevant State emergency management agencies.
We focus only on NSW since it is our home State, access to suitable disaster declaration data was readily available, we have a history of cooperation with the State emergency service authorities and understand their data needs and because NSW is the most populous State in Australia. Consequently, hazards and their associated disasters can be expected to affect significant proportions of the national population. Further, we are concerned with rapid onset disasters that are the remit of the Rural Fire Service (bushfires) and the State Emergency Service (floods and storms) to manage. We do not consider drought or heatwaves since they are not the responsibility of the State emergency service organizations to manage and we exclude tropical cyclones since none are known to have made landfall in NSW. We focus on the period from 2004 to 2014 since the dataset of disaster declarations is most complete for this period. Prior to this, only fragmented data on disaster declarations exists making temporal analyses unreliable.

Results
Total disaster declarations 2004-2014. Between 2004 and LGAs were included in natural disaster declarations 905 times -although the actual discrete number of separate events was 207. Table 1 provides the total number of LGAs experiencing disaster declarations by hazard type, by financial year. As indicated in Methods, this does not mean that there were 905 separate disasters. One event may have simultaneously affected four, or twenty or fifty LGAs. However, for the purposes of description and mapping which LGAs were affected, we note 905 declarations. The actual number of separate bushfire events was 108, of flood events was 44 and of storm events was 55. Figure 2 maps the disaster declarations noted in Table 1  LGAs experienced the highest number of disaster declarations at 10 and 9 events, respectively. Results of the Global Moran's I statistic allowed for rejection of the null hypothesis that disaster declarations were randomly distributed across NSW for all declaration types. The Getis-Ord ⁎ G i analysis by LGA showed a statistically significant cluster of high disaster declarations or 'hotspot' located in the northeast of the State and clusters of low disaster declarations or 'cold spots' located southwest of Sydney and in the central southern region of the State (Fig. 3D).

Bushfire declarations 2004-2014. In total,
LGAs were included in bushfire disaster declarations on 319 occasions (Table 1). Very few LGAs were entirely free of a bushfire disaster declaration during this ten-year period, but those that were, were mostly confined to the southeast of the Greater Metropolitan Region and the far southwest of the State (Fig. 4a). The LGAs that experienced the most number of bushfire disaster declarations were largely confined to the northeast of the State (Fig. 4a). Singleton, the Clarence Valley and Narrabri experienced nine, eight and eight bushfire disaster declarations respectively. The Blue Mountains, Lithgow and Port-Macquarie Hasting rounded out the top five bushfire declared LGA's with seven disaster declarations each.

The Blue Mountains and Hawkesbury
LGAs on the western fringe of the Sydney Metropolitan Region experienced the most bushfire disaster declarations at seven and five each (Fig. 4a). However, the Getis-Ord ⁎ G i analyses determined a bushfire disaster declaration hotspot in the northeast of the State and a large cold spot in the central south of the State (Fig. 3A).  Table 1.  (Table 1). Thirty-eight LGAs were entirely free of a flood disaster declaration during the study period and these were confined to the Greater Metropolitan Region (Fig. 5a). The LGAs that experienced the most number of flood disaster declarations were again largely restricted to the northeast of the State (Fig. 5a). The Clarence Valley, Bellingen and Nambucca experienced 12, 12 and 11 flood disaster declarations respectively. Ballina, Byron, Coffs Harbour, Lismore and the Richmond Valley followed with nine flood disaster declarations each. However, the Getis-Ord ⁎ G i analyses show a clear flood disaster declaration hotspot in the northeast of the State and a disaster declaration coldspot in the area to the west-south-west of Sydney (Fig. 3B).  (Table 1).

Forty-eight
LGAs were entirely free of a storm disaster declaration during this ten-year period and these were confined to the south and eastern LGAs of the Greater Metropolitan Region, the State's central regions and the central north (Fig. 6a). The LGAs that experienced the most number of storm disaster declarations were again largely restricted to the northeast and central coast of the State (Fig. 6a). Nambucca, Lismore, Greater Taree and the Shoalhaven each experienced six storm disaster declarations respectively. Ballina, Bellingen, Byron, Gloucester, the Great Lakes, Kiama and Wyong all experienced five storm disaster declarations. A hotspot of high storm disaster declarations along the northeast coast is clearly shown in the Getis-Ord ⁎ G i results (Fig. 3C). A local cluster of low storm disaster declarations (cold spot) is also evident in the central southeast region of the State. Figure 6b shows   (Table 1).

Bushfire, flood and storm disaster declarations and the ENSO. A non-significant correlation
between ENSO (as measured by the NINO3.4 index -see Methods) and bushfire disaster declarations of 0.19 (p = 0.27) was found, meaning that the current strength of ENSO is not a good predictor of the number of disaster declarations made. There is, however a difference between the two phases, with 45 disaster declarations made during active El Niño phases and 28 during La Niña phases. This means that during our study period, bushfire disaster declarations were more common during the El Niño, likely due to the hot and dry conditions that occur during this phase. This is in line with recent research that looked at the relationship between El Niño, extreme heat, and bushfires 16 .
The Spearman's correlation between flood disaster declarations and the strength of ENSO was inconclusive but somewhat negative at − 0.29 (p = 0.21), meaning that the strength of the ENSO phase is not a direct indicator of the total number of flood disaster declarations for a given month. However, disaster declarations for this hazard type were almost five times more common during La Niña phases (24 disaster declarations) compared to El Niño (five disaster declarations). Since La Niña phases are typically associated with higher than average rainfall, this result is not wholly unexpected. However, a five-fold increase in disaster declarations between the two active phases may provide some imperative to emergency services to prepare in advance.
Correlations with ENSO and storm disaster declarations were similar to flood declarations described above, also measured at − 0.29 (p = 0.14). The difference between El Niño and La Niña phases is less impressive (nine and 17 disaster declarations, respectively). However, it still reflects the known relationships between ENSO and eastern Australian rainfall. Such results may still be useful for emergency services preparedness, as storm disaster declarations were almost twice as likely during La Niña phases compared to El Niño. Storms are also smaller in spatial scale compared to floods, and so are influenced by other meteorological and climatological mechanisms that operate on similar scales.

Disaster declarations and socio-economic disadvantage. Fourteen rural LGAs in NSW have
IRSAD scores in the lowest decile, representing the 10% most socio-economically disadvantaged communities in Australia (Table 2). Significantly, six of these LGAs (43%) are located within the total disaster declaration hotspot (99% confidence) shown in Fig. 3D (see Table 2). This shows a large proportion of the most socio-economically disadvantaged LGAs are experiencing significant numbers of disaster declarations.

Discussion
The twenty-seven LGAs that experienced no disaster declarations were all located within the Greater Metropolitan Region, the part of NSW with the highest percentage of the overall State population. Consequently, during the study period of 2004 to 2014, the majority of the State's population had no direct experience of hazards that led to disaster declarations. Although contested, it is well documented that prior experience is a strong predictor of increased individual and community awareness and where awareness is higher, affirmative risk mitigation behavior is generally higher [17][18][19][20][21][22] . The reverse is also the case where low experience and risk awareness lead to lower adaptive behaviours. A potential compounding issue is that emergency service organizations and their local response units and the emergency managers themselves, are not regularly responding to hazards occurring within their own jurisdictions so there might be a drop in readiness and capacity to respond in the future when hazards impact these LGAs [23][24][25] . Consequently, LGAs, their communities and local emergency management officials may be less prepared and more vulnerable to future hazards. Resources should be provided to emergency management officials and communities to assist them to prepare. Having said this, we acknowledge the possibility that emergency service managers, staff and volunteers in these LGAs have volunteered their services during emergencies and disasters elsewhere in the State and across Australia (and internationally), thus gaining insights and experiences that would be useful to their home jurisdictions. Further, it may be that LGAs in the Greater Metropolitan Region have more human and material resources available to them to prepare and respond to hazards. In either case, further research is needed to determine levels of preparedness in NSW LGAs.
Floods result in the most frequent disaster declarations. This would suggest that a focus on community education and engagement and allocation of resources to the State Emergency Service to support them to manage future floods would be especially helpful -although not at the expense of preparation for other hazard types. Such a focus on flood risk would be of value, particularly given research after the catastrophic floods in Queensland in 2011 demonstrated the need for education, risk communication and community knowledge in order to empower local communities to increase their flood resilience, with a surprising number of people indicating they had no idea they were at risk from floods 26,27 .
Whilst there were no significant Spearman's correlations between bushfire, flood and storm disaster declarations and the respective phases of the ENSO, it is clear that bushfire and flood disaster declarations were more common during active El Niño or La Niña phases of the ENSO respectively. Several possibilities exist to explain this increased frequency of disaster declarations. First, it may be statistically true that more events occurred leading to disaster declarations. Second, it may be that the events that occurred were larger in geographic extent and therefore impacted more LGAs leading to greater numbers of disaster declarations. Third, the individual events may have been more intense leading to higher numbers of disaster declaration. Further research is needed to gain a better understanding of what exactly is happening and specifically, efforts to improve the number of data points and extend the time series would greatly help. What the results do suggest is that there is a complex relationship between meteorological conditions, early warning and a community's preparedness and whether an event results in a disaster. This has implications for emergency service organizations as the ENSO shifts phase.
The overall pattern of disaster declaration occurrences in NSW was spatially clustered, as determined by the Global Moran's I statistic. We are unable to explain why for each hazard type examined, there is a statistically significant 'hot spot' of disaster declarations located in the same northeast region of the State. It is likely that this has something to do with inherent locational vulnerability.   Whilst finer spatial scale analysis is required to establish the relationship between socio-economic disadvantage and disaster declarations, we consider the fact that 43% of the most disadvantaged LGAs are located within the disaster declaration hotspot to be significant from the perspective of disaster preparedness and response. This geographic overlap of disadvantage with frequent declaration of disaster presents challenges to emergency services, communities and governments that have to prepare for and respond to natural hazards. We do acknowledge the coarse scale of our spatial unit of analysis at the LGA level, noting that there will be heterogeneity within LGAs. Consequently, finer scale work is needed to examine local relationships at higher resolution scales such as 'mesh block level' 28,29 . Having acknowledged this limitation, we note that for the hotspot in the northeast of the State, our data mirrors recent high-resolution (postcode level) analysis of disadvantage 30 .
Those interested in and responsible for the management of hazards and their associated disasters are concerned with a number of key questions. For example, what does Australia's hazard landscape and disaster history look like? What types of hazard can we expect to affect Australia and what are their distributions, frequencies, magnitudes and return periods? What impacts and effects do they have on people, communities and the things we value? What technologies exist to forecast and predict events before they occur, and monitor and warn once an event has begun? What do individuals, families and communities understand about the risks associated with different hazards in Australia and what methods, tools, approaches and actions are available to enable them and our governments to reduce risk, increase resilience and safeguard us from disaster? Tied to this last question is what barriers and obstacles stand in the way of preventing individuals, communities and governments from taking actions that enhance resilience and reduce vulnerability? It is beyond the scope of the present study to attempt to tackle this list of questions (such a response is deserving of an entire publication in its own right). However, we note here that considerable research is on going to address these questions that builds upon a long scholarship in Australia focused on hazards and their accompanying disasters [31][32][33][34][35][36][37][38][39][40] . Such studies, as well as many others, have laid important foundations about the what, where, when, how and why of hazards and their accompanying disasters, against which more contemporary analyses and trends might be investigated and benchmarked. Further, they provide important insights and collectively track over time how well Australia is managing and mitigating the risks associated with hazards and their accompanying disasters.
Given the dynamic nature of our socio-ecological systems, this process never really ends. Consequently, there is a critical need to continuously assess what we know about hazards and the disasters they cause. This information can then be used to evaluate and modify if necessary, disaster risk reduction policies and practices. This study is a contribution to that effort.

Methods
Data Collection. Data about Natural Disaster Declarations in NSW were sourced from the NSW Ministry for Police and Emergency Services (MPES) website (2014). Any declaration that impacted multiple LGAs was counted once for each LGA in the declaration. That is, a single bushfire disaster (for example) might impact six LGAs resulting in us noting six bushfire disaster declarations (rather than one disaster declaration). Lord Howe Island, an autonomous island off the coast of NSW, was excluded, and declarations for LGAs amalgamated in the past ten years were attributed to the relevant LGA from the 2011 census. Declared disaster types were simplified in to one of three principal categories, 'Bushfire' , 'Flood' and 'Storm' . In relation to 'Storm' events, we collapsed a variety of event types in to the singular 'Storm' including 'severe weather' , 'storms and flooding' , 'storms' , 'storms, flooding and landslides' , 'floods and storms' , 'dust storms' , 'wind storms' , 'severe hailstorm' and 'severe hailstorm and wind storm' . Combined storm and flood events were counted once in each category, but only once for final disaster declaration totals. Totals by type and in aggregate form for all LGAs affected are shown in Table 2.
Collation and analysis of relevant climate data. The El Niño/Southern Oscillation (ENSO) phenomenon is a known driver of climate over eastern Australia 41 . ENSO is quasi cyclical, oscillating every 2-7 years between El Niño (generally hot and dry conditions), neutral (average conditions) and La Niña (generally cool and wet conditions) [42][43][44] . It is most active during the Australian summertime. However, its influence on the climate can extend outside this season. In order to understand any relationships between ENSO and disaster declarations in NSW, the NINO3.4 index was used to represent the state of ENSO at the monthly timescale. We use the pre-calculated index from the National Oceanic and Atmospheric Administration (NOAA; http://www. cpc.noaa.gov/data/indices/). NINO3.4 measures sea surface temperature anomalies in the Central Equatorial Pacific (5°S-5°N, 170°W-120°W), where standardized values smaller than − 0.5 represent La Niña conditions, and values greater than 0.5 represent El Niño conditions. Here we determine how many bushfire, flood and storm disaster declarations occur during active La Niña and El Niño phases of ENSO. We then calculated the nonparametric Spearman's correlation coefficient to determine whether the strength of the state of ENSO can provide information about the number of disaster declarations per hazard type. Correlations are deemed significant at the 5% level.
Mapping and geospatial analyses. Data about disaster declarations were linked with shapefiles of LGA boundaries from the Australian Bureau of Statistics' 2011 Australian Standard Geographical Classification data 45 in ArcGIS 10.2. The distribution of disaster declarations was mapped for four categories: bushfires, floods, storms and total disaster events, over the ten-year period from July 2004 until June 2014 (correlating with the Australian financial year). The number of disaster declarations for each hazard type was also mapped for each financial year to examine finer temporal scale variations.
The spatial distribution of disaster declarations in NSW was examined using the Global Moran's I statistic 46 to determine whether the overall pattern is clustered, dispersed or random. The null hypothesis was that disaster declarations were randomly distributed across NSW. The statistical significance of Global Moran's I was tested using the z-score, the standardized difference between the observed and expected values, and p-value. The Getis-Ord ⁎ G i , statistic, which measures the degree of spatial clustering of a local sample relative to the mean 47 , was then applied to test for the presence of local clusters. For both tests a binary weights case was applied using a fixed distance band weighting procedure. Appropriate lag distances were determined for each hazard type using incremental spatial autocorrelation. The lag distance or distance band defines the geographical area surrounding the LGA considered in the analyses and will reflect underlying processes influencing cluster patterns which may operate at different scales for each hazard type.
LGAs within the Greater Sydney Area were removed from the cluster analysis due to differences in scale. The lag distance of 158 km was specified for floods, storms and total disasters and 223 km for bushfire. The false discovery rate (FDR) criterion was applied to adjust for multiple testing 48 . Positive ⁎ G i values indicate statistically significant spatial clustering of LGAs with high disaster declaration occurrences and negative values indicate spatial clustering of LGAs with low disaster declaration occurrences.
The Australian Bureau of Statistics' 2011 Socio-economic Indexes for Areas (SEIFA) data was used to examine the spatial relationship between clusters of high disaster declarations and vulnerable communities. The SEIFA Index of relative socio-economic advantage and disadvantage (IRSAD) provided a proxy indicator for less resilient and more vulnerable communities following an established approach 49 .
LGAs with an IRSAD decile of 1, indicating the 10% most disadvantaged communities in Australia, were identified and compared with spatial clusters of high disaster declaration occurrence.