Rapid regional-scale assessments of socio-economic vulnerability to climate change

Assessing socio-economic vulnerability to climate change impacts to support regional decision-making is conceptually and practically challenging. We report on research that tested a rapid assessment approach of socio-economic vulnerability in Australia’s natural resource management regions. The approach focuses on regionally important economic sectors, identified using existing datasets, which are likely to be sensitive to climate change impacts. Disaggregated spatial representations of factors known to be associated with vulnerability function as multiple lines of evidence for highlighting intra-regional hotspots of high potential vulnerability. Our results show that a small number of factors based upon contextually relevant empirical evidence offers a low-cost, rapid assessment process, which is readily transferable across regions and provides end-users with guidance for interpreting the results within the context of regional conditions.


Introduction
Successful climate change adaptation is dependent upon the provision of high quality information at scales commensurate with the jurisdictions of organisations responsible for adaptation policy (Cash et al 2006, Measham et al 2011, Lemieux et al 2014. Consistent with moves towards regionalisation across a range of policy sectors (Morrison 2007), the 'region' has emerged as the preferred scale for many forms of decision-making such as natural resource management (NRM) in several developed countries (Gardiner et al 1994, Bryce et al 1999, Jennings and Moore 2000, Brunckhorst 2002, Clark et al 2005, Holmes et al 2005. Despite recent attempts to integrate vulnerability assessment approaches developed by the hazard and climate change communities (Costa and Kropp 2013, IPCC 2014, there is no accepted approach for assessing socio-economic vulnerability to climate change impacts and environmental hazards that may be used to inform regional priorities and decision-making.
To address this gap, we developed an approach to assess intra-regional socio-economic vulnerability using multiple lines of evidence derived from disaggregated spatial representations of regionallyrelevant factors known to be associated with vulnerability. We applied the approach to priority sectors in three Australian NRM regions, which were selected by applying the concept of resource dependency (Marshall et al 2007, Marshall 2011. Here, we demonstrate the utility of this approach by presenting a spatially explicit snapshot of the socioeconomic vulnerability of the South East Queensland (SEQ) horticultural sector. Multiple lines of evidence allows nuanced assessments of the differences in potential vulnerability between subregions. Crucially, the approach facilitates rapid assessment and renders it readily transferrable across NRM organisations with differing resources, capacities and funding cycles (Robins and Dovers 2007).
The field of climate change vulnerability assessment is characterised by a diversity of valid approaches that vary in their complexity and are operationalised at different scales to address climate change policy questions (Füssel and Klein 2006). Despite this diversity, rapid and cost-effective approaches to assess socio-economic vulnerability suited to regional-scale decision-making are rare. Responding to this pressing research need is conceptually and practically challenging. These challenges may be overcome by framing assessments using the concept of resource dependency and selecting regionally-relevant factors known to influence socioeconomic vulnerability to climate change impacts. In line with Australia's commitment to regional-scale NRM decision-making (Hajkowicz 2009), we tested how these techniques may be used to explore and assess socio-economic vulnerability in three NRM regions on the east coast of Australia. While we focus upon SEQ here, we demonstrate the transferability of our approach by presenting the vulnerability assessments for the Burnett Mary (Queensland) and Hawkesbury-Nepean (New South Wales) regions in this article's supplementary information.
2. Operationalising resource dependency at the regional scale Marshall et al argue that resource dependency is a useful means to measure climate sensitivity because social systems that are more dependent upon natural resources may be more sensitive to the changes in the biophysical systems anticipated under climate change (Marshall et al 2013. While the concept has been developed to measure climate sensitivity at the scale of individual actors (Marshall et al 2013), it has important utility for assessing vulnerability at larger scales because it provides an intuitive means by which regional-scale vulnerability assessments may be framed. To do this, we used employment data from the Australian Census of Population and Housing (2011) to determine upon which agricultural sector each NRM region was most socially reliant. We combined these data with value of agricultural commodities produced (VACP) data from the Australian Agricultural Census (2010-11) to determine upon which agricultural sector a region was most economically reliant. Spatial representations of these data provide an indication of the intra-regional distribution of the dominant agricultural sector against which spatial representations of factors known to influence socio-economic vulnerability may be compared.
In SEQ, horticulture was the most socially and economically important agricultural sector (table 1 and supplementary information). We do not suggest that the horticultural sector is the most climate sensitive sector in SEQ, rather operationalising resource dependency in this way functions as a prioritisation tool to target regionally important sectors. Simply put, we cannot focus on everything all of the time.

Identifying regionally-relevant factors for assessing vulnerability
Regionally-relevant factors for exploring and assessing socio-economic vulnerability were identified from a systematic review of the empirical, peer-reviewed climate change impacts and climate related hazards literature in our jurisdictions of interest (i.e., Queensland and New South Wales) (Smith et al 2015). Selecting factors on this basis is consistent with contemporary perspectives conceptualising vulnerability as a place-specific phenomenon (Adger 2003, Calgaro et al 2013. Only those factors that repeatedly were shown to influence socio-economic vulnerability were used in our assessment. Creating spatial representations of these factors was facilitated by freely available, national datasets that are appropriate for use at regional scales. The five factors and indicators used to represent them were: (1) reliance upon agriculture using the percentage of the labour force employed in agriculture; (2) geographic remoteness using the Australian Bureau of Statistics' (ABS) Remoteness Structure; (3) socio-economic disadvantage using the ABS' Index of Relative Socio-economic Advantage and Disadvantage (IRSAD); (4) economic diversity using the Hachman Index; and (5) age using workforce age profiles (table 2).
While composite indicators are often created to assess vulnerability at broader and finer scales using large numbers of indicators and statistical techniques to combine them (Rygel et al 2006, Solangaarachchi et al 2012, Arthurson and Baum 2015, they often obscure the key influences underlying the resultant vulnerability scores. Thus, our approach represents a deliberate departure from composite index approaches. Instead, to enhance rapid assessment and transferability to other regions the factor maps remain disaggregated, each representing one line of evidence in the overall assessment. Intra-regional vulnerability may be explored by interpreting the spatial heterogeneity of each map in line with the ways in which each factor influences vulnerability based upon the peerreviewed empirical literature (table 2). The rationale of the approach is that areas in which multiple vulnerabilities coincide highlight subregions of higher potential vulnerability compared to the wider regional context. Although the factors used here are directly relevant to the socio-ecological systems in our region of interest, the small number of factors for which there is empirical evidence included in this assessment invites some caution. In addition, there may also be instances where significant vulnerability in one area may distort the utility of a focus on intersecting lines of evidence. Thus, we qualify our terminology with the word 'potential', suggesting that this rapid assessment approach should be viewed as a precursor to more detailed, contextual vulnerability studies. Before presenting the socio-economic vulnerability assessment for the SEQ horticultural sector, we outline the methods and data compilation procedures. Figure 1 summarises the multiple lines of evidence approach to socio-economic vulnerability assessments in four steps.

The indicators
A region's reliance upon the agricultural sector was indicated using the percentage of the labour force employed in agriculture. Data were accessed from the ABS' 2011 Census of Population and Housing (Australian Bureau of Statistics 2011). Geographic remoteness was assessed using the ABS' Remoteness Structure which comprises five categories of relative remoteness: (1) major cities, (2) inner regional Australia, (3) outer regional Australia, (4) remote Australia, and (5) very remote Australia (Pink 2013a). The ABS' IRSAD was used to assess advantage/disadvantage. The IRSAD is a composite index comprising 25 variables that rank areas in terms of relative socioeconomic advantage and disadvantage, which the ABS defines as 'people's access to material and social resources, and their ability to participate in society' (Pink 2013b, p 3). High scores indicate areas with relatively high levels of advantage and relatively low levels of disadvantage. None of the variables included in the IRSAD overlap any of the other indicators used in this vulnerability assessment. Economic diversity was assessed using Hachman Index scores. The Hachman Index is a measure of how closely the employment distribution of the region of interest resembles the distribution of employment in a benchmark region. Hachman scores range from 0 to 1, where the economic diversity of the Australian economy was assumed to be equal to 1. Regional and subregional scores closer to 1 indicate a more diverse economy (Thomsen et al 2012). Employment data were accessed from the ABS 2011 Census of Population and Housing. 'Place of work' data were used, instead of 'usual residence' data, to provide a better reflection of the employment opportunities available in a given area (Australian Bureau of Statistics 2011). Finally, age was assessed using age profiles of the horticultural workforces. Age profiles were preferred to other options such as indicating the percentage of the workforce 65 years or older because it allows for a more nuanced analysis of the ways in which age influences different dimensions of vulnerability. Data were accessed from the ABS 2011 Census of Population and Housing (Australian Bureau of Statistics 2011). The total workforce was organised into two broad categories to differentiate between people who are likely to have decision-making responsibility and the wider workforce: owner/managers and employees. Six age groups were created: 15-24 years, 25-34 years, 35-44 years, 45-54 years, 65 years and older), and then combined into age cohorts broadly consistent with the findings of Clemens et al (2013). Percentages for four age cohorts were calculated: (1) owner/managers 65 years and older; (2) owner/managers aged 25-54 years; (3) employees 65 years and older; and (4) employees aged 25-54 years. Clemens et al (2013) conclude that adults of working age were disproportionately affected by the 2010/11 Queensland natural disasters when compared to older people. They hypothesise that this finding may reflect the 'greater likelihood that this age group participates in the labour force, owns an income-producing property, and is financially responsible for dependents' (Clemens et al 2013, p 554). Here, we use the age cohort 25-54 years to reflect those who are likely to represent this combination of commitments.

Maps
Regional maps for factors 1-4 were created in ArcGIS using geographic and tabular data freely available from the ABS' website. The reliance upon the agricultural sector, and socio-economic disadvantage were represented at the ABS' geographic unit statistical area 1 (SA1). Geographic remoteness was represented using the ABS' remoteness structure. In the case of economic diversity, the ABS' statistical area 2 (SA2) was used because this is the smallest scale at which 'place of work' data are available. The age profiles were presented using subregional aggregations of SA2s. An SA1 is a geographic unit in the ABS' Australian Statistical Geography Standard (effective from July 2011). They represent regions with populations in the range of 200-800 persons. SA2s are the next largest geographic unit, which represents regions with populations in the range of 3000-25 000 persons (Pink 2011).

Data analyses: multiple lines of evidence
Each factor and its corresponding map are considered to be one line of evidence for potential socio-economic vulnerability. Areas in which multiple lines of evidence intersect suggest higher potential vulnerability than areas in which fewer lines intersect. Areas of potential high vulnerability are then compared to the areas that are characterised by high reliance upon the horticultural sector. Reliance upon the horticultural sector is indicated by: (1) percentage of the gross value of horticultural commodities produced; and (2) percentage of the labour force employed in horticulture.
The data for percentage of the gross value of horticultural commodities produced were accessed from the ABS' Agricultural Census 2010-11 and mapped at SA2, the smallest geography at which the data is available. Horticultural commodities were defined by three categories: (1) nurseries, cut flower and cultivated turf; (2) fruit (including tree nuts); and (3) vegetables for human consumption (Australian Bureau of Statistics 2012). The percentage of the labour force employed in horticulture was assessed using employment data from the Census of Population and Housing 2011 and mapped at SA1 (Australian Bureau of Statistics 2011).

The study area: SEQ NRM region
The SEQ NRM region is on the Queensland/New South Wales border and encompasses the city of Brisbane, Queensland's capital and Australia's third largest urban centre. The region is home to approximately 70% of the state's population, but includes only 1.3% (23 200 km 2 ) of its land area (Department of State Development, Infrastructure and Planning 2014). The region's NRM body is SEQ catchments, a not-forprofit organisation.
The vulnerability of this region to climate-related hazards and climate change impacts was highlighted by the intergovernmental panel on climate change that considered SEQ as a 'hot spot' of high vulnerability by 2050 under a medium emissions scenario (Reisinger et al 2014). Current vulnerability was demonstrated by the December 2010/January 2011 floods, which caused flash flooding in many regional communities and Brisbane city, inundating thousands of homes and businesses, and causing 33 fatalities (Queensland Reconstruction Authority 2011).
The SEQ horticultural sector's sensitivity to climate impacts arises from the specific temperature requirements of most crops for the development of optimum yield and quality (Deuter 2008, Deuter et al 2011. The higher temperatures projected for SEQ under several Representative Concentration Pathways (Dowdy et al 2015) will likely impact the timing of crop developmental stages, increase the incidence of sunburn, and alter the prevalence and distribution of pests, weeds and diseases (Deuter 2008, Webb andWhetton 2010). Changed rainfall patterns and increased evaporation are expected to adversely impact soil moisture and runoff which may place pressure on water supplies (Webb and Whetton 2010). In turn, horticultural businesses may need to accommodate increased input costs for fuel, fertilisers and pesticides (Deuter 2008). If these impacts exceed horticulturalists' adaptive capacity at specific locations, horticultural production zones are expected to shift southwards (Deuter 2008, Webb and Whetton 2010).

The SEQ horticultural sector: hotspots of potential vulnerability
In 2011, the SEQ horticultural sector employed 46% of the region's agricultural workforce (4758 people), and generated 54% ($656 million) of the gross VACP in the region in 2010-11 (table 1). The region's horticultural sector employs 37% Queensland's horticultural workforce (10% of the total Australian horticultural workforce) and 29% of the value of horticultural commodities produced in the state (8% of the total Australian value of horticultural commodities). Thus, when compared to the wider Queensland horticultural sector, there is much at stake if the SEQ horticultural sector is adversely impacted by climate change (see supplementary information).
The multiple lines of evidence approach revealed three subregions that were characterised by high potential vulnerability: (1) Southwest, (2) Northeast, and (3) Northwest (figures 2 and 3). When compared to other areas within SEQ, these subregions were generally characterised by higher percentages of the labour forces employed in agriculture (often more than 20%, figure 3, Map a), higher levels of socio-economic disadvantage (often deciles 1 to 2 and 3 to 4, figure 3, Map c), and more specialised economies (often Hachman scores less than 0.40, figure 3, Map d). In addition, the northwest subregion was the most remote area in SEQ (outer regional, figure 3, Map b).
The age profiles of the horticultural workforces also contributed to the potential vulnerability of these three subregions. With the exception of horticultural employees in the Northeast, the three subregions were characterised by horticultural workforces comprising higher percentages of workers aged 65 years or older when compared to the total Australian workforce (table 3). These results are broadly consistent with the wider Australian agricultural sector which tends to have a considerably older age profile than other sectors (Palmer 2012). Figure 4 shows the spatial distribution of SEQ's horticultural sector by the value of horticultural commodities produced in 2010-11 (figure 4, Map e) and the percentage of the labour force employed in horticulture (figure 4, Map f). These maps allow validation of the multiple lines of evidence approach in the context of the wider SEQ horticultural sector and identifies the relative significance of vulnerability among sub-regions. In the Northeast and Southwest subregions, over 1100 people were employed in the horticultural sector (Northeast=1185; Southwest=1120), and up to 33% of the labour forces were employed in the horticultural sector (figure 4, Map f). In the Northwest, while up to 20% of the labour forces were employed in horticulture (figure 4, Map f), only 187 people were employed in horticulture.
The SEQ horticulture sector was similarly skewed towards the Northeast and Southwest when subregions' contributions to the regional value of horticultural commodities produced is examined ( figure 4, Map e). The Southwest contributed the largest percentage towards the regional value of horticultural commodities (approximately 33%), followed by the Northeast's contribution of approximately 20%. In contrast, the Northwest subregion contributed around 4% to SEQ's total value of horticultural production. Thus, from a regional perspective, the horticultural sectors in the Northeast and Southwest are of greater significance than the Northwest.
Focusing upon the two subregions in which high potential vulnerability coincided with high reliance upon the horticultural sector-the Northeast and Southwest-differences in the ways potential vulnerability is constituted in each subregion are observable. First, the larger contribution to SEQ's value of horticultural commodities from the Southwest suggests that from an economic perspective there was more at stake in the Southwest than Northeast if the horticultural sector was adversely affected by climate change. Second, the differing age profiles of the horticultural workforces in these subregions reveals that there were higher percentages of owner/managers aged 65 years or older, employees aged 65 years or older, and owner/managers aged 25-54 years in the Southwest when compared to the Northeast, which also suggests relative higher potential vulnerability in the Southwest (table 3).

Towards rapid, regionally-relevant vulnerability assessments
There are three key ways in which our approach advances knowledge about climate change socioeconomic vulnerability assessments. First, increased commitment to regional modes of policy and planning in developed countries means that tools and techniques designed to deliver information at this scale are pertinent to the effectiveness of climate change adaptation decision-making (Measham et al 2011). Thus, our approach specifically focuses upon regional-scale assessments to provide regionally-relevant information to organisations operating at the same scale. However, our approach goes beyond simply matching the scale of analysis to the jurisdictions of administrative organisations responsible for climate change adaptation policies and practices. The regional relevance of our approach is enhanced through the inclusion of techniques that target regionally important sectors and identify place-based factors associated with vulnerability.
Second, in our approach the factors used to assess socio-economic vulnerability are selected based upon empirical, place-based research examining the dynamics of vulnerability. From a conceptual perspective, this process more explicitly aligns with understandings that vulnerability to drivers operating at a range of scales is influenced by the set of social, economic and environmental conditions in particular places (Turner et al 2003). Two benefits arise from selecting factors in this way: (1) each factor has direct relevance to the region of interest; and (2) end-users obtain guidance about how the factors should be interpreted within the context of regional conditions (table 2).
One potential limitation is the small number of socio-economic factors that were included in this assessment, and consequently we may have omitted important determinants of vulnerability. This feature of the approach reflects the dearth of place-based research relevant to our area of interest (Smith et al 2015), rather than the unavailability of suitable secondary data upon which vulnerability assessments could be made. These features of the research bases in other regions/countries may differ, leading to the inclusion of a larger number of factors, which will strengthen applications of the approach in other contexts. In the case of Queensland, as the place-based vulnerability literature expands, additional factors may be incorporated to enhance examinations of socio-economic vulnerability.
While secondary data for a wider range of factors exists, comparing multiple time periods is currently more problematic. A fundamental change to the ABS' geographical framework introduced in 2011 hinders straight-forward comparisons with earlier time periods and, therefore, also hinders projections of how indicators may change in the future. Assessment of how individual factors have and may change over time coupled with climate change scenarios represent useful avenues to extend this multiple lines of evidence approach in future applications.
Third, the SEQ findings demonstrate that a small number of factors represented spatially (and based upon contextually relevant empirical evidence) offers a nuanced assessment process capable of identifying Decile 1 indicates higher levels of disadvantage and, therefore higher potential vulnerability; decile 10 indicates higher levels of advantage and, therefore, lower potential vulnerability. Data compiled at 2011 ABS SA1. Map d: economic diversity: Hachman Index. Light red indicates less economically diverse and, therefore, higher potential vulnerability; dark red indicates more economically diverse and, therefore, lower potential vulnerability. Data compiled at 2011 ABS SA2 using place of work data. Employees 25-54 years 63% (4% lower than Australian workforce) 65% (2% lower than Australian workforce) 62% (5% lower than Australian workforce) 1. Southwest 2. Northeast 3. Northwest intra-regional hotspots of potential vulnerability. Once subregions are identified, retaining disaggregated maps allows for closer inspection of the differences (and similarities) in potential vulnerability between them. In turn, adaptation actions may be prioritised. Alternatively, the rapid identification of vulnerability hotspots provides a sound rationale for more detailed, contextual analyses of the place-based dynamics of vulnerability at smaller scales.

Conclusions
Our research demonstrates that rapid, regionallyrelevant socio-economic vulnerability assessments are feasible using existing datasets to target important economic sectors that are likely to be sensitive to climate change impacts. The approach appeals for the following reasons. First, it is comparatively simple when compared to a wide range of other approaches for assessing socio-economic vulnerability, which are often time and resource intensive. Thus, it enables rapid assessments of spatial heterogeneity and the underlying factors that contribute to potential vulnerability-information that may be used for regionallevel decision-making and priority setting. Second, it appeals because of its reliance upon freely available, national datasets that can be easily customised to regional scales, rendering the approach highly transferrable across multiple regions and time periods. Thus, the approach is more accessible across organisations with differing resources and capacities.
Finally, utilising the concept of resource dependency as an assessment-framing device offers scope to extend the approach beyond the agricultural sector to other resource dependent sectors (e.g., tourism, mining, outdoor recreation). Investigation of the ways in which resource dependency might be used to integrate vulnerability assessments conducted at broader and finer scales also deserves researchers' attention.