1 Introduction

Drought, defined as a prolonged period of abnormally dry conditions when the amount of water available is insufficient to meet the usual needs of local socio-ecological systems (Wilhite and Glantz 1985; Mishra and Singh 2010), is a severe natural hazard in Australia (Van Dijk et al. 2013; Kirono et al. 2020) likely to be exacerbated by ongoing global climate change (Vicente-Serrano et al. 2020). Droughts are associated with financial hardship (Kingwell and Payne 2015; Hughes et al. 2020, 2022b) and irreversible damage to soils and vegetation (Tozer and Leys 2013; Auld et al. 2015). Social impacts of drought can be devastating and include permanent loss of services and employment opportunities in regional areas as well as long-lasting adverse impacts on physical and mental health (Alston and Kent 2004; Carnie et al. 2011; Hanigan et al. 2012). Drought adds weight to the existing personal and professional burdens experienced by regional communities, such as poor health, isolation and limited access to services and infrastructure (Kiem and Austin 2013). As such, regional communities are at high risk of experiencing adverse impacts of drought, particularly those communities that depend primarily on agricultural production for their livelihoods (Wittwer et al. 2002; Kingwell and Xayavong 2017; Kingwell et al. 2020). Droughts threaten agricultural production across the board (Sudmeyer et al. 2016), compromising yields or causing failed production in both cropping (Ray et al. 2018; Wang et al. 2018; Fletcher et al. 2020; Cohen et al. 2021) and livestock systems (Henry et al. 2018; Bowen and Chudleigh 2021). With droughts expected to become more frequent and more severe as climate change progresses (Iacona et al. 2018; Mukherjee et al. 2018; Vincente-Serrano et al. 2020), local-level planning for drought resilience is becoming increasingly important (Sugg et al. 2020).

Context-specific information at the scale of local-level decision-making units, e.g. local government authorities (de Sherbinin 2014), is needed to guide effective decision-making, planning and implementation for drought resilience. Many existing tools and resources provide climate change information to support decision-making (Aldum et al. 2014; Born et al. 2021; Stenekes and Kancans 2022). These tools frequently assume that data are available in accessible and understandable formats and that decision-makers are confident in interpreting such information, which may not always be the case (Uzzell 2000; Nerlich et al. 2010). Additionally, tools that only focus on synthesising climate data are often insufficient for supporting local-level planning and implementation. Many climate tools are not spatially explicit and are limited in their integration of socio-economic and environmental factors along with the climatic aspects (Pasquini and Cowling 2015; Bourne et al. 2017; Jones et al. 2020). Mainstreaming drought resilience at the regional level is greatly enhanced by (1) including relevant social, economic and environmental data alongside climate data and (2) producing spatial decision support tools at the scale of the known and familiar administrative boundary (Aldum et al. 2014; Bourne et al. 2016; Kasecker et al. 2018; Cardoso da Silva et al. 2020). Assessing drought resilience at administrative scales better aligns with decision-makers’ mandates, even though social and ecological processes routinely transcend administrative boundaries (Egoh et al. 2008; Hamann et al. 2015).

Spatial multi-criteria analysis (MCA) is a systematic, GIS-based planning approach that has been used to develop integrated decision support tools for land management and climate change adaptation planning worldwide (Joerin and Musy 2000; Malczewski 2006; Chandio et al. 2013). Spatial MCA tools are one way of facilitating consideration of multiple criteria by decision-makers and helping with stakeholder engagement (Ryan and Nimick 2019). The method involves transforming and combining different spatial datasets to produce a single, aggregated, multidimensional map (Chakhar and Mousseau 2008; Malczewski 2018; Nautiyal and Goel 2021). In South Africa, MCA has been used to identify regional priority areas for the implementation of ecosystem-based climate change adaptation measures (Bourne et al. 2016), promote more targeted investment in socio-ecological sustainability (Hamann et al. 2015) and inform systematic biodiversity conservation planning at the national level (Holness and Biggs 2011). Spatial MCA has also been used to decide where to locate landfills in Nigeria (Ajibade et al. 2019) and to support land use planning in the Lake Erhai basin in China (Wang et al. 2004) and urban green space planning in Hanoi, Vietnam (Uy and Nakagoshi 2008). In WA, MCA has been used to measure drought risk in the agriculture sector (Hughes et al. 2020), identify priority areas for carbon planting (Ecotones and Associates 2014) and map climate change adaptation corridors (Ecotones and Associates 2015).

Index-based and machine learning approaches to drought risk assessment are also common worldwide (Zargar et al. 2011, Nasrollahi et al. 2018, Mukherjee et al. 2018, Yihdego et al. 2019). In Australia, the Australian Bureau of Agricultural and Resource Economics (ABARES) has quantified and mapped drought risk, sensitivity and exposure across Australia using an index approach (Stenekes et al. 2012; Hughes et al. 2020). The outcomes of this work indicated that the northern regions of WA Wheatbelt and the temperate, inland regions of Eastern Australia were at the highest risk of drought impacts (Hughes et al. 2020). Similarly, index-based spatial modelling approaches have been used to assess drought risk in New South Wales (Hoque et al. 2021a), Queensland (Hoque et al. 2021b) and the Murray–Darling Basin (Dunne and Kuleshov 2023). The methods used by ABARES (e.g. the farmpredict model) and others are broadly similar to MCA methodology in terms of input requirements and data visualisation, but ABARES’ work additionally assesses risk as a function of reduction in farm profits (Hughes et al. 2019) and uses machine learning methods for modelling (Hughes et al. 2022a). In a recent study, ABARES combined agriculture exposure and sensitivity data (at the farm enterprise level) with data representing community sensitivity (agricultural dependence of a community) to produce an index of ‘potential drought impact’ that can be applied consistently at the local government area (LGA) level across Australia (Stenekes and Kancans 2022). Indicators of community adaptive capacity are to be included in the second phase, which had not been published at the time of writing.

Here, we use MCA to generate a suite of maps that address climate change, the environment and local government priorities for infrastructure and community development. The simpler MCA method was chosen over a potentially more nuanced or generalisable index-based approach in order to enhance stakeholder engagement with the input variables and resulting decision support products. We follow a conceptual framework based on best practice in global resilience planning and climate change vulnerability assessment (Ford and Smit 2004; Adger 2006; Biggs et al. 2015; Woodruff et al. 2018) and incorporate exposure, sensitivity and adaptive capacity data to the extent possible. The resultant maps systematically combine climate and agricultural production data with socio-economic priorities for local governments into spatial drought impact priority areas, providing a means of identifying locations for the implementation of drought resilience programmes to maximum effect where resources are limited. In this study, we have focused on the wheat-growing areas of southwest Western Australia (WA). This region is home to the WA grains industry, the largest agricultural contributor to the economy and one of the regions most impacted by climate change in Australia to date, experiencing consistent reduction in rainfall and recurrent drought over the last several decades (Dey et al. 2019; McFarlane et al. 2020; Scanlon and Doncon 2020). The work presented here was undertaken as part of the larger Future Drought Fund Regional Drought Resilience Planning (RDRP) programme in southwest WA (DPIRD 2021) and contributes towards Australia-wide work on this topic through the federal Future Drought Fund. Stakeholder engagement to sense-check spatial priorities identified through desktop MCA focused on three sub-regions in the Great Southern, Mid-West and Wheatbelt and included review workshops and participatory mapping processes.

2 Methods

Analyses and stakeholder engagement were conducted in 2021 and 2022. The spatial MCA method was selected over other index-based and machine learning approaches commonly used in Australia (Hughes et al. 2020, 2022a; Stenekes and Kancans 2022) because it is transparent, accessible and understandable for local decision-makers and farmers (Bourne et al. 2016). As the purpose of this study was to produce a community-informed decision support tool, we preferred a simple MCA approach that local stakeholders could feel confident in assessing and interpreting. Data analysis for the MCA focused on the Australian Bureau of Meteorology’s South West Land Division forecasting area. Within this area, targeted stakeholder engagement was conducted in the three RDRP focus sub-regions in WA, namely the Great Southern, Mid-West and Wheatbelt (Fig. 1). The Great Southern consortium focused on the Shires of Broomehill–Tambellup, Cranbrook, Gnowangerup, Jerramungup, Katanning, Kent, Kojonup and Woodanilling; the Mid-West consortium focused on the City of Greater Geraldton and the Shires of Northampton and Chapman Valley; and the Wheatbelt consortium focused on the Shires of Dumbleyung, Kondinin, Kulin, Lake Grace and Wagin. Together, the three sub-regions cover over 670,000 km2 and represent much of WA’s agricultural value.

Fig. 1
figure 1

Map of the study area showing the southwest land division forecasting area and the three focus sub-regions

Full size image

2.1 Assessing vulnerability to drought

An important step in drought resilience planning is vulnerability (risk) assessment (Ford and Smit 2004). We used the Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report’s definition of vulnerability (Parry et al. 2007; Field et al. 2014) as a function of exposure, sensitivity and adaptive capacity. The most vulnerable areas to drought are those where biophysical exposure (the nature and degree to which a system experiences drought) and sensitivity of regional communities (the degree to which human populations are affected by the impacts of drought) are highest; and where adaptive capacity (the ability of regional communities to adjust or accommodate drought behaviourally and/or technologically) is lowest (Jones et al. 2020). We developed a drought vulnerability assessment conceptual framework to guide this analysis (Fig. 2) and other work under the WA RDRP, drawing on past studies in Australia (Stenekes et al. 2012; Rickards 2013; Duxbury and Hodgson 2014; Hughes et al. 2020) and around the world (Nelson et al. 2007a; Bourne et al. 2016; Kasecker et al. 2018; Jones et al. 2020). The term ‘drought’ refers to a prolonged, abnormally dry period when the amount of available water is insufficient to meet our normal use (Mishra and Singh 2010, BOM 2022). Given the importance of rainfall in April to October for grain production in southwest WA, meteorological definitions of drought included here focused on total rainfall during the April to October growing season (Scanlon and Doncon 2020). For direct and proxy measures of drought sensitivity, we assumed that on-farm production per unit area correlates with drought severity, resulting in higher levels of sensitivity in historically marginal production areas (Dietz et al. 2021; Lawes and Kingwell 2012; Fletcher et al. 2020). Considering exposure, sensitivity and adaptive capacity together ensures that sufficient attention is paid to all important aspects of drought vulnerability and resilience in southwest WA.

Fig. 2
figure 2

Vulnerability assessment conceptual framework used to guide the GIS-based multi-criteria analysis for drought risk

Full size image

The inclusion of a consolidated impact area combining exposure and sensitivity follows the approach taken by ABARES (Hughes et al. 2020) and is utilised here for consistency across all components of a larger programme being implemented by the Australian Government’s Future Drought Fund. All data included in the MCA were categorised under exposure, sensitivity or adaptive capacity. Datasets followed those used in other Future Drought Fund projects where possible. For example, the measures for percentage change in rainfall and temperature were taken from the Future Drought Fund Climate Services for Agriculture project (DAFF 2023).

2.2 Creating the drought vulnerability priority areas map

The final drought resilience priority areas map is made up of a set of composite maps for (i) exposure, (ii) sensitivity, (iii) impact (combining exposure and sensitivity) and (iv) adaptive capacity, following the conceptual framework described above. The analysis was performed using the Multi-Criteria Analysis Shell for Spatial Decision Support (MCAS-S) tool developed by ABARES (MCAS-S Development Partnership 2018). The MCAS-S tool is a free desktop software for viewing and assessing mapped information. Priority areas, where vulnerability to drought is highest, lie at the intersection of all categories, where exposure, sensitivity and adaptive capacity overlap. The overlapping areas highlight where droughts are likely to occur most frequently (exposure) and have the largest impact on water resources and agricultural production (sensitivity). They also identify locations where regional communities may be more vulnerable to the impacts of drought due to socio-economic factors including relative remoteness, access to infrastructure and income (adaptive capacity).

All spatial data were prepared as layers for MCAS-S analysis using ArcGIS Pro 2.9.2 GIS software. Layers were classified into five classes. Lowest risk areas were assigned to Class 1 and highest risk areas were assigned to Class 5. Values were considered high risk when they contributed most to drought risk in terms of exposure and sensitivity or contributed least to adaptive capacity. Roughly equal intervals were used to portion the data for approximately half of the layers (23/44). Where relevant (n = 12), data were classed by presence or absence (e.g. whether or not a river or stream was present) or some other category (e.g. degree of accessibility and remoteness). For the remaining layers, data were binned after examining a frequency distribution graph and selecting suitable break points for an even or otherwise sensible distribution of the data. This was most commonly required for metrics with very large variation in values across the region (e.g. sheep and cattle numbers per LGA, soil salinity). Details are provided in the Online Resources (Table S1).

Reclassified input layers were exported to a standard geoTIFF raster format with a consistent scale (statewide 1 km grid) and coordinate system (EPSG:4283—GDA94) ready for analysis. The geoTIFF input layers were loaded into the MCAS-S software and combined to produce a series of composite maps. Standardised colour gradients were used throughout, with Class 1 (low risk/high adaptive capacity) shown in green and Class 5 (high risk/low adaptive capacity) in red. The final composite drought vulnerability map incorporates ten composite maps made up of 44 discrete layers.

The authors recognise that quite different data layers have been combined, but argue that all are directly relevant for identifying priority areas for drought risk and implementing drought resilience programmes. Care was taken to ensure that the data ranges and priority levels represented in the different summary layers were sufficiently similar to allow them to be sensibly combined, with each layer representing a consistent range from lowest priority through to highest priority for identifying drought risk.

Discrete data layers and composite maps were weighted according to their likely influence on drought resilience. Data layers and their respective weightings were identified initially by the authors following a review of the available published and grey literature and consultation with experts within local and state government. The list of data to include was then adjusted based on data availability and feedback from regional stakeholders during six review workshops, one cross-regional technical working group meeting and a series of project steering committee meetings. Review workshops were attended by 39 regional stakeholders representing 25 different organisations, including local government authorities, producer groups, agribusiness, government, NGOs and research institutions. The cross-regional technical working group was made up of 22 experts in the agricultural, climate, environmental and social sciences as well as in economics, mental health, disaster management and rural leadership. A project steering committee comprised of the authors and local government and regional development commission representatives from each of the three focus sub-regions met every six weeks between September 2021 and July 2022 to review drought risk assessment content. The data layers included in the drought resilience priority areas map, and their relative weightings, are described in Tables 1, 2 and 3. Detailed metadata descriptions, data processing notes and rationale for applied weighting for each data layer are provided in the Online Resources (Table S1). Most input data layers (33/44) were weighted 1 (neutral weight). Important water resources (composed of various data layers—see Online Resources Table S1) were weighted lower than the others (0.5) because natural water resources in the region are typically ephemeral, frequently saline and unlikely to contribute much to drought resilience as a result. The remaining four data layers were weighted higher than the others due to their relative importance. Average potential wheat yield received a higher weighting (1.5) due to the primary economic importance of wheat production in the agriculture sector in the focus regions. NDVI at the end of the growing season (20 year average and lowest 10%) received a higher weighting (1.5) due the detectability of drought impacts in NDVI data and the importance of vegetation cover for mitigating impacts of drought such as soil erosion. Finally, areas declared water deficient received a higher weighting (1.25) due to the importance of available water supplies for mitigating impacts of drought on livestock and on farming communities and the fact that water scarcity was identified as a high priority for regional stakeholders, particularly in the Great Southern and Wheatbelt sub-regions.

Table 1 Datasets contributing to the ‘exposure’ component, including all data sources
Full size table
Table 2 Datasets contributing to the ‘sensitivity’ component, including all data sources
Full size table
Table 3 Datasets contributing to the ‘adaptive capacity’ component, including all data sources
Full size table

Composite maps for exposure, sensitivity, impact and adaptive capacity contributing to drought risk and resilience were produced. Exposure was weighted higher than sensitivity (1.25:1) due to the importance of weather variables in determining drought risk. Exposure and sensitivity were combined to create the impact map, which was then weighted higher than adaptive capacity (2:1) in the final analysis due to higher confidence in both the component datasets and the relationship between included datasets and the impacts of drought. The impact and adaptive capacity maps were combined into a single composite map. Relative weightings of component maps were agreed through sense-checking results with local and regional experts and community members during the stakeholder consultation process described above. Priority areas for implementing drought resilience programmes lie in the overlap between impact and adaptive capacity in the final composite drought priority areas map. Data outside the South West Land Division forecasting area are often unreliable, absent or incomplete because these areas are outside of the main agricultural crop production zone in WA. As a result, comparable data are often collected sporadically or not at all, and to a much lower resolution, and should be ignored for the purpose of this analysis. These areas appear as faded on the maps that follow and in the Online Resources.

2.3 Checking the priority areas maps against community expectations

We conducted participatory mapping exercises (Cadag and Gaillard 2012; Fagerholm et al. 2021) during three workshops with 77 people in the Mid-West focus sub-region. These people represented 52 businesses and organisations, including Aboriginal community groups, broadacre farmers, local government and agribusiness. During workshops, participants were asked to refer to large printed topographical maps of their area and identify the locations they expected would be most vulnerable to drought in the future based on their past experiences with drought in the region. The maps created by workshop participants were converted into a data layer on the same scale as all other included data and compared with the final drought priority areas map.

3 Results

3.1 Drought priority areas map

The exposure map included datasets related to changes in growing season rainfall (historical and projected), maturing season temperature (historical and projected), average potential wheat yield (historical), timing of the autumn break (historical) and drought frequency (historical) as well as records of rainfall intensity (number of rainfall events > 10 mm and > 25 mm), root zone soil moisture and distance from the coast (Table 1). The sensitivity map included a suite of production-related datasets as well as data related to farm business characteristics, water quality and availability and relative remoteness (Table 2). The combined impact map shows that the areas most exposed and most sensitive to drought are the northern and eastern Wheatbelt (Fig. 3).

Fig. 3
figure 3

Drought priority areas map showing all the contributing datasets, outlining the workflow followed for combining all datasets. The workflow follows the overall Regional Drought Resilience Planning programme conceptual framework. Component maps are sized according to their weighting within each tier of analysis; primary = all individual component datasets and first-level composite maps, secondary = second-level composite maps made up of primary datasets and first-level composite maps, tertiary = composite maps made up of second-level composite maps, final = the final analysis drought vulnerability priority areas map made up of the tertiary maps

Full size image

The adaptive capacity map is made up of three composite maps for human capital, natural capital and physical capital (Table 3). The human capital map includes data on population age and employment. The natural capital map includes composite data related to water, soils and vegetation. The physical capital map includes data on proximity to road and water infrastructure. Areas with the lowest adaptive capacity to drought follow a less clear pattern and tend to be concentrated in the more populated areas along the coast. The Mid-West region is likely to be more vulnerable to drought from the perspective of adaptive capacity than the Wheatbelt and Great Southern sub-regions (Fig. 3).

The final drought priority areas map highlights the most vulnerable areas to drought and is composed from the two composite maps for impact and adaptive capacity. The drought priority areas map highlights locations in the landscape that are more likely to experience drought, are more sensitive to the impacts of drought and may have a lower ability to respond effectively during drought due to, for example, relatively poorer access to services and infrastructure, higher dependence on agriculture or relatively lower levels of income and employment. Areas at highest risk from drought overall are the northern and eastern Wheatbelt (Figs. 3, 4). Higher-resolution images and closer views of each component workflow are provided in the Online Resources.

Fig. 4
figure 4

Final drought risk priority areas map highlighting that the areas most vulnerable to drought are the northern and eastern Wheatbelt

Full size image

3.2 Checking the priority areas maps against community expectations

Workshop participants in the Mid-West region identified drought vulnerable locations in the Southern Rangelands, the areas north of Ogilvie and east of Tenindewa and the areas on poor soils. Inland northern Wheatbelt towns of Binnu and Yuna were specifically mentioned more than once, due to marginal soils, high exposure to hot and dry conditions and the relative remoteness of these communities. Participants also mentioned that the central and coastal grainbelt regions may be vulnerable to drought because they have not experienced it as often or as severely in the past and therefore might be less prepared. Overall, the area identified as most at risk from drought through participatory mapping (Fig. 5) was the northern and eastern Wheatbelt.

Fig. 5
figure 5

Results of participatory mapping in the Mid-West sub-region. Areas identified as most vulnerable to drought, based on regional expertise and past experience, were the northern and eastern Wheatbelt. Photograph by Annabelle Garratt, NACC NRM, Geraldton

Full size image

4 Discussion

We used MCA to systematically prioritise focus areas for implementing drought resilience programmes, highlighting locations for decision-makers that are likely to be worst affected by drought and benefit most from investment in resilience-building activities. In southwest WA, these areas are the northern and eastern Wheatbelt. We found substantial spatial agreement between the areas identified through desktop MCA and those identified by growers, community leaders and local and state government officials in the region.

We collated data showing southwest Western Australia’s highly variable climate and the consistent drying and warming trend experienced in the region since the 1970s, a result consistent with previous research (Dey et al. 2019; McFarlane et al. 2020; Scanlon and Doncon 2020). Our data support the conclusion that drought is a potentially serious natural hazard in the region (Van Dijk et al. 2013; Kirono et al. 2020), with potentially long-lasting social (Carnie et al. 2011; Hanigan et al. 2012), economic (Ding et al. 2011; Kingwell and Xayavong 2017; Edwards et al. 2018) and environmental (Tozer and Leys 2013; Auld et al. 2015; Jiao et al. 2020) consequences, particularly for rural and regional communities dependent on agriculture for their livelihoods (Wittwer et al. 2002; Nelson et al. 2007c; Kiem and Austin 2013; Hughes et al. 2022b). While the agriculture sector in southwest WA has done much to adapt to the impacts of climate change already (Morgan et al. 2008; Fletcher et al. 2020), drought resilience planning, supportive policies and timely intervention at the local level are important to ensure that regions are prepared for and able to recover from drought. Supporting such planning is a major focus of the state and federal government in Australia (Mukheibir et al. 2013, Botterill 2021, DPIRD 2021, Stenekes and Kancans 2022). Climate change mapping, spatially identifying regions most vulnerable to climate impacts, allows for targeted adaptation efforts and is a highly effective and widely used decision support system (de Sherbinin 2014; Bourne et al. 2016). Spatial prioritisation through MCA is useful to inform where and how support, incentives and investments can be provided by governments and communities. Thus, the results of this work can aid in ex-ante decision-making concerning the impacts of future droughts and improving the resilience and awareness of at-risk communities in WA.

We have a high level of confidence in the component datasets used for the impact part of the analysis. These datasets include climate data, agricultural production data and data relating to water quality and availability, levels of engagement in agriculture and economic and population trends in regional communities. These data are robust and regularly collected across the wheat-growing areas of southwest WA, including all three focus regions. The links between the component datasets and the impacts of drought are well understood. Low rainfall exacerbates drought risk (Van Loon 2015; Scanlon and Doncon 2020; Chancellor et al. 2021), as does a shorter or warmer growing season (Wang et al. 2018; Fletcher et al. 2020). Further reductions in rainfall and increases in temperature as a result of future climate change are likely to further exacerbate drought risk in marginal areas (Ludwig and Asseng 2006; Chen et al. 2020), and future climate projections are considered highly relevant for drought resilience planning in Australia (DAFF 2023). Moreover, drought is associated with reduced production and compromised farm incomes (Kingwell and Xayavong 2017; Edwards et al. 2018; Hughes et al. 2022a). Farming communities that depend directly on the land and the weather for their livelihoods are highly sensitive to drought (Archer et al. 2008; Kiem and Austin 2013; Carrão et al. 2016), and problems with water supply or water quality can leave regional communities more vulnerable to the effects of drought (Ali et al. 2012; Barron et al. 2012; McFarlane et al. 2012, 2020; Silberstein et al. 2012). The included datasets align well with regional communities’ perceptions of how they are affected by drought, identified through consultation workshops. Specifically, regional communities identified low rainfall, high temperatures, compromised agricultural production, financial stress and shrinking regional communities as key components of drought risk. There is a high level of agreement between the priority areas identified through this MCA, highlighted in the drought vulnerability map, and those identified through prior research (Hughes et al 2020) and a participatory mapping process involving a large number of regional stakeholders, many of whom have first-hand experience of historical drought in the region. Non-overlap areas to the north and east of the South West Land Division forecasting area can be explained because residents in the region identified these areas as being at high risk from drought based on their lived experience and knowledge of the region, but these areas are data-deficient in the MCA.

It must, however, be noted that many of the exposure and sensitivity measures included here are proxies of drought risk, rather than direct measures of drought impact. The inclusion of more direct measures of drought impact in future could improve the quality of the risk maps resulting from this analysis. Additionally, the direction of effect for drought risk is not always clear. For example, we have assumed in this analysis that lower-yielding, less productive areas will be more vulnerable to drought in the future (Archer et al. 2008; Chancellor et al. 2021; Hughes et al. 2022a), because of the devastating effects of past drought in the region. However, evidence to date of negative effects on wheat production in the more marginal areas of the southwest WA is equivocal (Fletcher et al. 2020), due largely to increases in efficiency and advances in plant breeding over time, and there are some cases globally where increased production is associated with reduced resilience (Lobell et al. 2014).

Including data layers representing adaptive capacity is important, because socio-economic, natural and physical characteristics of regions and regional communities all contribute substantially to overall resilience (Nelson et al. 2007a, b, c; Carnie et al. 2011; Munroe et al. 2012; Bourne et al. 2016; Edwards et al. 2018; Jones 2020; Kingwell et al. 2020). Important human capital components of adaptive capacity include the ability of individuals to access and use information to guide decision-making (Pasquini and Cowling 2015; Baudoin and Ziervogel 2016; Bourne et al. 2017), make effective decisions under conditions of uncertainty (Kingwell and Xayavong 2017; Caball and Malekpour 2019; Marchau et al. 2019; Rose et al. 2019; Kingwell et al. 2020; Phillips-Wren and Adya 2020), manage their finances and maintain employment (Edwards et al. 2018), engage with experts and in education and training and adopt new approaches, tools and technologies (Coelli 2021). Important physical capital components of adaptive capacity include mobile phone and internet connectivity (Riddlesden & Singleton 2014; Park 2017; Philip and Williams 2019), capital expenditure and access to infrastructure and equipment (Alston 2007; Bowen and Chudleigh 2021). Important financial capital components of adaptive capacity include access to savings or credit, earnings before income and tax (EBIT), the number of agricultural businesses and the diversity of agricultural commodities (Chancellor et al. 2021; Hughes et al. 2022b, a). Important natural capital components of adaptive capacity include soil health and capability, water availability, vegetation cover and native vegetation extent (Lavorel et al. 2015, 2019; Colloff et al. 2016; Jiao et al. 2020). Finally, important social capital components of adaptive capacity include membership of groups or clubs, levels of participation in community events, relative remoteness, population trends and social media activity (Paton & Johnston 2001; Magis 2010; Carnie et al. 2011; Hanigan et al. 2012; Casey et al. 2022). Keeping adequate feed and water reserves on hand and ensuring continuous ground cover through stubble retention or sowing cover crops are also critical (Coelli 2021). Where possible, these components have been included in the MCA, see Table 3. However, they can be difficult to measure (Nelson et al. 2007b; Brown et al. 2010; Engle 2011; Lockwood et al. 2015). Many are reported infrequently, if at all, and often only at small, local scales. They can be difficult to represent spatially and therefore to include in MCA. Our inability to include a large number of the adaptive capacity components identified as important in the literature and by regional stakeholders, due to limitations on data quality, resolution or availability, reduces our confidence in the adaptive capacity component of the MCA. It is likely to be incomplete.

A second feature of the data that reduces our confidence in the adaptive capacity component of the MCA is that the direction of effect of some components is ambiguous and our assumptions about the direction of effect may not hold for all included components in all regions or situations. For example, we have included the percentage of the population aged over 65 years in the human capital component, assuming that communities with an ageing population are more vulnerable (Filiberto et al. 2009; Polain et al. 2011; Berman et al. 2017; Rich et al. 2018). However, the experience that comes with age could be an advantage when preparing for and responding to drought (Frumkin et al. 2012). While all analyses could be improved with better resolution data, several of the datasets included for adaptive capacity are only available at the scale of the LGA boundary. This reduces the extent to which such datasets can contribute to fine-scale, local-level interpretation and inform decision-making.

The structure of the overall RDRP programme conceptual framework gives a larger relative influence to the adaptive capacity component than the exposure and sensitivity components. This is because, for consistency with other Australian drought vulnerability assessment frameworks (Hughes et al. 2020), exposure and sensitivity are first combined into a composite impact map before being combined with adaptive capacity to determine vulnerability. Given our higher confidence in the data layers included in the exposure and sensitivity components, and the importance of these components for understanding drought risk, we did not weight impact and adaptive capacity equally to generate the final drought vulnerability map. Instead, we weighted the impact component, in which we have more confidence, higher than the adaptive capacity component, in which we have less confidence. Adjusting the weighting aids interpretation of the resulting priority areas by allowing for the inclusion of all components while also accounting for differences in data quality, confidence and completeness in the final product.

The composite drought risk priority area map has the potential to be a powerful decision support tool for regional southwest WA. It clearly shows drought risk priority areas where exposure and sensitivity to drought are highest and adaptive capacity is likely to be lowest. Local authorities need information at the scale of the political and administrative units they manage (Ahmadalipour and Moradkhani 2018). There can be limited ability or willingness to take up information when it does not align well with political or administrative boundaries, priorities or mandates (Pasquini and Cowling 2015; Bourne et al. 2017). Further limitations occur when the available information does not cover the full area of responsibility and hence needs to be integrated with other data sources to provide a complete picture, or where the information extends beyond their jurisdiction and only a portion is relevant for the decision-maker (Baudoin and Ziervogel 2016). Presenting information at the scale of the sub-national administrative unit enables direct embedding of the priority areas identified into wider government and institutional processes (Ayers et al. 2014; Reid 2015; Bourne et al. 2016; Carrão et al. 2016). Local government authorities are the entities most directly responsible for local-level planning, management and implementation of drought resilience activities (Huq et al. 2007; Measham et al. 2011; Mukheibir et al. 2013; Wamsler et al. 2014). We argue that spatial products such as this drought risk priority areas map can provide significant support to decision-makers by collating complex climate, ecological and socio-economic information into a single image or series of images.

The products presented here provide detailed information at the practical scale of the familiar administrative boundaries within which local government officials and other regional stakeholders operate. The products were developed together with farmers and local government representatives. The methodological approach for identifying priority areas involves overlaying environmental and demographic data using standard multi-variate GIS analyses and publicly available datasets. Using industry standard software, publicly available data and straightforward methodological approaches allows for transparency, easy updating, revision, refinement and replication (Bourne et al. 2016). Most data used in the drought resilience priority areas analysis are publicly available at the national level in Australia or for southwest WA and readily applied to all 113 LGAs in the study area. For replication outside of southwest WA, the broad categories of data included under the conceptual framework, for exposure, sensitivity and adaptive capacity, could be used with adjustments to account for location-specific variations in socio-ecological systems, regional priorities and data availability. Additionally, given the large overlap between the identified areas at high risk from drought and low-rainfall zones, it is possible that low rainfall could be used as a proxy to guide decision-making in areas where these analyses cannot be completed (noting that a simplified analysis would lack nuance in terms of adaptive capacity particularly). Future studies could also consider including other more direct measures of drought exposure and sensitivity than the proxy measures used here. Where those data are readily available, they may include measured variation in production or on-farm incomes between drought and non-drought years (e.g. Hughes et al. 2022b).

5 Conclusion

The complex and multidimensional nature of drought requires the development of an integrated approach that focuses on the governments’ collaboration with different stakeholders (Mukheibir et al. 2013; Azadi et al. 2018). Adopting an integrated approach can improve drought risk management, decrease vulnerability and build resilience (Azadi et al. 2018). As a next step, spatially defined priority areas need to be linked explicitly with clear, site-specific implementation activities. Participatory and stakeholder-engaged local work is needed to identify the specific drought resilience actions to implement, when these are the most appropriate actions and to what extent they can be effective given the actual or potential limits of regional natural and social systems in terms of the extent of drought they can withstand (Jaeger et al. 2019; Vicente-Serrano et al. 2020). While the individual data layers may be useful on their own, the composite drought risk priority areas map should be used as the primary decision support tool from this MCA.