3.1 Intensive rapid onset hazards

The most prominent hazards are intensive and rapid onset hazards, such as cyclones, tsunami’s, major floods, and major earthquakes, which are characterized by their low frequency but large magnitude [53]. While the impacts of intensive rapid onset events affect all socio-economic groups, there is broad scientific evidence that shows that vulnerable socio-economic groups tend to live in higher risk areas (e.g. cyclones [26]; tsunami [15,54]; major floods [3,25]; major earthquakes [27,55]), have higher mortality rates [26,54,56,57], and rely more on ecosystem dependent sectors such as agriculture that are heavily impacted by large-scale events [10,11,58]. The magnitude and rapid onset of these hazards leads to an increased tension between doing what is urgently needed to prevent wide-scale devastation, and upholding distributive, procedural, and systemic equitability [16].

Despite the need for equitable adaptation, adaptation towards intensive rapid onset events usually focuses on large-scale adaptation measures or broadly applicable regulations and policies that are typically informed by cost-benefit analysis, which favours policies that protect higher-value assets and areas [3,5,14,15,25,27,55]. Faced with limited resources and large-scale impacts, adaptation decisions can thus quickly prioritize more affluent regions, increasing inequality [3,5,14,15,25,27,55]. For example, to prepare for major floods, the preferred method of adaptation is dikes, which are installed in areas of high value and impact coastal and riverine livelihoods that depend on access to water [3]. For hurricanes and earthquakes, policies are usually aimed at building levels, which actually might provide more opportunities to support poor and marginalized groups. But in reality governments often stimulate cost-effective, but expensive, retro-fitting measures which can usually only be applied to houses with a certain structural quality, disadvantageous to the poor who live in worse building conditions [27,55,59], or they selectively enforce building-codes in favour of the more affluent [14,55]. Adaptation measures that focus on the coping phase can also disadvantage the vulnerable. For example, hurricane Katrina showed that more affluent citizens of New Orleans had better opportunities to evacuate than poorer households [60]. Even in the post-disaster recovery phase of intensive rapid onset events, the constrained time and resources are often prioritized in higher value or urbanized regions [15]. Sovacool et al. [15] analysed the post-disaster recovery phase of four different rapid onset intensive events, and concluded that four processes aggravated distributional and procedural inequality; enclosure, where public funds are transferred to private actors, exclusion, where poor and marginalized groups are provided with limited access to the decision-making process, encroachment, where adaptation intrudes on biodiverse areas on which poor and marginalized groups depend, and entrenchment, where adaptation solutions aggravate pre-existing conditions. Furthermore, disaster relief funds for rapid onset intensive events tend to go to large NGO’s, and while they might be more (economically) effective, this overlooks small-scale local-based NGO’s which are better positioned to support poor and marginalized groups [61]. Also, in the wake of large-scale events, funding is diverted to reconstruction, pulling it away from general development and empowerment programs for poor and marginalized groups [61].

3.2 Intensive slow-onset hazards

Intensive slow-onset hazards are characterized by an unclear starting or endpoint, and typically involve droughts. Different types of droughts are defined, such as meteorological, agricultural, or hydrological droughts, which all affect large global regions simultaneously [62,63]. Droughts directly impact essential water use for poorer and vulnerable groups [64–67], and failing to implement adaptation in an equitable manner therefore quickly results in increased mortality and poverty [64,67,68]. The actual impacts of droughts are heavily driven by supply, demand, and access to water infrastructure, and therefore the distribution of impacts of droughts are even more a social construct than other hazards [64,67–69]. More so than intensive rapid onset hazards, intensive slow onset hazards disproportionately affect lower-income countries, with 90% of low-income countries being classified as at high or very high risk with respect to slow-onset hazards, in contrast to 5% of the high income countries [70]. In rural or agricultural areas, larger and wealthier water users can invest in larger water storage capacity, have better access to information, have better access to economic instruments, and have better access to authorities and power than poor and marginalized groups [68]. In urban areas, inequalities arise as more affluent neighbourhoods are better connected to water infrastructure, and richer households have better opportunities to reduce water consumption by cutting down on non-essential water use, while poor households would be directly impacted in their basic needs [64,67,68].

Equitable adaptation to slow-onset intensive hazards is complex. Similar to rapid onset intensive hazards, there is a tension between upholding equity standards when faced with resource constraints to address the magnitude of the event [64,71,72]. Adding to this, the lack of a clear starting point can lead to inadequate adaptation strategies that do not recognize the need of the poor and marginalized in time, as larger effects are only seen in a later stage [73]. By the time large-scale adaptive efforts are underway, the impacts might already have pushed the poor and marginalized in a worse position to cope with and recover from the drought. Moreover, as slow-onset hazards are spread out over time, they lead to complex interactions between other socio-economic processes and the impacts of hazards themselves. For example, adaptations aimed to improve water supply, such as storage and retention, are often only accessible by large-scale water users [68], geared towards more affluent regions and neighbourhoods [64,68], and even small-scale private or community water storage is often not affordable for poorer households [64,68]. On the other hand, reducing the demand of water is often not feasible for poor and marginalized groups, as they are already living on or near the minimum water demand levels [64]. Therefore, they cannot smooth their water use by reducing non-essential water use (e.g. washing cars), while more affluent households can [64]. Other adaptation solutions, such as water rights and water markets, can be effective measures to better match demand and supply, but again poor and marginalized groups often have no or limited access to these schemes [74].

3.3 Extensive rapid and slow onset hazards

In contrast to intensive hazards, extensive hazards are characterized by a higher frequency of occurrence and lower magnitude [28]. Typical examples of extensive hazards are small floods, landslides, wildfires, local or short-term heatwaves, and low-intensity or short-term droughts [68,75]. Similar to intensive events, poor and marginalized groups tend to live in riskier locations, for instance on slopes susceptible to landslides [76–78], areas that frequently flood [79,80], urban areas with higher temperature due to less green or less facilities for cooling such as air-conditioning [81,82], and in areas with less wildfire control [83]. While cumulative effects of extensive events can be as large as the impacts from intensive events [28,75,84], extensive events receive far less (inter-)national attention, are less researched, and are often not accounted for in disaster databases and therefore macro-economic assessments of risk [3,75,84]. Moreover, due to their small per-event loss, they are often not the focus of government or NGO-led adaptation, aid, or reconstruction efforts [75]. As a result of neglecting extensive events, the impacts are even more unequally distributed towards poor and marginalized groups, hindering development and perpetuating levels of poverty and insecurity [75]. Groups with a better socio-economic status might move away from frequently affected places [3], or they have opportunities to adjust their expenditure when faced with small impacts [3]. Poorer groups have less capacity to move away, and adjusting expenditure can often only be done by reducing essential expenditure on health or education [3]. Even though this smooths the impact on the short-term, it can increase the vulnerability of poor groups in the medium to long-term [3]. Furthermore, due to their more frequent occurrence, extensive events have a slow but deteriorating effect on transport, drinking water, power supply, food and fuel availability, leading to loss of workdays and reduced healthcare [14]. These slow deteriorating effects can also impact the effectiveness of existing social structures that would otherwise mitigate the impacts, such as community-based organizations or women groups [85,86]. Overall, the drivers of inequality for extensive events are therefore less related to the hazard itself compared to intensive events, but more to pre-existing socio-economic conditions and general lack of adaptation efforts [80,85].

Similar to intensive hazards, one of the causes for this is procedural inequality through the lack of representation of these groups in adaptation decision-making [14,28,76,84]. For instance, Anderson et al. [83] showed for wildfires how better positioned socio-economic groups have more opportunity to lobby or join adaptation decision-making structures, tilting adaptation measures towards better neighbourhoods. Similar evidence is found for small floods [28,29], landslides [76–78], heatwaves [81,82], and wildfires [83]. The difference with intensive events is that extensive events receive a general lack of attention [28,84]. Even the few adaptation strategies that are implemented also frequently exacerbate distributional and procedural inequalities [14,83], for instance by only implementing adaptation measures that prevent or reduce the hazard in more wealthier areas [76–78,80,82]. In contrast, some policies such land-use regulations, relocation, and eviction are reported to be enforced only in poorer neighbourhoods while wealthier neighbourhoods remain in place with better protection measures [14]. Moreover, such policies often overlook the area’s social cohesion and livelihood structure, leaving the poor and marginalized worse off after relocation [80].

3.4 Gradual processes

All hazard types are aggravated due to gradual climate or socio-economic change. The UNFCCC [52] distinguishes eight slow-onset processes (defined as gradual processes here); increasing temperatures, desertification, loss of biodiversity, land & forest degradation, glacial retreat, ocean acidification, sea-level rise, and salinization. Within countries, incremental negative effects on living conditions, such as declining fish stock, forest degradation, conflicts over scarce natural resources, and other effects can undermine poverty reduction and development efforts (Adamo et al., 2021 [87];Barnett J, Fincher R, Hurlimann A, Graham S, 2014 [88]; Solinska-Nowak et al., 2018 [89]). Gradual processes are foreseeable incremental phenomena, which potentially makes them easier to adapt to than rapid-onset events [30]. However, there are many challenges in adapting to slow onset events, not in the least because they tend to affect vulnerable groups disproportionally [87]. De Ruiter and Van Loon [90] also recognize slow-onset events as one of the three key dynamics of vulnerability. Moreover, due to their incremental nature, they are largely ignored in adaptation research and policy, which instead focuses on the hazard symptoms such as droughts and more frequent flooding [30]. As the incremental processes are allowed to continue, they continuously interact with anthropogenic parameters and stressors and their manifestations differ at the local scale, leading to increased complexity to adapt [91]. Adaptation to multi-hazard risk in the context of climate change presents added, unique challenges to long-term risk management. Adaptation must account for the non-stationarity of climatic processes, which contributes to both emerging and more extreme risks that we have not seen before, including those relating to tipping points. Next, the increasing uncertainty complicated mid- to long-term adaptation planning (Schlumberger et al., 2023) [92]. Understanding the role of gradual processes and their relationship with residual risk is crucial to develop comprehensive adaptation measures that address both short- and long-term vulnerabilities (Schipper 2020 [35]).

4.1 Cascading vulnerabilities

The importance and current lack of accounting for interactions between vulnerability dimensions and how these compound with risk is recognized in the current literature [e.g. 37].Even more so than for single hazards, adaptation to multi-hazards should focus on addressing the root causes of vulnerability, as during compounding or consecutive events, different vulnerability aspects or underlying vulnerability conditions interact and intersect with each other [20,90,98,103], and socio-economic vulnerability can cascade from one hazard to another. For example, disaster-induced unemployment can limit available financial resources to adapt to a next disaster, further reducing the capacity to prepare, cope and recover from multi-hazard events [104,105]. This is underlined by a recent stocktaking of compound climate disasters adaptation literature published between 2013 and 2019, in which Simpson et al. [20] identify 23 vulnerability aspects that have been observed to constrain the capacity to adapt under compound disasters, with low income, food insecurity and access to institutional resources and finance among the most prominent aspects. By addressing the root causes, those groups can shift towards lower vulnerability, reducing the cascading effect (Fig 1).

4.2 Cascading maladaptation and adaptation trade-offs

Even if adaptation measures are taken in an equitable manner for one hazard, they can have unintended negative consequences for a compounding or consecutive event [20,90,103,106]. These adaptation trade-offs are typically referred to as maladaptation [35] or adaptation a-synergies [32]. Examples are for instance because resources to support poor and vulnerable groups spent on one hazard cannot be spent on another hazard [103], or because farmers switch to more drought resistant crops which are more vulnerable to flash floods (de Ruiter, de Bruijn, et al., 2021 [32]; Wood & Good, 2004 [107]) Ward et al., 2020 [103]). The latter is particularly challenging when both floods and droughts become more frequent such as in the Sahel and Sub-Saharan Africa [108,e.g. 109]. Paradoxically, an adaptive response can thus lead to increased vulnerability, and due to its cascading nature can reduce future adaptive capacity [110]. This can be particularly impactful to an already vulnerable group when an adaptation strategy that aims to redistribute risk of the first hazard event actually aggravates the risk for vulnerable groups during the next event [35]. The lack of procedural inequality through underrepresentation and lack of agency enforces the potential for maladaptation or adaptation trade-offs, as the voice of the vulnerable is not heard twice over [111–113]. It is therefore key to better understand the effectiveness and challenges that multi-hazards pose for equitable adaptation [20,32,35,103,105,114,115].

4.3 Adaptation challenges due to varying temporal and spatial dimensions

Multi-hazards present further unique challenges for equitable adaptation due to varying temporal and spatial dimensions. Temporal challenges occur for example due to differing onset speeds of the hazards [22], mismatches between the temporal dimension of the hazard and the availability of funding and policy cycles [116], lags between hazard awareness and adaptation [117], the interaction between short- and long- term multi-risk adaptation strategies [35], and unequal recovery periods [118]. For example, slow onset events or extensive events can slowly exacerbate vulnerability, and when overlooked in adaptation efforts can put vulnerable communities at increased risk for a sudden intensive event like floods, making adaptation efforts for the sudden intensive event less effective [22]. In another example, Inuit communities overspecializing their harvesting practices as a climate change coping mechanism, decreases their short-term vulnerability but increases their long-term adaptive capacity [36,119]. The temporal difference can also lead to ineffective resource allocation, as sudden onset events might be prioritized over slow onset events, leaving communities that face slow-onset events at risk [116]. Similarly, for adaptation to be equitable, it needs to recognize and address the longer recovery phase for vulnerable populations [118] as this puts them longer at risk for compounding hazards as shown in Fig 1. The spatial dimension can also lead to specific challenges for equitable adaptation, due to the unequal distribution of hazards, for instance between coastal and inland communities [120], the disparities between infrastructure and services between urban and rural communities [8], the particular challenges for urban areas (high density population, heatwaves)[11], and the spatial mismatch between adaptation resources [120]. For instance, vulnerable communities in rural areas face relatively larger challenges from compounding events, as there is less infrastructure and public services, and due to low economic value they are less prioritized in adaptation planning. However, this might similarly apply to informal urban settlements such as slums, as they are not included in adaptation planning [11].

5.1 Qualitative approaches for complex socio-environmental systems

Qualitative approaches are particularly suitable to elicit a deep understanding of the social processes [64,121] and there are various qualitative methods that can support the analysis of the situation of vulnerable groups in relation to hazards, their coping capacity, and their needs. Some recent methodological advances promise to be particularly suitable for supporting knowledge on equitable adaptation in a multi-hazard context, as they already have specifically moved to focus on socio-environmental systems. First, participatory scenario planning (PSP) has been applied to study complex socio-environmental systems in the context of climate adaptation [122,123]. PSP has the potential to integrate diverse views, local context, complexities and uncertainties, and by doing so foster understanding and decision-support [122,123]. Recent developments focus on making PSP more participatory and inclusive, including diverse community members and designing adaptation pathways together [122]. Similarly, Participatory Action Research (PAR) has been used to understand climate adaptation planning and sustainable development in socio-ecological systems [123]. PAR also holds potential to focus more on power structures, social inequalities, and injustices, and how to overcome these together and transform them towards equitable adaptation. While PSP and PAR, have seen an increased use for equitable adaptation application, so far they have seen very limited use to specifically analyse equitable adaptation in a multi-hazard context. Partly, this is because the multi-hazard research is a young and evolving field of research [124–126], and complex interactions between adaptation measures for compounding and consecutive events are not well known yet (de Ruiter, de Bruijn, et al., 2021b [32]; Schipper, 2020b [35]; Ward et al., 2022 [124]). As this fields matures, quantitative methods are needed to integrate knowledge on adaptation synergies and adaptation trade-offs in the qualitative approaches, to discuss with individuals and communities the implications for equity. PSP and PAR are particularly well-suited to be applied in low-resource context. For example, place-based PSP has been successfully applied in Latin America, Africa, and Asia [127–129], and facilitates common understanding of adaptation and learning within communities [127–129]. Similarly, PAR is particularly suitable to be applied in low-resource context, and they prioritizes the value of experiential knowledge for tackling problems caused by unequal and harmful social systems (for a review, see Cornish et al. [130]). These applicability of qualitative methods increases the geographical reach of the method, and facilitates the usefulness for communities to design equitable adaptation measures in a multi-hazard context.

5.2 Quantitative approaches for complex socio-environmental systems

Quantitative modelling approaches can aid in understanding the complex interactions between multi-hazard adaptation and equity, and support qualitative methods and vice versa. However, such models need to be able to integrate a range of concepts and input, from multi-hazard maps to the socio-economic effects of adaptation decisions. A field of modelling that is particularly suited for this are Socio-Environmental System (SES) models [131,132] or otherwise referenced as coupled human and natural systems (CHANS) models [133,134]. These modelling concepts depart from the premise that the social, economic, and environmental systems have complex feedbacks and interactions and need to be modelled within one framework. Some examples that can be classified under these terms are integrated assessment models (IAM), system dynamics models, and agent-based models. Especially agent-based models have evolved in recent years as a strong modelling tool to understand complex non-linear processes in adaptation decision-making (An, 2012 [133]; Anshuka et al., 2022 [135]; Filatova et al., 2016 [131]; Haer et al., 2019, 2020b [44,106]; Schrieks et al., 2021 [136]; Taberna et al., 2020 [137]). While not applied yet to study equitable adaptation in a multi-hazard concept, they are particularly suited to do so as (1) they capture heterogeneity of agents in characteristics and decision-making, allowing for representation of individuals and groups living in vulnerable situations (2) they are designed for complex feedbacks and interaction between agents, such as vulnerable individuals, communities, and government action (3) they allow for integration of environmental data and processes, allowing for the inclusion of multi-hazard data and processes, and (4) they capture emergent behavior, in this case the (in)equity of adaptation decisions by different agents and to different hazards, which potentially occur consecutively or compounding.

Moreover, agent-based models lend themselves to strengthen procedural equity through participatory approaches, which can mean involving groups living in vulnerable situations in the model design and outcome analysis [67]. In the conceptualization, design, analysis, and validation and calibration phase stakeholders can be included [67]. While ABM development requires considerable training and know-how, they have been successfully applied in LMICs. For example, Hailegiorgis et al., [138] use ABM to provide insight in adaptation strategies for rural households in Ethiopia, Wens et al., [139] do similar work for Kenya, and Amadou et al. [140] how the applicability for ABM on northern Ghana. Here, the capacities of groups living in vulnerable situations do need to be considered, to align capacities with the phase of contribution. Combining ABM with focus group discussions or household surveys are often a direct and useful way to integrate local knowledge and context. The more technical sections in ABM development, such as design and analysis can be done with stakeholder groups that represent the most vulnerable, while conceptualization and even validation and calibration can be supported by group discussions, workshops, or more participatory approaches such as serious game design of agent-based models.

5.3 Decision-making under deep uncertainty

As the interaction between multi-hazards and equity increases complexity, specialized decision-making tools are needed that specifically address deep uncertainty. In recent years, methods for Decision-making Under Deep Uncertainty [141] have emerged as effective tools for decision-making for short- and long-term adaptation. However, they have been criticized for overlooking the organizational and individual context [141]. To address equity in adaptation, DMDU approaches should therefore be combined with the qualitative and quantitative methods described above.

Examples of DMDU approaches are Robust Decision Making (RDM), Dynamic Adaptive Planning (DAP), Dynamic Adaptive Policy Pathways (DAPP), Info-Gap Theory (IG) and Engineering Options Analysis (EOA). We highlight here the DAPP, as recent advances in this field also address the multi-hazard context [92]. DAPP approaches recognize that adaptation decision-making needs to be flexible and robust when faced with long-term uncertainty, such as under the context of climate change [142]. The method develops pathways, with decision-triggers, and adaptation tipping points, which reflects moment in time to act to remain flexible and robust in the future. Initially, the usefulness of these methods was shown for one hazard, but the method is now expanded to the multi-hazard context (DAPP-MR) [92]. The DAPP-MR expands the method by also considering interaction effects between hazards, and the effects therefore on different economic sectors, to meet certain objectives like minimized costs, minimized damages, minimized crop or production loss. The first analysis of this methodology shows that the framework is useful to show non-linear effects. Similarly, we propose here that this DAPP-MR framework could be expanded to include equity considerations, thereby addressing equitable adaptation in a multi-hazard context. Specifically, this means that objectives have to be integrated that measure the distributive effects of adaptation on groups living in vulnerable situations.

The DAPP-MR approach also lends itself to increasing procedural justice in adaptation decision-making when combined with participatory approaches. In the development stage, this could be done through qualitative approaches such as focus group discussions, storyline development, and others as discussed in [92]. Once developed, the DAPP can be represented as a metro-map, with visuals that are comprehensible for a broader audience. This offers further opportunities to involve groups living in vulnerable situations, by checking in a participatory way whether the views, opinions, and thoughts are represented in the DAPP, and whether the outcome objectives reflect the needs of people living in vulnerable situations. While DAPP was developed and first applied in high-resource countries, it is increasingly used to support adaptation decision-making in LMICs, with application in Africa, Asia, and South America [141]. Recognizing that the capacity to apply the full DAPP framework might be limited in low-resource context, the DAPP can also be amended towards a more qualitative framework [143]. This supports also the argument for a mixed-methods approach to elicit equitable adaptation in a multi-hazard context.