Drought vulnerability and risk assessments: state of the art, persistent gaps, and research agenda

Reducing the social, environmental, and economic impacts of droughts and identifying pathways towards drought resilient societies remains a global priority. A common understanding of the drivers of drought risk and ways in which drought impacts materialize is crucial for improved assessments and for the identification and (spatial) planning of targeted drought risk reduction and adaptation options. Over the past two decades, we have witnessed an increase in drought risk assessments across spatial and temporal scales drawing on a multitude of conceptual foundations and methodological approaches. Recognizing the diversity of approaches in science and practice as well as the associated opportunities and challenges, we present the outcomes of a systematic literature review of the state of the art of people-centered drought vulnerability and risk conceptualization and assessments, and identify persisting gaps. Our analysis shows that, of the reviewed assessments, (i) more than 60% do not explicitly specify the type of drought hazard that is addressed, (ii) 42% do not provide a clear definition of drought risk, (iii) 62% apply static, index-based approaches, (iv) 57% of the indicator-based assessments do not specify their weighting methods, (v) only 11% conduct any form of validation, (vi) only ten percent develop future scenarios of drought risk, and (vii) only about 40% of the assessments establish a direct link to drought risk reduction or adaptation strategies, i.e. consider solutions. We discuss the challenges associated with these findings for both assessment and identification of drought risk reduction measures, and identify research needs to inform future research and policy agendas in order to advance the understanding of drought risk and support pathways towards more drought resilient societies.


Introduction
Droughts are recurring slow-onset hazards that can potentially have major direct and indirect impacts on human and natural systems, including terrestrial and freshwater ecosystems, agricultural systems, public health, water supply, water quality, food security, energy, or economies (e.g. through tourism, transport on waterways, forestry) (Schwalm et al 2017). While drought generally refers to a lack of water compared to normal conditions (Van Loon et al 2016), droughts are commonly grouped into four major types, including (i) meteorological or climatological, (ii) hydrological, (iii) agricultural or soil moisture, and (iv) socioeconomic drought (Wilhite and Glantz 1985). They are characterized in terms of their frequency, severity, duration, and extent (Zargar et al 2011). According to existing conceptual models (Wilhite andGlantz 1985, Van Loon et al 2016), these drought types generally occur in a particular sequence: climate variability leads to a precipitation deficit that instigates a meteorological drought, which when combined with high potential evapotranspiration leads to an agricultural or soil moisture drought. Hydrological droughts occur as a delayed hazard associated with the effects of temperature anomalies, precipitation shortfalls, and/or anthropogenic demand pressures on surface or subsurface water supply, such as streams, reservoirs, lakes or groundwater. Socioeconomic drought is associated with the impact of an inadequate supply of some economic goods resulting from meteorological, agricultural, and hydrological droughts (Wilhite 2000, Zargar et al 2011, Van Loon et al 2016, Wang et al 2016. However, despite the progress that has been made in classifying and characterizing different drought types, no commonly accepted definition of what comprises a drought hazard exists (Mukherjee et al 2018).
Over the past decades, drought events across the world have caused damage to human wellbeing, the environment, and the economy. While there is ambiguity regarding drought trends in the past century (Andreadis and Lettenmaier 2006, Sheffield et al 2012, IPCC 2013, Trenberth et al 2013, McCabe and Wolock 2015 due to a lack of direct observations and the dependency of trends on drought index choice, it is expected that drought hazards will increase in both frequency and severity in many regions across the globe in the coming decades as a result of climate change (Sheffield and Wood 2008, Dai 2011, IPCC 2012, Trenberth et al 2013, UNCCD 2016. Despite the high uncertainty regarding future trends, risk assessments are needed in order to understand and ultimately reduce the risk of negative impacts associated with droughts. Today it is widely acknowledged that risk, i.e. the potential for adverse impacts or consequences, is not driven only by natural hazards (droughts, floods, etc), but results from the interaction of hazards, exposure, and vulnerability (IPCC 2012(IPCC , 2014. According to the Intergovernmental Panel on Climate Change (IPCC), exposure in this context refers to the 'presence of people, livelihoods, species or ecosystems, environmental functions, services, and resources, infrastructure, or economic, social, or cultural assets in places that could be adversely affected' by such hazards (IPCC 2014, p 5). Vulnerability is the predisposition to be adversely affected, resulting from the sensitivity or susceptibility of a system and its elements to harm combined with a lack of short-term coping capacity and long-term adaptive capacity (IPCC 2014). Due to its complex, multi-dimensional nature (Turner et al 2003, IPCC 2014, drought risk can therefore not be adequately represented solely by a single factor or variable, such as a rainfall deficiency or poverty (Chambers 1989). Rather, it is often driven by a variety of context and impact-specific factors, including environmental, social, economic, cultural, physical and/or governance-related aspects (Birkmann et al 2013, Hagenlocher andCastro 2015).
Cross-sectoral and impact-specific assessments of who and what (e.g. people, agricultural land) is at risk to what (e.g. meteorological or soil moisture drought), as well as where and why, will be key for the identification of targeted drought risk reduction, resiliencebuilding, and drought adaptation strategies (IPCC 2014, González Tánago et al 2016, UNCCD 2016. The need to understand, assess, and monitor drought risk is underscored by relevant international agreements and initiatives such as the Sendai Framework for Disaster Risk Reduction 2015-20304 (UNISDR 2015 or the 2018/19 UNCCD Drought Initiative 5 . A range of approaches exist for assessing vulnerability and risk in the context of climate change and natural hazards such as droughts. These include quantitative, qualitative, and increasingly mixed-methods approaches that combine both (Schneiderbauer et al 2017). Promoting and integrating a plurality of approaches can produce complementary information to better explain the complexity of processes that mediate vulnerability and risk. The choice of the approach depends not only on the scale of analysis (local to global), but also on the scope of the assessment, such as understanding root causes, identifying spatial and temporal patterns and hotspots of risk, etc. Qualitative vulnerability and risk analysis often makes use of a wide array of data collection techniques such as interviews, focus group discussions (FDGs), or storylines to reveal context-specific root causes of risk. In contrast, quantitative assessments tend to apply criteria and indicators to assess vulnerability and risk, often in a spatially explicit manner.
In addition to assessing current patterns of risk such as risk hotspots, the analysis of past trends and dynamics and the development of future scenarios in vulnerability and risk have sparked increasing interest and attention in recent years for a number of reasons. The analysis of past trends or risk dynamics through repeated risk assessments can support the monitoring and evaluation of risk reduction and adaptation options (Hagenlocher et al 2018b). Future risk scenarios can provide useful inputs for precautionary, preventive, and adaptive planning (Garschagen andKraas 2010, Birkmann et al 2015). A recent review of climate risk assessments concluded that while the number of studies that include temporal dynamics is growing, the majority of future-oriented assessments do not consider scenarios of exposure and vulnerability (Jurgilevich et al 2017) instead focusing on the hazard element of the risk concept.
Many of the steps in quantitative drought risk assessments, such as data imputation, outlier treatment, normalization, weighting of indicators or proxies, and aggregation, introduce uncertainty into the modeling/analysis result. Statistical validation-in the form of both sensitivity/uncertainty analysis and the regression of risk assessment outcomes against observed impacts or losses (e.g. crop losses, number of people affected)-has proven to provide relevant information on the reliability, validity, and methodological robustness of risk assessments and their outcomes (Schmidtlein et al 2008, Fekete 2009, Tate 2012, Hagenlocher and Castro 2015, Welle and Birkmann 2015, Feizizadeh and Kienberger 2017. However, its application in the field of risk assessment remains largely underdeveloped.
Over the past decades, a number of review articles have been published focusing on (i) drought classifications and definitions (Mishra and Singh 2010), (ii) the assessment and monitoring of drought hazards in general (Rossi et al 1992, Mishra and Singh 2011, Zargar et al 2011, Li and Zhou 2014  However, a review of existing concepts, methods, approaches, and studies on drought vulnerability and people-centered integrated risk assessments is still lacking.
This paper seeks to close this gap by analyzing the state of the art and identifying key gaps regarding the assessment of drought risk with a focus on people. Furthermore, the paper aims to evaluate to what extent existing drought risk assessments suggest potential solutions for drought risk reduction or adaptation. A synthesis of the findings informs a recommended agenda for future research.

Methods
A systematic literature review was conducted to synthesize and better understand (i) how peoplecentered drought risk is currently conceptualized and assessed in the scientific literature, (ii) how existing assessments are linked to the identification of drought risk reduction or adaptation strategies and measures, and (iii) what gaps and research needs exist. The following questions guided the analysis: 1. How are existing assessments distributed across geographic regions (e.g. continents, countries) and spatial scales (local to global)?
2. How is drought risk conceptualized?
3. Does each assessment specify the drought type analyzed, and if so, which type of drought hazard was considered?
4. Which drivers of vulnerability and drought risk are used in existing risk assessments?
5. Which assessment approaches (e.g. qualitative, quantitative, or mixed methods; index-based assessments versus. dynamic simulations) were used? Was sensitivity and/or uncertainty analysis or any form of validation of results applied?
6. Are temporal dynamics considered (e.g. past trends, future scenarios of drought risk) or is the focus largely on evaluating current patterns and hotspots of drought risk?
7. To what extent are assessments of drought vulnerability and risk linked to the identification and planning of drought risk reduction and/or adaptation options? When they are, which measures are proposed?
8. Which key gaps exist in understanding, characterizing, and assessing drought risk?
Peer-reviewed research articles were identified from the Web of Science and Scopus databases covering the period from January 1970 to December 2018 based on a set of pre-defined search terms focusing on people-centered drought risk assessments (table 1). The search was conducted between December 2017 and January 2018 and re-run during the revision process in February 2019. A systematic approach that only includes peer-reviewed articles was selected to ensure transparency, reproducibility, and quality of the analysis following an adapted workflow for systematic literature reviews as proposed by Rudel (2008), Hofmann et al (2011) andPlummer et al (2012).
In a second step, the titles, keywords, and abstracts of the identified articles were screened independently by three researchers and allocated to a 'YES', 'NO', or 'PER-HAPS' list based on each author's judgment of relevance to the search criteria. The respective decision was crosschecked by the two other researchers and assessed for its relevance for the review. Whenever an article was allocated to the PERHAPS list by one of the three authors, the full article was read by all three researchers in order to decide whether or not to include it in the review (YES list) or not (NO list), and the outcomes discussed and cross-checked. In a third step, a coding scheme focused on the aforementioned guiding questions was developed for in-depth content analysis of the final set of articles and implemented in MAXQDA software (VERBI Software 2017). Finally, the information was analyzed using descriptive and statistical methods in Excel software. The following sections are structured according to the eight questions outlined above.
In order to respond to question number four on vulnerability factors a classification scheme was developed to inform the content analysis of the articles, drawing on a scheme proposed by González Tánago et al (2016). In a first review of factors of vulnerability in the context of droughts they grouped vulnerability factors into biophysical and socioeconomic dimensions and 11 sub-dimensions. Based on their work and the more recent grouping of drought vulnerability indicators into social, economic, and infrastructural dimensions by Carrão et al (2016), the finale scheme applied here encompasses a list of seven dimensions and 24 sub-dimensions or vulnerability factors (table 3).

Bibliometric analysis
Based on the systematic search protocol, a total of 1141 articles were identified, including 568 articles from Web of Science and 573 from Scopus. Following the multi-step process described above, the number of articles considered for the final review was reduced to 105 (table 2; supplementary material 1 is available online at stacks.iop.org/ERL/14/083002/mmedia).
Overall, more than 95% of the assessments were published after 2005-the year the Hyogo Framework for Action 6 (HFA) (UNISDR 2005) was adopted by 168 governments-and almost 60% of all assessments were published in the past four years, i.e. between 2015-2018 (supplementary material 1). This is not surprising given the strong call for risk assessments in the HFA 2005-2015 and in the Sendai Framework for Disaster Risk Reduction 2015-2030 (UNISDR 2015), which was adopted in 2015. Figure 1 shows the geographic distributions, by climate zone and by spatial scale, of all the assessments reviewed. The most assessments (46%) were conducted in Asia, followed by Africa (29%) (figure 1), and in mainly dry (34%) or tropical (19%) climates or across climates. As such, the studies are highly concentrated in a few countries, namely China (18), India (11), the United States (9), Ethiopia (6), and Brazil (5). In terms of spatial scales, assessments at the sub-national level are dominant, with only very few studies that draw conclusions at the global or local/community level.

Conceptualization of drought risk
The review demonstrates that a variety of different risk definitions have been used as a conceptual underpinning for characterizing and assessing drought risk and highlights two contrasting developments (figure 2). First, there is an increasing number of studies that follow the conceptual understanding of risk as  promoted by the IPCC. Second, there is an increasing number of drought risk assessments that do not specify how drought risk is conceptualized in their assessment (i.e. they do not provide any definition of risk). The majority of articles that provided a definition of drought risk used the IPCC concepts of 2001 (IPCC 2001) and 2007(IPCC 2007. However, since the publication of the IPCC SREX Report (IPCC 2012) and the subsequent Fifth Assessment Report (IPCC 2014), there has been a shift in the conceptualization of risk towards a stronger focus on assessing the risk of specific consequences or impacts that may harm a system, wherein risk is a function of (drought) hazard, exposure, and vulnerability (IPCC 2014). This has been reflected to some degree in studies assessing drought risk ( The ambiguity in definitions is also reflected when analyzing how vulnerability-as a key component of risk in the IPCC AR5-is conceptualized and operationalized in existing drought risk assessments. Of the articles reviewed, 34% consider sensitivity and/or susceptibility, 25% consider adaptive capacities and only 14% consider coping capacity as sub-components of vulnerability. Eleven percent of all papers include drought hazard characteristics and 14% include exposure 7 as part of vulnerability. The review reveals that although different types of drought hazards are acknowledged in the scientific literature, more than 60% of the assessments published on drought risk do not explicitly specify the type of  7 Here, exposure is understood based on the IPCC (2014) definition as 'exposed elements'. Thus, even if authors used the term 'exposure', it was not considered to have been conceptually applied if only hazard characteristics were used as proxies. drought hazard that is addressed ( figure 3). This is particularly relevant for drought given that the different drought types have very different implications in terms of potential impacts and policies to mitigate these impacts (Wilhite 2000).
Although it is increasingly acknowledged that droughts cannot be seen as purely natural hazards (Van Loon et al 2016) and there is a need to consider the complex interactions between natural and human systems when analyzing vulnerability and risk (Turner et al 2003), the review clearly shows that the majority of existing drought vulnerability and risk assessments still focus largely on the social dimension and do not apply an integrative social-ecological systems (SES) perspective. Out of the 105 articles that were reviewed, only 18 (17%) applied an SES perspective. This confirms a persistent gap in vulnerability and risk assessments that was recently highlighted by Sebesvari et al (2016) in their review of vulnerability assessments in coastal river deltas.
3.3. Assessment of drought risk 3.3.1. Assessment approaches The review of existing drought risk assessments revealed that the majority of studies applied quantitative (56%) or mixed-methods (32%) approaches, while purely qualitative approaches are rather rare (11%) and have mostly been applied at the subnational level with results extrapolated to explain phenomena at broader spatial scales (Nelson and Finan 2009, Saha et al 2012, Ayantunde et al 2015, Birhanu et al 2017. In terms of assessment methodology, more than half of the assessments used an index-based approach (62%) to tackle the complexity of drought risk, followed by dynamic simulation methods (12%) and lastly the more qualitative method of using narratives or story lines (8%). For example, Carrão et al (2016) use a static, index-based approach to map the global patterns of drought risk by integrating hazard, exposure, and vulnerability indicators into a composite risk index. Meanwhile, Martin et al (2016) apply a processbased, spatially-explicit social-ecological model for analyzing system dynamics contributing to drought risk for pastoral households in Morocco. In contrast, Ayantunde et al (2015) use qualitative methods (FDGs, community workshops, seasonal calendars, etc) to analyze the patterns and causes of drought risk in three agro-pastoral communities in Western Africa.

Factors and indicators to characterize drought vulnerability and risk
The review of literature conducted here has revealed that factors related to poverty and income (49%), technology (47%), education levels (34%), or the availability and quality of infrastructure (34%) were deemed important drivers of vulnerability and risk by almost one third of all reviewed assessments (table 3).
Following the classification scheme of table 3, 65 different indicators (18 belonging to the social dimension, 13 to the economic dimension, seven to the physical dimension, two to the crime and conflict dimension, eight to the governance dimension, nine to the environmental dimension, eight to the farming practices dimension) were identified during the review which can serve as a basis for future vulnerability and risk assessments (see supplementary material 2 for the complete list of indicators).
In order to identify and incorporate the potentially varying relevance and contribution of factors and indicators to vulnerability and risk in the context of natural hazards, a wide variety of weighting schemes have been developed (OECD 2008). These schemes can be categorized as being based on statistical models (e.g. regression analysis, principal component analysis) or on experts and/or community participatory consultation (e.g. ranking, budget allocation, Delphi methods). In most of the assessments reviewed here (57%) the authors did not explicitly specify their weighting methods, which is also in line with findings from a recent review of disaster risk, vulnerability, and resilience indices (Beccari 2017). Thirty-two percent of the reviewed assessments used statistical methods and ten percent used participatory, expert-based approaches.

Past trends, current patterns, and future scenarios
Fifty-four percent of the reviewed drought risk assessments are static, that is, they represent a snapshot in time. For the remaining 46%, most studies focus on assessing past trends (32%) and only 11 articles (10%) explore future scenarios of drought risk. Four percent of the articles do not specify the time frame of their analysis. Similar to other future-oriented risk assessments (e.g. in the context of sea level rise, flooding, etc)-where the focus is often on the modeling-based analysis of Table 3. Vulnerability dimensions and sub-dimensions used in the 105 studies considered in this review.

Physical
• Availability and quality of infrastructure (e.g. transportation; water and sanitation; energy; water tanks; reservoirs; wells; water quality)

Validation of risk assessments
Our analysis shows that less than 20% of the drought risk assessments reviewed here have conducted any form of validation of their results and only 12% have conducted a statistical sensitivity or uncertainty analysis. To date, only four studies (less than four percent) have conducted both a validation of the outcomes of the risk assessment against observed impacts and sensitivity analysis (

Drought risk reduction and adaptation
Effective drought risk assessments are those that center around the ultimate objective of being used or useful for disaster risk reduction (DRR) 8 and/or adaptation 9 strategies. While strategies should be based on context-specific empirical findings-taking into account both drivers and patterns of risk-the assessments should also consider what actions individuals and institutional bodies are already taking and their effectiveness. Less than half (40%) of the assessment papers reviewed make a direct link to drought risk reduction or adaptation strategies. Those that do comprise a wide array of structural (i.e. engineering-based or technological) and non-structural (e.g. capacity building, ecosystem-based approaches) solutions (table 4). Table 4. Drought risk reduction and adaptation options proposed by the authors of the reviewed studies.

DRR or adaptation solution Examples
Structural measures • Implementation and use of irrigation infrastructure • Water supply systems (e.g. dams, pipelines, cisterns) • Maintenance of water supply systems (desalinization and wastewater treatment plants, reducing leakage rates) • Early warning systems • Farming technology (use of, investment in) (e.g. machinery) Non-structural measures (individual, household, or farm level) • Water conservation • Diversification of livelihood strategies • Education and training (e.g. in water conservation, farming practices, drought awareness, drought risk management) • Fertilizer/manure (use of, increase in) • Pesticide/herbicide/pest control (use of, increase in) • Migration (temporal, permanent) Non-structural measures (government level) • Providing better access to credits and financial instruments • Implementation of social assistance and social protection programs • Access to finance instruments (credit, savings, markets) • Implementation of crop/climate risk insurance schemes • Investment in research and development • Water management practices/policies • Drought, water and climate change adaptation plans/policies • Mainstreaming indigenous and local knowledge into policy planning • Drought/emergency response and preparedness (equipment, facilities, funds) • Risk-informed (land use) planning Non-structural measures (ecosystem-based) • Soil conservation practices • Changing farming practices (e.g. crop diversification, drought resistant crops, adjusting planting dates, climate-smart agriculture, horticulture, intercropping, rotations) • Reclamation of degraded land • Water harvesting • Expanding the number and coverage of protected natural areas 8 Disaster risk reduction aims at preventing new and reducing existing disaster risk and managing residual risk (based on UNISDR terminology; https://unisdr.org/we/inform/terminology). 9 Here, adaptation refers to the process of adjustment to changing drought frequency, intensity, duration, or extent (based on IPCC 2014).

Discussion: persisting gaps, and research agenda
Existing review articles on the topic so far have primarily concentrated on (i) drought concepts and definitions ( . This paper complements these reviews by conducting a systematic review of peoplecentric drought risk assessments published between January 1970 and December 2018. Despite the boost in drought risk research over the past decades, the review has revealed and re-confirmed a number of persistent knowledge gaps of conceptual, methodological, and practical nature and relevance. In synthesizing these gaps, a number of needs have been identified that should be addressed in future research. Table 5 summarizes persisting gaps and the related needs from a conceptual, methodological and practical perspective.

Conceptual gaps and needs
Our analysis shows that more than 60% of the reviewed studies do not explicitly specify the type of drought hazard that is addressed and re-confirms that a broad variety of definitions of drought vulnerability and risk are used. This creates not only terminological and taxonomic confusion when operationalized in assessments, but also complicates the comparability of assessments and their outcomes-a gap that has also been emphasized in previous studies (Ebi and Bowen 2016, Bacon et al 2017. While context is crucial and other operational definitions of risk may be more appropriate depending on region and purpose (Wilhite 2000), providing a definition is important for producing scientifically rigorous and comparable work. There is increasing recognition that the causes of drought impacts on people and factors that dictate severity are complex, interact with each other, and are often features of coupled SESs (Van Loon et al 2016). The majority (83%) of existing people-centric drought risk assessments still focus largely on the social dimension and do not necessarily apply an integrative approach when characterizing drought hazards, vulnerability, or risk. As demonstrated in table 3, only 13%-14% of the reviewed articles considered factors such as soil conditions or quality or the protection of ecosystems in their assessments. Particularly when assessing drought risk in the context of agricultural systems (including people whose livelihood depends on agriculture), which are by definition SES, an SES perspective could help to understand and evaluate the role of degraded ecosystems as a driver of drought risk. Furthermore, an SES perspective can help to better understand the role of ecosystems and their regulating services as an opportunity for drought risk reduction-a gap that has also been highlighted by Asare-Kyei et al (2017). These gaps demonstrate the need for enhanced conceptual models that underscore the complex, differential interplay between drought hazards, exposure, vulnerability, and impacts while acknowledging the relevance of human-environmental interaction in each of these components. The latest definitions put forward by the IPCC in its Fifth Assessment Report (IPCC 2014), widely acknowledged by both the DRR and climate change adaptation communities, can help to overcome the existing terminological confusion.

Methodological gaps and needs
When dealing with droughts, embracing complexity is necessary for understanding the multi-dimensional nature of drought risk. Over recent years, index-based approaches have been promoted as useful tools to measure, compare, and monitor the complexity of risk associated with natural hazards and climate change (Sherbinin et al 2017) and have been gaining in popularity. Our analysis confirms this trend, with more than half of the reviewed assessments using index-based approaches (62%). However, their usefulness for policy support has also been subject to criticism (Hinkel 2011), given that indices are static in nature and do not capture the complexities and dynamics (e.g. nonlinearities and feedback loops) of vulnerability and risk (Hagenlocher et al 2018a). It is thus crucial to develop and apply methods, such as Bayesian or system dynamics modeling, that are able to both capture complexity and deliver simple messages for policy-making and allocation of resources.
The analysis has also shown that the relevance of individual hazard, exposure, and vulnerability indicators for explaining different drought impacts is poorly understood and tackled in assessments: 57% of the indicator-based risk assessments that were reviewed did not explicitly specify any weighting method. Future research should tackle this gap by exploring different ways for evaluating indicator weights (e.g. expert-based versus statistical approaches) and compare the findings by means of sensitivity analysis to evaluate the effect of weighting schemes.
Preventive planning for risk reduction and of adaptation measures requires a forward-looking perspective, and ideally should be based on different scenarios of future drought risk for a given region and impact-a need that has been increasingly emphasized over the past years (Garschagen andKraas 2010, Birkmann et al 2015). In addition, the monitoring of risk trends and changes in risk components and indicators over time can contribute to the monitoring and evaluation of risk reduction and adaptation measures. This has also been recently highlighted as a pressing need (Hagenlocher et al 2018b). Interestingly, 54% of the existing drought risk assessments are static in nature, i.e. they represent a snapshot in time, while the evaluation and development of future scenarios of drought risk (ten percent of all studies) is a rather recent phenomenon (the first paper in our review to develop future scenarios was published in 2009) and heavily underdeveloped aspect. In order to support the planning of adaptation strategies, scenarios of future risk pathways-in all components of hazard, exposure, and vulnerability-are urgently required.
The validation of risk assessments presents another persisting gap given the need of decision makers and practitioners for up-to-date and reliable data and information. Despite major progress in sensitivity and uncertainty analysis in the context of risk research (Fekete 2009, Tate 2012, Feizizadeh and Kienberger 2017, our analysis has shown that less than ten percent of all risk assessments reviewed here have conducted any form of validation of their results using impact data and only 12% have conducted a statistical sensitivity or uncertainty analysis. These findings are in line with gaps identified by Asare-Kyei et al (2017).

Practical gaps and needs
Risk assessments should ideally not be an end in themselves, but be linked to the identification, planning and prioritization of options for preventing and managing drought risk or adapting to changing conditions. The IPCC AR5 (IPCC 2014) identified the lack of assessments focusing on the actual implementation of adaptation measures and their potential positive or negative effects, a finding further confirmed in this review. While just under half of the studies reviewed here (40%) make a direct link to drought risk reduction or adaptation strategies, only Table 5. Summary of knowledge gaps of conceptual, methodological, and practical nature and identified needs related to people-centered drought vulnerability and risk assessments that could inform future research and policy agendas. 1. Linking of future research on exposure, vulnerability and risk to scenarios of relevant planning processes and a consideration of global change 2. Less than half of the assessments provide entry points for potential solutions (e.g. drought risk reduction or adaptation measures) 2. Provision of guidance on how risk assessments can support the identification, planning, monitoring and evaluation of risk reduction and adaptation strategies 3. Ecosystem-based solutions for risk reduction and adaptation are underrepresented 3. Further research on the role of ecosystembased solutions very few of these articles consider or recommend ecosystem-based approaches, leaving the potential of nature-based solutions (NbS) for drought risk reduction and mitigation (Kloos andRenaud 2016, UN 2018) far from being realized. Hence, more research is needed to evaluate the role of ecosystems and their services not only as drivers of drought risk, but also as an option for drought risk reduction and adaptation.

Conclusions
Reducing drought risk and associated direct and indirect impacts through targeted risk reduction and adaptation has become a global priority, as reflected by recent global initiatives and frameworks (e.g. the 2018/19 UNCCD Drought Initiative, Sendai Framework for Disaster Risk Reduction 2015-2030, Sustainable Development Goals, and the upcoming 2020 GAR Special Report on Drought) as well as by the steadily increasing number of drought risk assessments over the past decades. Efforts to reduce drought risk and adapt to changing environmental conditions by prioritizing and allocating funding and resources should be based on a sound understanding, characterization, and assessment of the drivers, patterns, and past trends as well as projected future patterns of drought risk. However, despite major advances over the past decades in terms of developing better methods and tools for characterizing individual components of risk, the review has revealed and re-confirmed a number of persistent knowledge gaps-of conceptual, methodological, and practical nature-which need to be urgently confronted in order to advance the understanding of drought risk for people, improve its assessment, and support pathways towards more drought resilient societies.