Exploring drought‐to‐flood interactions and dynamics: A global case review

This study synthesizes the current understanding of the hydrological, impact, and adaptation processes underlying drought‐to‐flood events (i.e., consecutive drought and flood events), and how they interact. Based on an analysis of literature and a global assessment of historic cases, we show how drought can affect flood risk and assess under which circumstances drought‐to‐flood interactions can lead to increased or decreased risk. We make a distinction between hydrological, socio‐economic and adaptation processes. Hydrological processes include storage and runoff processes, which both seem to mostly play a role when the drought is a multiyear event and when the flood occurs during the drought. However, which process is dominant when and where, and how this is influenced by human intervention needs further research. Processes related to socio‐economic impacts have been studied less than hydrological processes, but in general, changes in vulnerability seem to play an important role in increasing or decreasing drought‐to‐flood impacts. Additionally, there is evidence of increased water quality problems due to drought‐to‐flood events, when compared to drought or flood events by themselves. Adaptation affects both hydrological (e.g., through groundwater extraction) or socio‐economic (e.g., influencing vulnerability) processes. There are many examples of adaptation, but there is limited evidence of when and where certain processes occur and why. Overall, research on drought‐to‐flood events is scarce. To increase our understanding of drought‐to‐flood events we need more comprehensive studies on the underlying hydrological, socio‐economic, and adaptation processes and their interactions, as well as the circumstances that lead to the dominance of certain processes.

Both increased moisture transport and more active rain-producing systems may play a role in dry-wet transitions (e.g., Dong et al., 2011;Ma et al., 2019;Maxwell et al., 2017).
Whether these meteorological conditions develop into a drought-to-flood hazard and associated impacts depends on hydrological, socio-economic, and adaptation processes.There are some studies that provide examples of human adaptation to drought (or flooding) and how this affects the other extreme (De Ruiter et al., 2021;Garcia et al., 2022;Ward et al., 2020).However, the underlying physical and social processes that play a role in droughtto-flood events have been studied less and an overview of their reducing and enhancing effects is lacking.In this study, we aim to synthesize the current understanding on the hydrological, impact and adaptation processes behind drought-to-flood events, as well as their interactions.Based on a literature analysis and a global review of historic drought-to-flood cases (for a description of the methods, see Supporting Information), we show the diversity of processes that govern how drought affects flood risk and assess the circumstances under which positive and negative effects and feedbacks occur.The overview is subdivided in sections on "Catchment processes" (Section 2), "Impacts" (Section 3), and "Adaptation processes" (Section 4).Besides providing new insights on drought-to-flood interactions, we identify research gaps that should be addressed to better understand and improve the management of droughtto-flood events (Section 5).

| CATCHMENT PROCESSES INFLUENCING DROUGHT-TO-FLOOD EVENTS
This section explores how drought can alter catchment processes and how these changes can affect a subsequent flood event.In some cases, a drought can exacerbate subsequent flooding.In other cases, a change in catchment processes because of a drought can reduce subsequent flooding or make it less likely to occur.An overview of the different processes is provided, as well as an exploration of when and where different processes may be dominant.We make a distinction between storage and runoff process.An overview of the processes discussed in this section, and the factors that influence when and where they are dominant, is provided in Table 1.
T A B L E 1 Summary of catchment processes and the factors influencing when and where they are dominant.

Vegetation response Decrease
Increase in places with low aridity, high baseflow, a shift from snow to rain or resilience of high-elevation runoff.
Decreased infiltration and increased surface runoff

Increase
Snow related processes Decrease in case of a lower snowmelt peak because of below normal snowmelt or when precipitation falls as rain instead of snow Increase in case of rain falling on snow

| Storage processes
The hydrological conditions at the end of a long drought are not the type of conditions that are prone to result in flooding when heavy rainfall occurs.For example, subsurface storage will be low and will need to be replenished.With dry antecedent soil moisture conditions, a lower flood is expected (Berghuijs et al., 2016;Blöschl et al., 2015;Evans et al., 1999;Pathiraja et al., 2012), making drought-to-flood events less likely or less extreme.For example, the flash flood in the Netherlands in the summer of 2010, was less severe than it could have been because the first rainfall was stored in the dry soils (Brauer et al., 2011).In the UK 2010-2012 drought-to-flood event, initial rainfall led to soil moisture recovery.It was only prolonged heavy rainfall that resulted in flooding (Parry et al., 2013), with variability in time to peak discharge between quickly and slowly responding catchments (Parry et al., 2016).This suggests that, with greater storage depletion after more severe drought, recovery takes longer (Bravar & Kavvas, 1991), as found by Parry et al. (2016) in the United Kingdom.However, in some catchments in Australia, there was no such buffer effect and the flooding occurred soon after the drought (Yang et al., 2017).
The process of flood mitigation due to empty or low water storage caused by preceding droughts is not frequently mentioned in our case review of past drought-to-flood events.This is likely because this process alleviates flooding, thereby reducing or eliminating impacts.These events are therefore not included in our database (since it only includes cases with reported impacts).There were only two cases, in Saskatchewan, Canada and Iowa, United States of America, where the increased storage availability because of the drought was mentioned as reducing subsequent flooding (CBC News, 2015;Danielson, 2014).
One of the mechanisms that would explain an additional delay in the response of the hydrological system to excessive rainfall after drought is decreasing groundwater levels during drought.This leads to groundwater disconnecting from surface water (Eltahir & Yeh, 1999;Parry et al., 2013), which causes contraction in the stream network, shrinkage of the saturated partial contribution area, and slower pathways for water.One of the reasons for this is that it takes more water for saturation excess flow to occur (Saft et al., 2016).This would imply that heavier and longer duration rainfall is necessary for flooding to develop as a result of saturation overland flow.However, pluvial flooding from infiltration excess in the case of short heavy rain events would not be affected by this contraction.In addition, with the contraction of the stream network the contribution of groundwater to the stream becomes lower up to the point that the stream loses the connection with the groundwater and groundwater no longer contributes to the discharge.Large amounts of recharge are needed for groundwater and stream to reconnect and for groundwater to start contributing significantly to the discharge (Poeter et al., 2020).Drought recovery tends to be asymmetric (Eltahir & Yeh, 1999), with changes to the stream network taking longer to recover than to develop.Therefore, it would take longer for flooding to develop after a drought.

| Runoff processes
In the literature, there is also evidence for drought-induced changes in the land surface activating quick runoff pathways.This would suggest that catchments do not need to fully recover, with storage completely replenished, before flooding can occur (Parry et al., 2013).One mechanism that could play a role is vegetation response.In Australia, for example, many catchments remained in a lower runoff state for more than 7 years after the Millennium drought and showed no sign of recovery to the pre-drought runoff state (Peterson et al., 2021).In this case, the lack of recovery is not controlled by refilling of subsurface storage, but rather the authors postulate that it is related to a vegetation response leading to increased evapotranspiration (Peterson et al., 2021).However, the authors only investigated changes in annual runoff, and not changes in runoff extremes.In a study in the southern Appalachians, in California, Scaife and Band (2017) find that stormflow response is higher after a drought, because of a reduction in transpiration.Maurer et al. (2022) find that in some basins in California, the rainfall-runoff response during drought is higher than expected compared to the rainfall-runoff response under normal conditions.This occurs in basins with low aridity, high baseflow, a shift from snow to rain (i.e., more precipitation falling as rain instead of snow), or resilience of high-elevation runoff (i.e., no significant decreases in high-elevation runoff during a drought).Whether a drought increases or decreases the rainfall-runoff response, therefore seems to depend on the climatic and catchment conditions.
Many studies report increased runoff after drought because of decreased infiltration rates and increased surface runoff (Descroix et al., 2009).This can be due to, for example, increased soil compaction (Alaoui et al., 2018) or hydrophobicity (Evans et al., 1999).Also, soil cracking can result in quick vertical flow of rainwater (Miller et al., 1997) and increased flooding.Similarly, land cover changes can affect infiltration excess runoff generation, surface runoff routing and spatial connectivity, usually increasing (but in some cases decreasing) flooding (Rogger et al., 2017).Wildfires, often related to drought, have been found to change flow pathways (Murphy et al., 2018) and increase the risk of other hazards including flash floods, debris flows, and landslides (Moftakhari & AghaKouchak, 2019) (see Box 1 for more details on wildfires).The question remains whether these changes in pathways persist long after the end of the drought (Worrall et al., 2007) and whether they are important at the catchment scale (Alaoui et al., 2018;Blöschl et al., 2007).
In snow-dominated regions, other mechanisms can explain the relation between drought and flooding.Early or below-normal snowmelt in dry and warm years, could lower the snowmelt peak, reducing snowmelt floods ( Van Loon et al., 2015).The same occurs when precipitation falls as rain instead of snow, resulting in decreased snowfall and snowpack (Tabari, 2020).On the other hand, early snowmelt could also lead to higher peak runoff.Hatchett and McEvoy (2018) found that many snow droughts were characterized by lower snow fractions and midwinter peak runoff events.Under higher winter and spring temperatures (often co-occurring with drought), rain-on-snow flood events are expected to shift in occurrence.They shift from spring to winter (Freudiger et al., 2014) and from lower to higher elevations (Musselman et al., 2018).We can also speculate that drought in snow-covered areas could potentially lead to compacted snowpacks that could result in more severe rain-on-snow floods when rainfall occurs (Rössler et al., 2014).However, more research would be needed on this process.

| Which process is dominant, when, and where?
The availability of studies on effects at the catchment scale is limited.Therefore, it is difficult to answer the question of where and when storage depletion processes or runoff processes are dominant.Rainfall-runoff relationships are a good overarching metric to study effects at the catchment scale.Rainfall-runoff relationships often decrease during multiyear droughts.In these cases, the same amount of rainfall will produce lower runoff after a prolonged dry period compared to normal circumstances.This has been observed in Australia (Peterson et al., 2021;Saft et al., 2015) and Chile (Garreaud et al., 2017).In Algeria, however, rainfall-runoff relationships were found to increase, leading to less infiltration, more surface runoff and higher flood hazard (Sofiane et al., 2019).Although in this case it is not clear which specific process may have caused this decrease in infiltration and increase in runoff.Stronger decreases in rainfall-runoff relationships were found to be related to aridity (Garreaud et al., 2017;Saft et al., 2016), broader and flatter valley catchments (Saft et al., 2016;Yang et al., 2017), deeper soils, and more groundwater variability (Saft et al., 2016).Therefore, BOX 1 The role of wildfires and landslides in drought-to-flood events Drought increases the frequency and severity of wildfire (Riley et al., 2013), increasing flammability and interacting with other fire spread controls such as the prevalence of fire weather conditions (Littell et al., 2016).Wildfire drives changes in vegetation and soil properties making overland flow the dominant flow path postwildfire (Rountree et al., 2000) and increasing the risk of flooding (McGuire et al., 2021;Moftakhari & AghaKouchak, 2019).
This was illustrated in Australia where the 2019-2020 wildfires were a consequence of a widespread drought and a heatwave.While at first, torrential rain assisted in containing the fires, the wildfires had decreased infiltration capacity and flash floods soon followed (Alexandra & Finlayson, 2020).
Landslide risk increases up to several years after a wildfire.Wildfire followed by extreme precipitation causes increased run-off generated debris flows and slope instability (Rengers et al., 2020).In a study of Southern California, Rengers et al. (2020) found that slopes that burned 3 years prior to an extreme rainfall event, had the highest landslide density.Handwerger et al. (2019) report an increased risk in landslides during droughtto-flood events even without wildfires, because a rapid shift from drought to extreme rainfall may trigger the acceleration of landslides.
Our case review indicates that wildfires occur more often when a drought has propagated to a hydrological drought (Figure 3a).In case wildfires are reported during the drought event, landslides are also more often reported after the flood event (Figure 3b).
this may indicate that in areas with larger subsurface water depletion due to an arid climate, permeable geology, or deeper soils, the effect of filling up of depleted storage is more dominant, whereas in wetter, more quickly-responding catchments the balance might tip to runoff increase after drought.
In the global case review, we also find that the type of flood is important in drought-to-flood events.We define the type of flood generating process following the typology proposed by Merz and Blöschl (2003).Short-rain flood was the most frequently reported flood process (120 cases), whereas long-rain flood (34), tropical cyclone (20), and snowmelt flood (3) were reported less often as the dominant process causing flooding.The more frequent occurrence of short-rain floods points to high intensity rainfall events and limited infiltration.This may indicate that, in general, across all climate types, drought-induced changes to the land surface activating quick runoff pathways are a dominant process.Flood events reported as flash flood or pluvial flood (rather than a riverine flood or a combination of flash and/or pluvial with riverine flooding) were mentioned more often in case of a multiyear drought than a within year drought (Figure 1a).In addition, flash floods are more often reported to occur during than after a drought (Figure 1b).This may be because during a multiyear drought the land surface is affected more severely than during a 1-year drought.For example, through changes in vegetation or soil properties, causing quick-runoff generation processes to be dominant.However, it could also be that flash floods and pluvial floods after a drought instead of during a drought are not reported in our case review because the excessive rainfall refills depleted storage and did not cause severe impacts.In case of riverine flooding (which was reported much less often), storage depletion processes may be more dominant.This may especially be the case in arid regions, as discussed in the previous paragraph.We find some evidence for this, since groundwater drought is most often reported in semi-arid cases, as shown in Figure 2.There are no reports of groundwater drought in arid cases, but this may be due to the low number of cases with an arid climate in our case review.This may indicate an increase in available storage before a subsequent high rainfall period.However, there are also several cases with sub-humid and humid climate conditions that are reporting groundwater drought, which indicates there may be other factors, besides aridity, that influence the importance of storage depletion process.
All of these processes and their importance depend on the speed of drought recovery and the time lag between the drought and the heavy rainfall.The rate and duration of drought recovery are strongly dependent on climate and catchment properties, such as elevation, slope, average catchment wetness, and soil conditions (Ganguli et al., 2022;Parry et al., 2016;Yang et al., 2017).Additionally, human activities, including land use change, groundwater abstraction, and reservoir operation, have been found to prolong recovery (Apurv et al., 2017;Margariti et al., 2019) and shift seasonality (Wang et al., 2020).In many locations across the globe, the recovery time (i.e., the time between the drought and the flood event) is decreasing.This means that there is less time for the system to recover before the next extreme happens (Rashid & Wahl, 2022).We find that a flash flood by itself or in combination with pluvial flooding (rather than pluvial or riverine alone or flash flood in combination with riverine) is more often reported during a drought than after a drought (Figure 1b).This may be an indication of runoff processes being dominant and shows that heavy rain can lead to floods without terminating the drought.Riverine and compound flooding occurred mostly after the end of a drought (Figure 1b).This may indicate that for riverine floods to develop, depleted storage first needs to be replenished.
Both storage processes and land-surface conditions are influenced by human activities, especially during drought, but there is no empirical research on the effects of human activities on consecutive drought and floods.For example, studies on changes in the rainfall-runoff relationship (Garreaud et al., 2017;Saft et al., 2016;Yang et al., 2017) excluded "impaired" catchments.
Recent research has started to quantify the effects of human activities on hydrological drought (Rangecroft et al., 2019).Van Loon et al. (2022), for example, show that groundwater abstraction makes a hydrological drought more severe.In addition, human activities can slow drought recovery (Margariti et al., 2019).Flash floods occur more frequently and/or are more severe in human-influenced catchments (Jodar-Abellan et al., 2019;Mohamed & Worku, 2021).Human activities seem to enhance both storage and runoff processes.In human-dominated catchments, increased storage depletion leads to lower floods after droughts, and land-surface changes lead to higher floods after drought.These human activities are discussed in detail in Section 4 on adaptation processes.

| IMPACTS OF DROUGHT-TO-FLOOD EVENTS
In this section, we explore how drought impacts affect flood impacts.Droughts are generally long-lasting and have a range of cascading impacts on both the physical and societal system (De Brito, 2021;Stahl et al., 2016;Sugg et al., 2020).These impacts can continue into the period after the drought ended.If society is hit by a flood event, the impact may be higher due to the preceding drought impacts.

| Drought impacts on the physical system
Drought can impact the physical system, thereby making subsequent flood impacts more likely.Drought can cause damage to infrastructure, including infrastructure for flood risk mitigation.For example, during droughts, dikes and levees can suffer from cracking, which can increase the probability of failure (either during the drought itself or during subsequent wet periods).Examples of this are abundant in the literature, especially for the Netherlands (e.g., Van Baars, 2005;Van Baars & Van Kempen, 2009), Australia (e.g., Hubble & De Carli, 2015;IWMI, 2017;Jaksa et al., 2013), and the United States (Vahedifard et al., 2016;Vahedifard et al., 2017).This happens mostly in lowland areas with dikes made of peat and clay or on peat/clay soils that are prone to cracking during drought.
Moreover, drought-to-flood events can increase impacts by degrading water quality and creating favorable conditions for the development of diseases that impact human health.In our case review, we find that impacts on human health were most often reported in low-income countries (Figure 4).We also found examples in the scientific literature.For example, Effler et al. (2001) showed that the Escherichia coli outbreak in Swaziland and South Africa in 2000 was preceded by intense precipitation following a three-month drought period.It was shown that high concentrations of pathogens occurred during the drought period because livestock used human water sources.This increased human exposure during consequent heavy rainfall (Levy et al., 2016).Drought-to-flood events can also result in large-scale simultaneous hatching of mosquito eggs.This may lead to the transmission and outbreaks of, for example, Rift Valley fever virus (Stanke et al., 2013).In addition, these events can alter water quality due to deposition of pollutants within the soil during the drought period and their consequent discharge into the river during flood events (Mishra et al., 2021).Moreover, consecutive dry and wet periods result in elevated phosphorus release and consequent water quality degradation (Laudon et al., 2005).Alteration of river nutrient concentrations can cause severe fish mortality (Laudon et al., 2005) and eutrophication of surface water bodies (Wurtsbaugh et al., 2019).

| Drought impacts on the socio-economic system
Impacts of drought on the physical system (e.g., water supply infrastructure) can also impact water supply, leading to water insecurity.Drought impacts on agriculture and public drinking water supply were most often reported (Figure 4).Drinking water supply is often disrupted during drought, especially if communities or households rely on local sources that are not connected to a network (Mullin, 2020).This may not only be due to lower water levels, but also to more sediments in the water intake, more breakage of handpumps, and more damage to pipes because of soil subsidence (Thomas et al., 2020;Wlostowski et al., 2022).Drinking water companies then need to repair pumps and pipes and/or invest in finding other sources of drinking water, thereby depleting financial reserves (Koehler et al., 2018).For example, in the 2011-2017 drought in California (USA), many community water services requested emergency funding from the state to maintain water delivery to low-income populations (Mullin, 2020).This left drinking water supply services in a vulnerable position in case a subsequent flood destroys drinking water infrastructure (Fekete, 2019;Njogu, 2021).For women and girls in rural communities in low-income countries, walking distance to drinking water sources increases during drought (Arku & Arku, 2010;MacAllister et al., 2020).This creates many cascading impacts, for example, increased exposure to sexual harassment, missing school, and less time to work on the land to provide food for the family or to work on other sources of income (Tallman et al., 2023).All of these can potentially increase vulnerability to flooding (and other hazards).
The percentage of cases of a certain drought duration (1 year vs. multiple years) reporting a certain drought type, separated by aridity group.Meteorological drought was reported in all cases, whereas soil moisture, surface water and groundwater drought are reported in a decreasing percentage of the cases.There is a clear difference between the propagation for multiyear and 1-year events.
Drought impacts lead to changes in vulnerability and increase the susceptibility to subsequent flooding.This occurs through multiple processes.Multiyear droughts cause long-term impacts such as physical and mental health issues, long-term financial struggles, lack of education, erosion of social coherence, and increase in social conflicts (e.g., Matan o et al., 2022;Sena et al., 2017).This is especially the case in low-income countries.Sena et al. (2017), for example, found that in Brazil health and well-being are lower in regions that experience drought more regularly than the rest of the country.This implies that drought increases vulnerability.In contrast, repeated extreme flooding has been found to reduce vulnerability in subsequent flood events due to enhanced awareness and adaptation (Kreibich et al., 2017).These studies focus on multiple occurrences of the same hazard, but to understand the social processes behind drought-to-flood events we need to understand cascading impacts and changes in vulnerability between different hazard types (De Ruiter & Van Loon, 2022;Siegel et al., 2003).One study by Rockström (2003) mentions how drought-induced unemployment leads to increased financial struggles during flood, but more research is needed on cascading vulnerabilities.
A combination of impacts on agriculture and livestock, from both the drought and the flood event is reported quite often and most frequently in low-and lower middle-income countries (see Figure 4).The vulnerability of livestock to flooding may be increased by a preceding drought.This occurred in Queensland in 2019 (Cowan et al., 2022), where cattle were weakened due to a lack of food during a drought.This increased their vulnerability during the following flood, causing many of them to die from exposure.Vegetation conditions can be an important factor in drought-to-flood events.Agriculture that has suffered during drought can be further impacted by extreme rain events.Cobon et al. (2016) found that, in coffee plantations, alterations of extreme dry and wet periods reduce coffee bean size.Droughtto-flood can have severe effects on crop yield, based on the intensity of the consecutive hydrological extremes, soil conditions, and crop growth stage (Gao et al., 2019;McCarthy et al., 2021).Impacts on agriculture cascade further to health, financial and societal impacts.Thereby potentially increasing the vulnerability of the population.If the drought is followed by a flood, food availability could decrease even further and access to water, food and health care could be reduced even more (Matan o et al., 2022).
These effects are not uniform throughout society.Natural hazards do not affect people equally (Neumayer & Plümper, 2007) and differences in societal vulnerabilities can exacerbate the inequality of impacts of consecutive droughts and floods.Low-income groups are more at risk of damages caused by hydrological extremes due to structural inequalities (Andrijevic et al., 2020).Moreover, social groups are impacted differently by hazards because of the differences in access to resources for preventing, mitigating, or recovering from extreme events (Masozera et al., 2007).As a result, groups that possess resources for adaptation against one hazard type can become less vulnerable to another consequent hazard, compared with groups that do not have such means.In addition, the recovery of people and economies after a drought ends is important.Societies that quickly bounce back or even forward will be less vulnerable to a next F I G U R E 3 Percentage of cases with a certain drought type also reporting fires (a) and percentage of cases with and without reports of fire reporting land-or mudslides (b).If a case reports a groundwater drought it becomes more likely that there are also reports of fires.In case there are reports of fire, it is more likely that there are also reports of mud-or landslides.
event than societies that recover slowly (Di Baldassarre et al., 2018).The latter are more prone to collapse when faced with subsequent extreme events (Kuil et al., 2016;Weiss & Bradley, 2001).The differences between groups (or countries) with different amounts of resources is also reflected in our case review.While we do not find evidence of specific drought impacts causing an increase in a specific flood impact, we do find evidence of income levels as a common driver to the type of impacts that are reported most in a country.Figure 4 shows that drought impacts on public water supply, energy/industry and ecosystems are more often reported in higher-income countries (the darker colored upper right and lower right quadrants) than in lowerincome countries (the lighter colored lower left and upper left quadrants).This coincides with a higher reporting of the flood impacts of damage to properties and road infrastructure and more reported evacuations.In contrast, damage to agriculture as a flood impact is more often reported in lower-income countries and often coincides with reported agricultural impacts from drought (which is generally high across all income groups).In addition, drought impacts on human health and public safety are mostly reported in low-income countries and they often coincide with a high frequency of reported loss of life, as well as damage to agriculture and/or livestock due to flooding.
Percentage of cases per income group reporting both a certain drought impact (x-axis) and flood impact (y-axis).The four quadrants represent the four income groups.An empty quadrant means zero % of the cases in the corresponding income group report this combination of impacts and a full quadrant means 50% of the cases report this combination of impacts.Note that for the first column of drought impacts "Agriculture and livestock farming" a full quadrant means 100% of the cases report this impact combination.This was done for visualization purposes.
In this section, we discuss how human adaptation to drought can affect flood risk.Besides leading to more impacts and increased vulnerability (Section 3), long, multiyear droughts allow ample time for drought responses, management, and adaptation (Watts et al., 2012).Human adaptation processes can affect the catchment processes and impacts discussed in the previous sections, both positively and negatively.Responses to drought that increase flood risk are sometimes called maladaptation (Adger et al., 2005;Ward et al., 2020).Ward et al. (2020) provide an extensive review of how drought adaptation measures can influence flood risk (and how flood adaptation measures can influence drought risk).Here, we provide a summary of their findings and compare them with the evidence from the case review.

| Adaptation processes affecting hydrological processes
As mentioned in Section 2, increased groundwater abstraction in response to a drought can potentially reduce flooding.Groundwater abstraction is often increased during drought, which depletes storage and potentially slows drought recovery (Apurv et al., 2017;Wendt et al., 2021).This makes drought-to-flood events less likely, depending on the local geology and the type of flood event that follows the drought.Groundwater abstraction is one of the drought adaptation measures that was reported most often (Figure 5).It is mostly reported in higher-income (i.e., both high-and higher middle-income) countries (Figure 6).Ward et al. (2020) discuss how increased groundwater abstraction can also increase the risk of flooding, because it can lead to subsidence.Stopping abstraction can also lead to flooding, for example, in the case of the foggaras (a traditional irrigation system) of Bouda in Algeria.An initial increase of groundwater extraction through boreholes caused the foggaras to run dry and their long disuse caused a state of disrepair.When the F I G U R E 5 Number of cases reporting a drought adaptation measure.In total 192 drought-to-flood events were included in the case review.
groundwater pumping through boreholes was reduced, the foggaras started flowing again.However, they could not handle the amount of water due to their state of disrepair and this caused flooding of sebkhas and palm groves (Boutadara et al., 2018).The opposite of groundwater abstraction, managed aquifer recharge, or the construction of other water storage infrastructure, such as sand dams, can be beneficial for both drought and flooding (Ward et al., 2020).The construction of new water infrastructure in response to a drought was found in several cases (Figure 5).There are several examples of drought adaptation measures that can have a negative effect on subsequent flooding by influencing catchment processes.Examples are reservoir operation strategies, agricultural practices and land use change (Ward et al., 2020).

| Adaptation measures affecting impacts
Ward et al. ( 2020) also discuss how awareness and risk perception of flooding can be influenced by drought experience, and how a focus on drought preparedness can decrease the preparedness for flooding.These processes can increase the vulnerability to flooding.There are also several drought adaptation measures or coping strategies that increase exposure to flooding, such as migration (Ward et al., 2020).Flood exposure could be increased by drought-induced migration resulting in people living in floodplains (FGS, 2018), but migration is complex and not always directly attributable to drought (Black et al., 2013).However, it was reported as a drought response in some instances in the case review (Figure 5).Another example of a drought coping mechanism that results in a higher exposure to flooding is when farmers decide to delay planting until the rains start.When the rains do start, but are extreme, they may wash away seeds or make sowing impossible.This happened, for example, in Bundhelkhand, India (Pateriya, 2016).Similarly, the uptake of drought resistant crops which may be vulnerable to heavy rain can worsen flood impacts (Tirado & Cotter, 2010).In addition, food aid reducing outmigration in drought-struck and flood-prone areas (Salite & Poskitt, 2019) is another measure that can increase exposure to flooding.We found that in low-income countries, food aid or cash aid is the most reported adaptation measure (Figure 6).Government and NGO support could be exhausted during the drought, meaning that limited aid would be available during a subsequent flood (Matan o et al., 2022).On the other hand, responses that aim to reduce vulnerability to drought can also be beneficial during floods.Mavhura (2019) found, for a case in Northern Zimbabwe, that vulnerability to flood and drought was influenced by the same drivers, such as low incomes or few savings.Measures and policies that address these drivers, would be beneficial for reducing both flood and drought vulnerability.
Another adaptation measure frequently reported is water rationing or the prioritizing of certain users (Figure 6).This measure is more often reported in high-income countries and semi-arid countries.The latter could be because these places may be more experienced with dry conditions and have regulations in place for the rationing and prioritizing of water use in case of a drought.However, there is also a high percentage of cases with a humid climate reporting this measure, which may indicate that high-income is a more important factor.This could be related to the fact that high-income countries more often have a regulated public water supply system in place.These regulations may lead to unequal impacts because water users may face different levels of reduction in water supply This depends on the legislation, prevailing water rights and social inequalities (Lund et al., 2018;Savelli et al., 2021).Therefore, different water users may experience different levels of increases in vulnerability, which may influence their ability to cope with a subsequent flood.These unequal effects hold for adaptation measures in general, just like for impacts.Adaptation measures benefit some groups more than others and responses that are beneficial for one group may not be for another group (Masozera et al., 2007;Savelli et al., 2021).In California, for example, cuts in water supply did not severely affect the agricultural sector, who used groundwater as an alternative.However, the combination of cuts in water supply and the effects of adaptation measures taken by the agricultural sector (such as groundwater extraction) led to increased vulnerability of ecosystems and the people that rely on them (Christian-Smith et al., 2015).Another example of increased vulnerability of ecosystems because of human adaptation is the case of Guadiana wetlands in Spain, where impacts of groundwater abstraction on ecosystems were found to be four times higher than the impacts of drought ( Van Loon & Van Lanen, 2013).
In general, risks are experienced differently and lead to varying adaptation strategies, depending on the management structure (e.g., public, private, or community responses) and the type of hazard (e.g., drought or flood).In a review on pre-disaster planning and preparedness for droughts and floods, Raikes et al. (2019) found that drought management tends to be responsive, while flood management is more risk-based and includes prevention and preparation.In addition, some countries may have a more centralized approach to flood and drought risk management, such as the Netherlands and Poland.In other countries, such as the United States, risk management is increasingly privatized (Raikes et al., 2019).This affects the type of adaptation measures that may be adopted.

| Main findings
This review has provided an overview of the processes and feedbacks that are important in drought-to-flood events.We find that drought can increase subsequent flood occurrence and severity due to increased surface runoff, but it can also decrease subsequent flooding due to storage depletion.There is little evidence as to which process may generally be dominant.Storage depletion processes seem to be more prevalent in arid places, as well as in broader and flatter valley catchments and places with deeper soils or more groundwater variability.This also depends on the type of flood that occurs during or after the drought event.For high-intensity rain events and flash floods, storage depletion processes are less important.Both increased runoff and storage depletion processes seem to mostly play a role when the drought is a multiyear drought and when the flood occurs during the drought (e.g., the drought is not ended by the flooding).Human activities seem to enhance both processes in human-dominated catchments, with increased storage depletion leading to lower floods after droughts and land-surface changes leading to higher floods after drought.
In terms of impacts, there is clear evidence that drought-to-flood events cause more impacts than if the hazards would occur on their own.Examples are breaking levee systems and drought-to-flood related water quality problems.In general, the socio-economic processes that underlie drought-to-flood events seem to be mostly related to changes in vulnerability.The drought may cause increases in the vulnerability of people, crops or livestock, which causes impacts of a subsequent or co-occurring flood to be worse than from a flood event that is not preceded by a drought.This depends on the initial vulnerability and how quickly communities are able to recover.Impacts and changing vulnerability are not the same for everyone.Which process is dominant where, when, and for whom, is strongly dependent on the context of the drought-to-flood event.Characteristics like climate, geology, land use, and the socio-economic, cultural, and political context determine the impacts and responses and therefore the interactions between drought and flood impacts.F I G U R E 6 Percentage of cases per income group (top) and aridity group (bottom) reporting a certain drought adaptation measure.Food and cash aid are reported more often in lower income countries while water rationing and groundwater pumping are reported more often in higher income countries.Water rationing is reported slightly more often in semi-arid countries.
The case review has shown that there is a huge diversity in the type of adaptation measures that are implemented.Several adaptation measures are focused on reducing the overall vulnerability of the population, which makes them beneficial for both flood and drought risk management.In addition, there are several measures that focus on storing excess water, which can reduce both drought and flood risk (e.g., managed aquifer recharge, sand, and earth dams).Groundwater abstraction (reported most often in higher income and arid places) can have a positive effect on flood risk, by increasing storage space, but can also increase subsidence.Finally, the awareness or memory of flood risk may decline due to a drought, which can negatively affect flood risk.Adaptation processes in response to a drought can affect socio-economic processes in several ways, changing the risk perception, exposure and vulnerability to flooding.

| Outlook
We argue that more research is needed on drought-to-flood events.In particular, the catchment characteristics and the socio-economic, cultural, and political context that determine which drought-to-flood processes and feedbacks are important should be further investigated.In terms of hydrological processes, it would be interesting to further explore when and where storage depletion and runoff processes are dominant and how human interventions affect this.There is still a lot unknown about processes related to impacts.Further research is needed into whether certain drought impacts make certain flood impacts more likely, or into whether there are common drivers of these impacts and when and where these occur.In addition, changes in vulnerability related to drought-to-flood events require further investigation (De Ruiter & Van Loon, 2022).This review does not provide much evidence of which adaptation measures are taken when and where and how this affects flood risk.Characteristics of the social and hydrological system that influence the feasibility and effectiveness of adaptation measures, as well as their influence on flood risk could be further investigated.In addition, adaptation measures may be beneficial for some while they are harmful to others.When and where this is the case is something that could be addressed in future research.
Methods that can increase our understanding of the specific hydrological and social processes happening during drought-to-flood events include detailed case study analyses, as suggested by Mostert (2018).Qualitative case studies can also be compared to find generalizable results, especially if and when a larger body of case study analyses becomes available, for example using qualitative comparative analysis (Srinivasan et al., 2012).System characteristics (both social and hydrological) that lead to certain processes being dominant, or certain adaptation measures being feasible or not, can be investigated using large-scale comparative studies.This approach has already been applied in hydrological studies to, for example, investigate which characteristics influence the seasonality and magnitude of maximum annual flows (Berghuijs et al., 2016).These qualitative and quantitative case studies can be combined to both create new insights and report the findings to different audiences (e.g., Grainger et al., 2016).Finally, management alternatives can be explored using models that incorporate both the hydrological and social processes (e.g., Mazzoleni et al., 2021).Using these models, scenarios can be developed and explored.This can be done through the exploration of adaptation pathways (Werners et al., 2021), where not only the effects of management and adaptation choices on drought risk are taken into account, but also the effects on flood risk.
Improving our knowledge on drought-to-flood events and their interactions will help in designing more robust measures that do not have adverse effects on the opposite risk, now or in the future.The necessity for the analysis of system-wide, short-term and long-term implications is already recognized for both drought and flood risk management separately.Here, we argue that this should also be expanded to the study and management of drought and flood risk together.Identifying drought-to-flood processes and their characteristics would help in the design of robust measures that hold under future risk scenarios.In addition, it would help identify opportunities for reducing both drought and flood risk at the same time, through the reduction of overall vulnerability or by implementing measures such as managed aquifer recharge.Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Johanna Koehler: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Marleen De Ruiter: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Margaret Garcia: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Maurizio Mazzoleni: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Jeroen Aerts: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Philip Ward: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Giuliano Di Baldassarre: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Rosie Day: Conceptualization (supporting); writingoriginal draft (equal); writingreview and editing (equal).Anne F. Van Loon: Conceptualization (lead); formal analysis (supporting); funding acquisition (lead); writingoriginal draft (lead); writingreview and editing (lead).
Percentage of cases reporting a certain flood type differentiated by drought duration (a), and percentages of cases reporting a certain flood type differentiated by flood timing during or after the drought (b).Highlighted in red are the event types (flash flood and pluvial flood) that occur more often in relation to a multiyear drought than to a 1-year drought (a), and the event types (flash flood and flash and pluvial flood combined) that occur more often during than after a drought (b).