Hydrological Drought‐To‐Flood Transitions Across Different Hydroclimates in the United States

Floods following on streamflow droughts can have severe impacts. While they have been prominently featured by the media in recent years, we know little about their spatio‐temporal variability. In this study, we analyze the occurrence and drivers of such drought‐to‐flood transitions by calculating transition lengths from droughts to floods for natural and regulated catchments across the Contiguous United States between 1970 and 2022. We find that drought‐to‐flood transitions strongly vary in their lengths and their spatial distribution. We identify snowmelt as the main driver of transitions in high‐elevation catchments, while transitions in low‐elevation catchments are more variable in their time of occurrence and drivers. Reservoir management reduces the number of short drought‐to‐flood transitions, particularly in catchments with a high amount of snow where snowmelt is crucial for filling reservoirs in early summer. These findings suggest that projected changes in the snowmelt season will lead to changes in transitions from streamflow droughts to floods and that reservoir management may be used to adapt to these changes.


Introduction
Droughts and floods have severe impacts on society, economy, and ecosystems (e.g., Cammaleri et al., 2020;Hammond et al., 2015;Jones & Van Vliet, 2018;Mosley, 2015;Tradowsky et al., 2023), which can be particularly pronounced if they occur in close succession, that is, when droughts rapidly transition to floods (Barendrecht et al., 2024;Ward et al., 2020).The negative impacts of drought-to-flood transition events are illustrated by recent examples such as the floods in winter 2016-2017 following on the multi-year drought of 2011-2016 in California (He & Sheffield, 2020;Swain et al., 2018), the drought-to-flood transition in Kenya in 2017-2018 which led to a humanitarian crisis (Matanó et al., 2022), and the May flood in 2023 in the Italian region Emilia-Romagna which followed on a prolonged dry period (Povoledo, 2023).
Such consecutive drought-to-flood events pose a challenge to emergency and water management because droughts and floods require different water management strategies.When managing opposing hydrological extremes, drought protection measures such as long-term water storage have to be weighted against short-term flood control measures (Di Baldassarre et al., 2017;He & Sheffield, 2020;Mazzoleni et al., 2021).This management challenge is especially present for large reservoirs, which provide water supply and flood protection alike (e.g., Wivenhoe Dam near Brisbane, Australia (Van den Honert & McAneney, 2011)) and can strongly impact hydrological extremes.While reservoirs can play an important role in flood protection by attenuating the flood hydrograph when reservoir levels are low, for example, during a drought (Valdes & Marco, 1995;Ward et al., 2020), optimization for one type of extreme may have adverse impacts on the other type of extreme.For instance, increasing water security implies keeping more water stored in the reservoir which at the same time reduces its capacity to absorb flood peaks (Wu et al., 2023).Considering regulation effects is especially relevant in regions such as the United States where rivers have been increasingly regulated and fragmented during the last century (Spinti et al., 2023;Steyaert et al., 2022).Brunner (2021) has shown for the United States that the presence of reservoirs influences drought and flood occurrence and characteristics, which suggests that reservoirs may also influence the transition time from droughts to floods.
While the impact of reservoir regulation on individual drought and flood events is undisputed, it is yet unclear how such regulation affects the transition from droughts to floods.Drought-to-flood transitions can be harmful events, where the extreme events may amplify each other's impact (Barendrecht et al., 2024).For example, flood protection measures can be weakened by the occurrence of a drought, for example, due to an increased risk of dike failure or a decreased infiltration capacity of soils (Ward et al., 2020).Drought-to-flood transitions may also lead to a deterioration of water quality, when accumulated nutrients in dry areas are mobilized and transported by floods (Mosley, 2015;Pulley et al., 2016).Overall, the impacts of drought-to-flood transitions can be particularly severe in regions where the socio-economic vulnerability is high (Matanó et al., 2022).Despite the potentially severe impacts of drought-to-flood transitions, floods and droughts are often studied separately from each other.
In the rare cases where droughts and floods have been jointly considered, they have mostly been studied from a meteorological perspective by looking at transitions from precipitation deficits to wet spells on large spatial and temporal scales.For example, De Luca et al. (2020) found that wet-to-dry event transitions are usually slower than dry-to-wet event transitions at a global scale and a coarse spatial resolution (2.5°).Other studies focused on dryto-wet or wet-to-dry transitions in individual regions, such as Northern China (Shi et al., 2021) and South Asia (Ansari & Grossi, 2022).Both studies found strong spatial differences in the occurrence of these transitions.These existing studies focus on precipitation only and lack a hydrological perspective.This hydrological perspective is an important one because the occurrence of hydrological extremes is not only dependent on meteorological drivers, but also depends on land-surface characteristics including topography, antecedent conditions and water storage such as snow cover or reservoirs (e.g., Berghuijs et al., 2016;Brunner, Gilleland, et al., 2020;Van Loon et al., 2015;Van Loon & Laaha, 2015).
While hydrological drought-to-flood transitions and their drivers remain understudied, the atmospheric processes leading to drought-termination have been discussed in the literature.Dettinger (2013) has shown that droughts along the US West Coast are often ended by a single wet month or storms caused by atmospheric rivers, while Holgate et al. (2020) have shown that droughts in Australia are often terminated by marine moisture contributions.The termination of droughts is also influenced by human activities, for example, water abstraction, which can prolong drought-termination (Margariti et al., 2019).Parry et al. (2016) have focused on streamflow droughts and considered drought termination as the period from a streamflow deficit to normal streamflow conditions.They acknowledged though, that even long droughts can end rapidly due to short but strong precipitation events and subsequent streamflow extremes.Floods as drought terminators can be hazardous but have not yet received much attention.Specifically, a large-sample study focusing on floods following on streamflow droughts is still missing.Therefore, it is largely unknown which catchments are most frequently affected by hydrological drought-to-flood transitions and which catchment characteristics favor their occurrence.
The development of sustainable adaption measures against drought-to-flood transitions requires an improved understanding of their frequency and time of occurrence.In this study, we therefore explore the frequency and seasonality of drought-to-flood transitions across different hydroclimatic zones and for natural and regulated catchments in the Contiguous United States (CONUS).Specifically, we ask "when, where and how often do transitions from streamflow droughts to floods occur?."To answer this question, we first identify drought and flood events in a large-sample streamflow data set over the CONUS.Then, we calculate transition times between droughts and floods as the time elapsing between the end of a drought and the beginning of a flood and link these transition times to different hydroclimatic parameters as well as the regulation status of the respective catchments.By differentiating between natural and regulated catchments, we assess if, where and when reservoir regulation has an effect on the occurrence of drought-to-flood transitions.Such human influences on drought-to-flood transitions can only be quantified by taking a hydrological instead of a meteorological perspective, which highlights the added value of taking a hydrological perspective in compound event studies.

Data and Study Region
The Contiguous United States show a large variability in topographic and climatic features, leading to a wide range of hydrological regimes (Addor et al., 2017), which makes them an ideal study area to analyze the different drivers of drought-to-flood transitions.We selected catchments for which streamflow data are recorded through the United States Geological Survey (USGS), which have at least 90% of data available between 1970 and 2022, are included in the GAGES-II database (Geospatial Attributes of Gages for Evaluating Streamflow) (Falcone, 2011), and were either classified as "HCDN" or "Regulated" by Ryberg et al. (2020).Catchments classified as "HCDN" (also referred to as "natural") are part of the USGS's Hydro-Climatic Data Network (HCDN), which means that they are only marginally influenced by human-activities (Lins, 2012).Catchments classified as "Regulated" show a high degree of regulation and low urbanization (Ryberg et al., 2020).The 1,002 catchments selected are almost equally split between being classified as natural and regulated, with 519 regulated and 483 natural catchments, respectively.The GAGES-II database provides further information about catchment topography and climatology, such as information on elevation, temperature or the amount of precipitation falling as snow.We have chosen to analyze the same time period of 1970-2022 for all discharge stations because it is characterized by a good data availability over many stations.For each of these 1,002 catchments, we downloaded daily streamflow time series from the USGS database (USGS, 2022).

Definition of Droughts, Floods and Drought-To-Flood Transitions
We define a "drought-to-flood transition" as the subsequent occurrence of a flood after a hydrological drought, here also referred to as streamflow drought.Before identifying such transition events, we identify drought and flood events for each catchment in our data set (Figures 1a and 1b).For droughts, we use a variable threshold approach which has been recommended for catchments with a seasonal flow regime (Van Loon & Laaha, 2015) and has been applied in various hydrological studies (e.g., Brunner et al., 2023;Heudorfer & Stahl, 2017;Van Huijgevoort et al., 2012;Van Loon et al., 2015).Before calculating the variable drought threshold, we calculate the 30 days moving average of the flow to avoid identifying minor drought events.Then, we calculate the threshold by using the 20th flow percentile of the averaged flow for each day of the year, considering also the 15 days prior and after the day of interest (e.g., Brunner et al., 2023).We define drought events as periods where the averaged streamflow undercuts the calculated variable threshold.For each drought event, we calculate its start date, end date, duration and flow deficit.We define the flow deficit as the accumulated difference between the averaged flow and the threshold from the beginning to the end of a drought event.To focus on events relevant for drought management, we constrain the number of droughts by only including events which (a) have a minimum duration of 30 days (e.g., Brunner et al., 2023) and (b) have a deficit larger than zero.We include the latter constraint to avoid identifying droughts in periodic rivers at times of the year when they have fallen dry in most years.These zero-flow events do not represent streamflow anomalies and are hence here not considered as hydrological droughts.The drought identification approach used here results in 0.68 events per catchment per year on average (first quartile: 0.58 events, third quartile: 0.77 events) (Figure S3a in Supporting Information S1).
We define floods as events where discharge exceeds a certain absolute threshold, which is defined as the 50th percentile of annual maximum discharge in each catchment (Figure 1b).We identify floods as events where the daily discharge exceeds this threshold.This approach is similar to the approach used by Brunner, Gilleland, et al. (2020) and uses a minimum time lag of 10 days between events to guarantee event independence (Diederen et al., 2019).With this approach, we obtain about 0.74 flood events per catchment and year on average (first quartile: 0.66 events, third quartile: 0.83 events) (Figure S3b in Supporting Information S1).Last, we identify drought-to-flood transitions as events when a flood follows on a drought, independently of whether the two events are causally related or not (Figure 1c).We only consider events where a flood event follows on a drought, without Figure 1.Overview of the event and season definitions used in this study: (a) Droughts defined using a variable drought threshold, (b) floods defined using a fixed flood threshold, (c) transition times defined as the time between the end of a drought and the beginning of a flood event, (d) high-and low-flow seasons defined as the 91 consecutive days with the highest and lowest discharge, respectively and (e) summer and winter seasons defined as the periods between May and October and November and April, respectively.

10.1029/2023WR036504
another extreme interrupting the sequence.If, for example, two droughts follow on each other before a flood occurs, only the last event pair is counted as a drought-to-flood transition.We calculate the transition time of each drought-to-flood transition as the time between the end of a drought and the onset of a flood in days.

Analysis of Drought-To-Flood Transitions
We first look at transition times from droughts to floods for four example catchments with different streamflow regimes to get an impression of the transition-time variability within a catchment.Second, we distinguish between transition times of different lengths and introduce the terms "rapid transitions" and "seasonal transitions" for drought-to-flood transitions with a length of up to 14 days and up to 90 days, respectively, to study spatial variations in transition times and their time of occurrence.Third, we analyze the occurrence of rapid and seasonal transitions for different seasons and their dependence on hydroclimatic factors, such as the amount of precipitation falling as snow.We compare transitions for the summer and winter seasons, which we define as May-October and November-April, respectively, because CONUS is located in the Northern hemisphere (Figure 1e).Furthermore, we differentiate between the low-and high-flow season (Figure 1d) to discuss potential impacts of drought-to-flood transitions assuming that flooding in the low-flow season is usually less expected and has hence a larger effect of surprise than floods happening in the high-flow season.We define the low-and highflow seasons for each catchment as the 91 consecutive days with the lowest and highest average flow, respectively, during the period of 1970-2022 (Satoh et al., 2022).Last, we analyze the effect of reservoir regulation on drought-to-flood transitions by comparing the number of transitions occurring in natural versus regulated catchments.

Duration of Drought-To-Flood Transitions
The calculated transition times vary widely between 1 and 2479 days (6.79 years) over all natural catchments (median = 122 days, mean = 191 days, 1st quartile = 42 days, 3rd quartile = 263 days).Most catchments have experienced both very short and very long drought-to-flood transitions.This large within-catchment variability is observed across all hydroclimatic zones in the study area as illustrated by the distribution of transition times for four example catchments (Figure 2).That is, all example catchments show a large spread of transition times between 1 day and 1,365 days.To study the variability of transition times across catchments, our further analyses focus on the number and seasonality of transitions of a certain length and we introduce the terms rapid (≤14 days) and seasonal transitions (≤90 days), respectively.Instead of computing average transition times that are not very representative of the overall transitions behavior of a catchment, we count the number of rapid and seasonal transitions and assess the seasonality of these two types of transitions to characterize the transition behavior of a catchment.

Frequency and Seasonality
The number of drought-to-flood transitions of different lengths varies substantially among catchments and does not show any clear spatial pattern, neither for natural nor regulated catchments (Figures 3a and 3b).Natural catchments have experienced 2 rapid drought-to-flood transitions over the study period (53 yrs) on average, with 15.7% of the catchments showing no rapid transitions and only 2.2% of the catchments showing more than five rapid transitions (Figures 3c and 3d).The timing of these rapid transitions varies regionally and shows spatial patterns (Figure 4).Rapid transitions at the West Coast occur mainly from fall to winter; in the Rocky Mountains from late spring to summer; and in the mountainous Northern US and the North East in spring.The Great Plains show a large variability in the seasonality of the occurrence of rapid transitions, with rapid transitions occurring all year round except in winter.At the East Coast and in Southern Florida, rapid transitions occur primarily in fall.The timing of rapid drought-to-flood transitions appears to be related to elevation (Figure 4b).Catchments with a median elevation below 2,000 m (a.s.l.) are characterized by average rapid drought-to-flood transition timings between February and mid-April.Above 2,000 m, the average transition timing is between May and mid-June.While the mean timing changes with elevation, the standard deviation of the transition timing varies by up to 3 months.At high elevations (>2,000 m), the standard deviation of the timing of rapid transitions is substantially smaller, but so is the sample size.

Low-and High Flow Season
The potential impacts of drought-to-flood transitions are expected to differ depending on their time of occurrence.That is, a drought-to-flood transition in the low-flow season is most likely a transition from a low-flow to a highflow state with potentially pronounced impacts, while drought-to-flood transitions during the high-flow season do not necessarily have to involve low-flow states, which is why they might have fewer impacts.Therefore, we further analyze the seasonality of transition events and differentiate between drought-to-flood transitions occurring during the typical high-flow and low-flow season (see Section 2.3 and Figure 1d).
Rapid drought-to-flood transitions have occurred more often during the high-flow than during the low-flow season (1.17 and 0.19 events per catchment on average, respectively) (Figure 5).66% of the natural catchments have experienced at least one rapid drought-to-flood transition during the high-flow season but only 15% of the catchments have experienced rapid transitions during the low-flow season.In the high-flow season, most rapid transitions have occurred in the North, the Rocky Mountains, the South and the North East (Figure 5a).Many catchments in these regions have experienced more than one rapid drought-to-flood transition, with the largest number of rapid transitions occurring in the North East of the US.In contrast, hardly any transitions have occurred during the low-flow season except for some catchments in the East, Pacific Northwest and South (Figure 5b).

Seasonal Drought-To-Flood Transitions
The analysis of the timing of rapid transitions presented in Section 3.2.2highlights that catchments at highelevations have a more distinct seasonality in the occurrence of drought-to-flood transitions than low-elevation catchments.Such high-elevation catchments are usually characterized by a substantial amount of precipitation falling as snow (here called snow-dominance), which suggests a relationship between the snow characteristics and drought-to-flood transition behavior of a catchment.Therefore, we want to further assess the relationship between snow influences and drought-to-flood transitions by considering the number of transitions in addition to   transition seasonality.To establish a link between the fraction of snow of a catchments and the number of droughtto-flood transitions it experiences, we move from analyzing rapid transitions (≤14 days) to seasonal transitions (≤90 days) which increases the sample size from 2 to 5.9 transition events per natural catchment on average.The number of seasonal transitions per catchment ranges from 1 to 13 across CONUS.Notably, more seasonal transitions occur in mountainous regions near a coast, for example, in the North East and North West (Figures 6a  and 6b).The number of seasonal transitions is slightly lower in catchments with a median elevation between 2,000 and 3,250 m in both, natural and regulated catchments, compared to catchments at other elevations, but does overall not show a strong relationship with elevation (Figures 6c and 6d).
The occurrence of seasonal transitions in natural catchments is linked to their snow-dominance, that is, the fraction of precipitation falling as snow (Figure 9a).In summer, catchments with a low snow-dominance show a small number of seasonal transitions (Figure 9a).This number increases once the snow-dominance rises above 40%.The number of seasonal transitions is highest in catchments with a snow-dominance of over 60%, with a median of 5 seasonal transitions per catchment, which corresponds to roughly one seasonal transition every 10 yrs.In winter, seasonal transitions occur most frequently in the less snow-dominant catchments, with a peak of a median of 7 seasonal transitions in catchments with a snow-dominance between 30% and 40% (Figure 9b).With increasing snow-dominance, the occurrence of transitions decreases rapidly during winter (Figure 9a, lower panel).Most catchments with fractions of snowfall between 60% and 70%, that is, the class of catchments that showed the highest number of seasonal transitions during summer, have not recorded any seasonal transitions during the winter months.
We also linked the number of seasonal transitions to several other catchment characteristics than the fraction of snow, such as aridity, yearly precipitation, average temperature or mean catchment slope.While the link between snow-dominance and the number of transitions is very clear, we could not find any relation between these other parameters, or a combination of different parameters, and the frequency of occurrence of drought-to-flood transitions (Figure S6 in Supporting Information S1).

Reservoir Influences
Catchments influenced by reservoirs show fewer transition events than natural catchments both in the case of rapid and seasonal transitions.Specifically, they experience fewer rapid transitions than natural catchments (natural: 2, regulated: 1.6 rapid transitions, Figures 3b and 3d), with 54% of all regulated catchments showing fewer than two transitions in the study period compared to 42% of all natural catchments (Figures 3c and 3d).The seasonality of rapid transitions is less expressed in regulated than in natural catchments (Figures 4 and 7).
Especially, catchments between 1,500 and 2,750 m of elevation have a weak seasonality, with transitions   4b and 7b).At elevations above 2,000 m, the variability of timing is particularly high in regulated catchments (58 days) compared to natural catchments (30.6 days).The number of rapid transitions is also reduced in regulated catchments as compared to natural catchments, with on average 0.87 instead of 1.17 rapid transitions during the high-flow season and 0.18 instead of 0.19 rapid transitions during the low-flow season (Figures S5 in Supporting Information S1 and Figure 5).Rapid transitions show a weaker relationship with snow dominance in regulated catchments than in natural catchments (compare Figures 8a and 8b).
The number of seasonal transitions is 16% higher in natural than in regulated catchments (natural: mean 5.9, regulated: mean 5.1 seasonal transitions); difference significant according to the exact Poisson test (α = 0.05, R Core Team ( 2023)) (Figure 6, comparison between panels a and b).In summer, regulated catchments across all snow-dominance levels show a similar number of seasonal transitions (1-3 events) (Figure 9b).The fewest seasonal transitions occur in regulated catchments with a moderate snow-dominance (30%-50% fraction of snow).The distribution among different snow-dominance levels in regulated catchments in summer is overall much more similar in regulated than in natural catchments (Figure 9).During winter, regulated catchments with a medium level of snow-dominance between 10% and 30% show the most seasonal transitions (median values of 3-5 events).The pattern between the number of seasonal transitions and the snow-dominance is similar between natural and regulated catchments for winter, with an overall reduced number of seasonal transitions in regulated catchments.

Drought-To-Flood Transition Frequency Varies Widely in Space and Is Linked to Snow Fraction
In this study, we have analyzed the occurrence of streamflow drought-to-flood transitions in the Contiguous United States and identified snowmelt as the main driver of drought-to-flood transitions in high-elevation catchments.The number of drought-to-flood transitions is highly variable between catchments (Figures 2 and  4b) and only shows weak relationships to specific catchment characteristics despite some spatial patterns (Figure 5a).Similarly, the seasonality of transitions also varies substantially in space.However, in contrast to the spatial variability in the number of transitions, the variability in the seasonality of transitions shows more expressed spatial patterns, which are closely linked to the seasonality of flooding (Villarini, 2016) and to catchment characteristics, in particular the fraction of snow (Figure 9).In high-elevation catchments, drought-toflood transitions have a distinct seasonality because they are driven by snowmelt.The relationship between snowmelt and transitions is indicated by the match of the timing of transitions with the timing of snowmelt which is particularly expressed in catchments with elevations above 2,000 m which comprise 16% of the catchments studied here (Figure 4b).Snowmelt-driven transitions occur at high-elevations, that is, catchments with a high snow-fraction, in summer, while they occur in winter at lower-elevations where snowmelt occurs earlier in the year.The importance of snowmelt as a main driver of drought-to-flood transitions is further highlighted by the observation that drought-to-flood transitions occur almost exclusively during the high-flow season (Figure 5a), which coincides with the snowmelt season in high-elevation catchments (Stewart et al., 2004) (Figure 4b).Snowmelt is the main cause of flooding in these regions, specifically in the Rocky Mountains (Berghuijs et al., 2016;Villarini, 2016).Drought-to-flood transitions could here be favored by freezing temperatures, which prevent snowmelt and lead to anomalously low streamflow, that is, streamflow droughts (Brunner et al., 2023;Van Loon et al., 2015;Van Loon & Van Lanen, 2012).The remaining snowpack can then later contribute to streamflow floods and drought-to-flood transitions.
In low-elevation catchments, the timing of drought-to-flood transitions is much more variable and we could not identify a clear and unique hydro-meteorological driver of these events as in the case of high-elevation  catchments.However, also in low-elevation catchments, drought-to-flood transitions are more likely to occur in the high-than in the low-flow season (Figure 5b).Identifying the atmospheric drivers of drought-to-flood transitions remains difficult due to the complexity of individual drought-to-flood transitions and the large variability in their characteristics (He & Sheffield, 2020).
Droughts and floods are not just influenced by atmospheric and terrestrial processes (e.g., Brunner et al., 2022;Stein et al., 2020;Tarasova et al., 2019), but also human actions (e.g., Di Baldassarre et al., 2017;Van Loon et al., 2016).These different factors also influence drought-to-flood transitions, as illustrated by our finding that regulated catchments show fewer transitions than nearly natural catchments (Figures 3c and 3d), in particular in summer (Figures 8 and 9).Human activities can also alter the timing of drought-to-flood transition events as indicated by the larger variability in the timing of rapid transitions in regulated compared to natural catchments (Figures 4 and 7).In natural catchments, transitions occur later in the year with increasing elevation but only at elevations above 1,500 m (Figure 4b).In regulated catchments, this transition-elevation pattern is much more irregular (Figure 7b).
These findings highlight that flow regulation impacts the occurrence of drought-to-flood transitions which implies that (a) it is important to consider regulation when analyzing drought-to-flood transitions and (b) transition events can be managed and their negative impacts potentially mitigated.A decrease in the frequency of transitions, or of flood volume in general, can either be achieved by short-term event management or be a side-product of long-term management which we want to illustrate on the example of hydropower reservoirs at highelevations, which rely on snowmelt inputs: Reservoirs used for hydropower generation are typically emptied during the winter and get refilled during the snowmelt season in late spring to early summer.In catchments regulated for hydropower production, snowmelt is an essential component for refilling reservoirs and is accounted for in long-term reservoir management planning, implying that reservoir levels are usually low before the snowmelt season.A flood event, for example, caused by snowmelt, which occurs upstream of the reservoir, would hence rise the water level in a reservoir, but the flood peak would not be transferred downstream because it can be absorbed by free reservoir capacity.A drought event occurring downstream of a reservoir before or during the snowmelt season indicates that reservoir outflow is reduced to store water in the reservoir, suggesting that  reservoir levels are lower than usual and the reservoir has some free capacity.Because such reservoir management schemes reduce the occurrence of floods downstream of reservoirs, drought-to-flood transitions are occurring less frequently during the snowmelt season in regulated high-elevation catchments.In other seasons, reservoir levels might be higher and reservoirs might have a smaller capacity for reducing the discharge during flood events.The effect of reservoirs in reducing the number of drought-to-flood transitions is hence much weaker in low-elevation catchments where the discharge is less dependent on snowmelt (Figure 9b).

Challenges in Analyzing Drought-To-Flood Transitions
Drought-to-flood transitions have recently received more attention in the scientific community, but as of today, there are no recommendations available on how to define or how to characterize them.In this study, we have introduced the terms "rapid" and "seasonal" transition as two different types of transitions.We have chosen a 14day threshold to represent rapid, and presumably impactful events, and a 90-day threshold to present less rapid but more numerous events.While the choice of the exact threshold for rapid transitions is up for debate, it does according to our sensitivity analysis, in which we tested thresholds of 7, 14, and 21 days, not substantially impact our results.The number of identified transitions increases almost linearly with the number of days between a drought and a flood (Figure S7 in Supporting Information S1) and characteristics such as the seasonality of rapid events are not substantially affected by choosing potential alternative thresholds of 7 or 21 instead of 14 days (Figure S2 in Supporting Information S1).Using our chosen threshold of 14-day, rapid transitions occur less often than every 25 yrs in more than half of the catchments and occur only every 7 yrs in the catchments most affected by rapid transitions (Figure 4b).
The sensitivity analysis highlights that rapid drought-to-flood transitions are rare phenomena.While additionally working with seasonal definitions helps to slightly alleviate this problem, the number of transition events available for analysis remains limited.In particular at high-elevations, where gauging stations are rare, the sample size is limited with, for example, only 74 observed rapid drought-to-flood transitions above 2,000 m.This hampers robust statistical analyses about their occurrence and drivers.For such analyses, longer observational records or large-ensemble simulations, such as ensembles generated with stochastic models (Vogel, 2017), ensembles derived using reforecast pooling (Brunner & Slater, 2022;Kelder et al., 2022;Thompson et al., 2017), or Single Model Initial-condition Large Ensembles (SMILEs) (Brunner et al., 2019;Deser et al., 2020;Maher et al., 2021), would be highly beneficial.In such a large-ensemble context, the concepts of "seasonal" and "rapid" transitions introduced here could be used to further improve our understanding of spatial variations in transition occurrences and their hydroclimatic drivers.Furthermore, a larger sample size would enable analyses on a continuous scale beyond the two types of transitions.
The number of transition events in general depends on the definition of droughts and floods as illustrated by our sensitivity analysis (see Figure S1 and Tables S3 and S4 in Supporting Information S1).For this analysis, we calculated the number of droughts, floods and drought-to-flood transitions for different drought and flood thresholds.We used the 15th, 20th and 25th percentile as variable drought thresholds and the 25th, 50th and 75th percentile of the yearly discharge maxima as fixed flood thresholds.The sensitivity analysis shows that the number of droughts and floods increases with a higher threshold for droughts and lower threshold for floods, respectively (Figure S1 and Tables S1 and S2 in Supporting Information S1).The number of drought-to-flood transitions decreases both with an increase in the flood threshold and a decrease in the drought threshold.That is, the stricter the thresholds, the fewer transition events there are available for the analysis.The identification of droughts does not only depend on the chosen percentile threshold but also on the chosen minimum event duration.
The number of identified droughts, and hence also drought-to-flood transition events, depends on this minimum duration, with a higher duration leading to fewer, but potentially more impactful events.
The analysis of the effect of reservoir regulation highly benefits from a subdivision into regulated and nearly natural catchments.In this study, we rely on the natural-regulated classification approach by Ryberg et al. ( 2020) who consider the storage volume with respect to the discharge as an indicator of regulation.While this approach considers reservoir volume, our study does not consider reservoir purposes or the distance of gauging stations to reservoirs.This means that we are not able to look at variations in reservoir effects on drought-to-flood transitions for different reservoir purposes and that the reservoir regulation signal can not be properly isolated because it may be blurred by other processes taking place between the reservoir and the gauging station.Furthermore, natural and regulated catchments are not spread homogeneously across the CONUS and its different hydroclimatic zones. Water Hence, it is possible, that climatic influences also imprint on our observed effect of reservoir regulation on the occurrence of drought-to-flood transitions.

Future Drought-To-Flood Transitions
The transition behavior of droughts-to-floods may change in a future climate due to projected future changes in streamflow regimes and droughts and floods.How climate change will influence drought-to-flood transitions is yet unclear because of counteracting influences of changes in the snowmelt regime and an increase in the number of rain-on-snow events.Changes in the timing of the snowmelt season have been documented and even earlier snowmelt is expected in the future with less water stored as snow (Cho et al., 2021;Hale et al., 2023).Considering the higher number of drought-to-flood transitions during summer in higher-elevation and snow-dominated than in lower-elevation and rainfall-driven catchments (Figures 8 and 9), we expect an overall decrease in snow-related drought-to-flood transitions in a warming world.In particular, we expect less summer transitions in highelevation catchments (above 2,000 m) where an earlier snow-melt season likely leads to an earlier occurrence of drought-to-flood transitions and hence more winter transitions.Additionally, an increase in winter temperatures, and associated shifts from snowfall to rainfall, leads to an increase in rain-induced flood risk (Brunner, Melsen, et al., 2020;Knowles et al., 2006;Musselman et al., 2017).The number of rain-on-snow events could increase at high-elevation catchments where a seasonal snow-cover persists, but decrease in low-elevation catchments where the amount of snow decreases (Musselman et al., 2018).Overall, climate change may lead to shifts in the drivers of drought-to-flood transitions in catchments at high-elevation, primarily due to changing contributions of snowmelt as a cause for floods (Chegwidden et al., 2020).Human interference on natural flow could also strongly influence these change pattern as indicated by the effect of reservoir regulation on transitions (Figures 8b and 9b) and previous studies highlighting the large effects of regulation on river-regimes which can surpass the effects of climate change (Arheimer et al., 2017).
From a meteorological perspective, Rashid and Wahl (2022) have identified North America, including the US, as one of the hotspots for dry-to-wet transitions.They have found a historic increase in dry and wet extreme events and a negative trend in the recovery time between these extremes.They point out that the risks of harmful dry-towet transitions will increase if these trends continue.In Southern North America, these trends can be attributed to increased variability of precipitation and potential evapotranspiration and an increase of such transition events has been projected for the future (Chen & Wang, 2022).These previous studies have worked with monthly and climatic indices.That is, they focused on changes in rather long water availability anomalies which are not necessarily linked to flood events.How exactly changes in climate and reservoir regulation are going to individually and jointly affect future hydrological drought-to-flood transitions hence still remains to be assessed in targeted modeling experiments.
For future analyses of transitions from streamflow droughts to floods, we suggest to characterise drought-to-flood transitions also by the streamflow development between the end of a drought and the onset of a flood, because transition events with the same transition length can have very different transition phases.For example, a transition event could have a gradual increase in streamflow levels between a drought and a subsequent flood.In contrast, streamflow can leave drought conditions but remain on a below average level before suddenly increasing to flood levels leading to a transition event with a length comparable to the previous example.In the scheme used in this study, these two events are treated as similar events, but they can have very different impacts due to the associated change of water management practices and the surprise effect of a sudden flood in contrast to slowly rising water levels.

Conclusions
In this study, we have analyzed the occurrence of transitions from hydrological droughts to floods in natural and regulated catchments in the Contiguous United States.We show that (a) the length of drought-to-flood transitions varies strongly in space, with short and long transitions occurring in all hydroclimates, (b) catchments at high elevation (>2,000 m) and with high fractions of snow (>40%) experience rapid transitions mainly during the snowmelt-season, (c) snowmelt floods are used to fill reservoirs which reduces the number of drought-to-flood transitions in snowmelt-driven catchments and (d) low-elevation catchments show a much more variable timing of rapid drought-to-flood transitions than high-elevation catchments.The latter finding suggests that multiple and different processes can lead to drought-to-flood transitions in catchments where snow plays a less important role Water Resources Research 10.1029/2023WR036504 in the hydrological cycle.Further analyses of transitions from streamflow droughts to floods would benefit from a better knowledge about which kind of drought-to-flood transitions are most impactful and challenging for water management, for example, transitions from low-flow to high-flow anomalies or transitions from low-to highwater levels in absolute terms.Future research could then be tailored toward the interest and needs of water managers and future changes in transitions be investigated in order to identify suitable water management and adaptation measures.Potential measures include reservoir regulation schemes taking into account drought and flood protection measures alike and strategies for the improvement of water quality after high material inputs due to floods following on dry periods.

Figure 2 .
Figure 2. Drought-to-flood transition time distributions for four example catchments (a), within CONUS highlighted on the map (b), and characterized by different streamflow regimes (i.e., smoothed hydrographs) (c).The transition times are ranked per catchment to indicate the different numbers of transitions across stations.The introduced definitions of "seasonal transitions" (≤90 days) and "rapid transitions" (≤14 days) are indicated by gray boxes, with seasonal transitions also comprising rapid transitions.

Figure 3 .
Figure 3. Number of rapid drought-to-flood transitions between 1970 and 2022.The maps show the spatial distribution for (a) natural and (b) regulated catchments while (c) shows the number of catchments with a certain number of transitions for natural and (d) for regulated catchments.

Figure 4 .
Figure 4. Timing of rapid transitions in natural catchments: (a) Map showing the mean seasonality of rapid drought-to-flood transitions.The color of the points indicates the average transition timing at that catchment and the point size indicates the standard deviation of the timing within a catchment.Please note that a standard deviation of zero can mean that only one rapid transition has occurred in a catchment.(b) Mean and variability of timing of rapid transitions with respect to median catchment elevation.Catchments are binned according to elevation with a bin-size of 250 m.The x-axis labels indicate the upper boundary of each bin.The numbers at the top of the figure indicate the number of events in each bin.

Figure 5 .
Figure 5. Number of rapid transitions during the (a) high-and (b) low-flow seasons (see Figure 1d) in natural catchments.

Figure 6 .
Figure 6.Number of seasonal transitions (<90 days) between 1970 and 2022.The maps show the number of seasonal transitions for (a) natural and (b) regulated catchments, while the corresponding histograms show the number of seasonal transitions per elevation bin for (c) natural and (d) regulated catchments.

Figure 7 .
Figure 7. Timing of rapid transitions in regulated catchments: (a) Map showing the mean seasonality of rapid drought-to-flood transitions.The color of the points indicates the average transition timing at that catchment and the point size indicates the standard deviation of the timing within a catchment.Please note that a standard deviation of zero can mean that only one rapid transition has occurred in a catchment.(b) Mean and variability of timing of rapid transitions with respect to median catchment elevation.Catchments are binned according to elevation with a bin-size of 250 m.The x-axis labels indicate the upper boundary of each bin.The numbers at the top of the figure indicate the number of events in each elevation bin.

Figure 8 .
Figure 8. Number of rapid drought-to-flood transitions in summer (top, orange) and winter (bottom, blue) with respect to the amount of precipitation falling as snow in (a) natural and (b) regulated catchments.Each cross represents a catchment.

Figure 9 .
Figure 9. Number of seasonal drought-to-flood transitions in summer (top, orange) and winter (bottom, blue) with respect to the amount of precipitation falling as snow in (a) natural and (b) regulated catchments.Each cross represents a catchment.The dashed green line connects the median values of the snow-fraction bins.