1. Introduction
Extreme streamflow estimation is essential for the design of hydraulic infrastructures and flood warning systems, as well as for proper risk assessment in flood zones. Several studies based on the historical record or on climate projections have shown how climate change can impact the hydrological cycle (Chen et al. 2012; Croitoru and Minea 2015; Donat et al. 2013; Markonis et al. 2019). Rainfall and extreme rainfall changes are especially critical to the management of water resources (Arnbjerg-Nielsen 2006; Milly et al. 2008; Samuels et al. 2009; Wang et al. 2013). These changes can substantially affect streamflow regimes in terms of mean flow, seasonality, as well as intensity and frequency of extreme runoff (Bormann 2010; Gobiet et al. 2014; Guerreiro et al. 2018; Moustakis et al. 2021; Strasser et al. 2018). Changes to projected future rainfall display complex spatial patterns (Rajulapati et al. 2020) that are globally dependent on latitude. At higher latitudes, a transition from snowfall toward rainfall is expected in the winter over many regions (e.g., Dougherty et al. 2020; Riboust and Brissette 2015; Rasmussen et al. 2014; Minville et al. 2008), although many other factors have an influence such as local topography and local climate zone. There is also a strong scientific consensus that deep convection will become more frequent and intense. in a warmer climate, leading to more extreme precipitation, and particularly so for short-term duration and longer return periods (Cannon and Innocenti 2019; Cuo et al. 2011; Fildier et al. 2017; Martel et al. 2021; O’Brien et al. 2016; Pendergrass 2020; Pendergrass et al. 2016; Semie and Bony 2020). This shift toward more extreme and shorter-duration rainfall will have a significant impact on summer–fall floods in smaller catchments (Bertola et al. 2020; Martel et al. 2021; Prein et al. 2017).
The Intergovernmental Panel on Climate Change (IPCC) Fifth Assessment Report (IPCC 2007) stated that the frequency of extreme precipitation is expected to grow over midlatitude regions by the end of the century. Increasing temperature impacts moist convection by increasing evapotranspiration and convective available potential energy (CAPE), thereby strengthening convective processes (Brooks 2013; Capek 2021; Diffenbaugh et al. 2013; Tippett et al. 2015; Trapp and Hoogewind 2016). This is consistent with the Clausius–Clapeyron (CC) equation, which states that atmospheric water vapor pressure increases at a rate of 7% °C−1 of warming. However, the relationship between warming and precipitation, and particularly with extreme precipitation, is complex and depends on many factors such as type, return period, and duration of rainfall.
Berg et al. (2013) divided observational rainfall into two groups: convective and large-scale stratiform rainfall. They postulated that with increasing temperature, the probability of occurrence of convective extreme rainfall rises, leading to more intense precipitation, with shorter duration (Berg et al. 2013; Westra et al. 2014). The rate of increasing extreme precipitation is not uniform for stratiform and convective precipitation. Anticipated increases in extreme stratiform precipitation align with the Clausius–Clapeyron rate, while extreme convective precipitation could potentially surpass this rate, achieving super-Clausius–Clapeyron levels. As a result, extreme convective precipitation may play a more significant role in shaping future precipitation patterns compared to its influence in today’s climate (Guerreiro et al. 2018). Prein et al. (2017) examined the impact of climate change on a mesoscale convective system using a convection-permitting regional climate model. They found that subdaily convective extreme rainfall is expected to increase over most of North America, and particularly so in the northeast United States and Canada. Similar conclusions have been reached in a number of observation-based studies (e.g., Diffenbaugh et al. 2013; Feng et al. 2016; Gensini and Mote 2015). There is increasing scientific agreement that extreme rainfall scaling is related to both duration and frequency (Cannon and Innocenti 2019; Martel et al. 2021), with short-duration, low-frequency subdaily rainfall most likely to follow super-CC scaling. This finding is supported by the observational record and regional climate model projections (Lenderink and Van Meijgaard 2008, 2010; Panthou et al. 2014; Westra et al. 2014). The amplification of short-duration, long-return-period rainfall is expected to be one of the most important impacts of climate change (Forestieri et al. 2018; Ganguli and Coulibaly 2019; Martel et al. 2021, 2020).
To properly evaluate the impact of changes in rainfall magnitude on future floods, rainfall duration and catchment size need to be jointly considered, since rainfall durations commensurate with catchment reactivity (time of concentration) are most likely to maximize streamflow extremes. Smaller catchments are therefore most likely to face an increased flooding risk from convective storm cells, whereas catchments whose size is larger than that of storm cells are potentially less affected due to flood wave attenuation across the catchment. The reduction of spatially averaged rainfall intensity over progressively larger areas (e.g., catchments) has been the subject of many studies, and most notably through the use of areal reduction factors (ARFs) (e.g., Ball et al. 2016; Wright et al. 2014). ARFs are defined as the ratio of rainfall depth across a given area (for a given duration and return period) compared to point rainfall, as typically measured by weather stations. ARFs start from 1 (local scale/very small catchments) and approach 0 as the considered surface areas increase and become progressively larger than storm cells. ARF values have been found to decline faster for short-duration events due to their highly convective nature and small spatial extent (Mineo et al. 2018; Ramos et al. 2005). Catchment size is therefore a critical aspect to consider in flooding related to extreme rainfall (Fowler et al. 2021; Westra et al. 2014). Failing to consider shorter-duration rainfall amplification may lead to potentially significant underestimations of future flood probability in smaller catchments (Cheng and AghaKouchak 2014).
Currently, a significantly large body of literature is available on the impacts of climate change on water resources. However, most of the work was conducted at the daily time scale, since, until recently, most climate model projections outputs were available at this time step. Therefore, few studies have explored the impacts on water resources at the subdaily time step. Since the largest increases in extreme rainfall are expected to be for subdaily durations, we can thus predict a disproportionate impact on small catchments, whose hydrological response is rapid and commensurable with that of convective precipitation systems. These findings point to major climate change impacts on small rural and urbanized watersheds, including in cities and urban areas that are already very vulnerable to rainfall extremes. For most of these catchments, a daily simulation time step may be too coarse. Accordingly, the main objective of this work is to look at the impact of the amplification of extreme precipitation on runoff as a function of rainfall duration and catchment size. This main objective is split into three specific objectives, namely, 1) quantify future changes in extreme rainfall at the subdaily time scale; 2) simulate the impact of changes in extreme rainfall on streamflow; and 3) explore the relationship between catchment size, rainfall duration, and future streamflow increases.
2. Methodology
The methodological framework of this study is presented in Fig. 1. A database of 133 catchments with surface areas ranging between 66.5 and 9886 km2 is set up within the North American computational domain of the ClimEx experiment (Leduc et al. 2019), which is described later herein. An hourly database of precipitation, temperature, and streamflow data is used in each catchment to calibrate a hydrological model over a common 1980–2003 reference period. Following a bias correction step, the hourly precipitation and temperature outputs from ClimEx are used to generate climate scenarios over the 1980–2003 reference and 2075–99 future periods. The scenarios both contain 1200 years (50 members times 24 years per member) representing climate conditions over the 24-yr reference and future periods. Extreme rainfall and streamflow values are then analyzed as a function of the return period, rainfall duration and catchment size. Each methodological step is detailed below.
Methodological framework of this study.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Methodological framework of this study.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Methodological framework of this study.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
a. Catchment database
A total of 133 catchments dispersed across the northeastern United States were selected from the Model Parameter Estimation Experiment (MOPEX) database (Duan et al. 2006). The MOPEX database was chosen since it contains a quality-controlled database of hourly precipitation averaged at the catchment scale. MOPEX precipitation is derived from the combination of daily and hourly weather station datasets from the National Climate Data Center and the Natural Resources Conservation Service. Daily precipitation data were disaggregated from the nearest hourly gauge, and a minimum number of gauges per unit area was required for catchments to be included in the database. All chosen catchments are located within the ClimEx northeastern North America computational domain. A maximum of 5% of missing data for precipitation, temperature, and streamflow data was used as a threshold for inclusion in the database. Figure 2 presents the location of these catchments. The catchments cover four distinct climate zones of the Köppen climate classification. The catchment areas vary between 66.5 and 9886 km2. To investigate the impact of the catchment size, all 133 catchments are separated into three different size groups: smaller than 500 km2, between 500 and 1000 km2, and larger than 1000 km2. The three groups respectively contain 12, 25, and 96 catchments. The median values for total annual precipitation and mean annual temperature were respectively 1247, 1072, and 1049 mm and 11.9°, 10.9°, and 11.5°C across all three size classes. Using three groups containing an equal number of catchments was also considered, but by doing so, the mean catchment size of each group was deemed too large to appropriately separate catchments with a clear subdaily response (<500 km2) from those with a daily response (>1000 km2). The 500- and 1000-km2 values are somewhat arbitrary, considering that catchment response time depends on many factors in addition to size (e.g., average slope, land use), but they were found to be adequate for the study area in the absence of large mountain ranges. Many catchments are located in the Appalachian range, but none occupy areas with long continuous slopes that would significantly impact response time.
Location of the centroid coordinates of all selected catchments. The color and size of the circles are functions of the catchment surface area.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Location of the centroid coordinates of all selected catchments. The color and size of the circles are functions of the catchment surface area.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Location of the centroid coordinates of all selected catchments. The color and size of the circles are functions of the catchment surface area.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
b. Observed hydrometeorological data
In this study, hourly precipitation, temperature and streamflow datasets over a common 1980–2003 period are used to define the reference period. This 24-yr reference period is shorter than the typically recommended 30-yr duration typically used to define the reference climate. This choice was dictated by hourly precipitation data availability. It is worth noting that several climate change impact studies have moved toward shorter durations (typically 20 years) for various reasons (e.g., Martel et al. 2020). Hourly precipitation and discharge data for all catchments are taken from the MOPEX database (Duan et al. 2006). In the MOPEX dataset, precipitation is averaged at the catchment scale. Since hourly temperature data is not available in the MOPEX database, hourly temperature from the ECMWF Reanalysis v5 (ERA5) dataset was used as a substitute. Tarek et al. (2020) showed that ERA5 temperature is just as accurate as the weather station temperature for hydrological modeling. This is not entirely surprising since ERA5 incorporates measured surface temperature in its assimilation scheme. The spatial and temporal resolutions of ERA5 are respectively 30 km and 1 h. Hourly temperature was averaged at the catchment scale using all grid points inside the catchment boundaries. For the smaller catchments, if no grid point was found within their boundaries, the closest ERA5 grid point to the catchment centroid coordinates was used. The reference period dataset is used for the hydrological model calibration as well as for bias-correcting ClimEx hourly precipitation and temperature outputs.
c. Hydrological model and calibration
The rainfall–runoff model selected for this study is GR4H (Génie Rural à 4 paramétres Horaires). It was first implemented by Mathevet (2005) specifically to be run at the hourly time step and is based on GR4J (Génie Rural à 4 paramétres Journalier), a widely used daily lumped continuous rainfall–runoff model introduced by Perrin et al. (2003). GR4J has been used in multiple studies (Harlan et al. 2010; Kunnath-Poovakka and Eldho 2019; Traore et al. 2014), including in many climate change impact studies (Brigode et al. 2013; Li et al. 2013; Tian et al. 2013). GR4H has been optimized to the hourly time step (Bennett et al. 2014), but its structure is very similar to that of GR4J. GR4H is therefore also a lumped conceptual continuous rainfall runoff model. Its structure includes two reservoirs (production and storage) and two unit hydrographs which are used for flow routing. It has four parameters that need to be adjusted for optimal model performance: the maximum capacity of the production and routing reservoirs (X1 and X3), a groundwater exchange coefficient for the production store (X2) and the time base of the unit hydrograph (X4) for the flow routing reservoir. Snowpack accumulation and depletion are simulated by coupling GR4H with the two-parameter CemaNeige degree-day snow model (Valéry 2010). The CemaNeige-GR4H model requires potential evapotranspiration and precipitation at the hourly time step as inputs. The PET formulation of Oudin (Oudin et al. 2005) was chosen for this work. It is a radiation-based formula which uses hourly (or daily) temperature as its sole input. This formulation was specifically developed to work with the GR4J model and has been shown to provide better simulation results as compared to other possible alternatives (Oudin et al. 2005).
d. Climate model data
This work examines the impact of climate change on extreme precipitation in small catchments with a need for an hourly modeling time step. Accordingly, the ClimEx Single Model Initial-Condition Large Ensemble (SMILE) was chosen (Leduc et al. 2019). ClimEx is a high spatial (12 km) and temporal (1 h for precipitation and 3 h for temperature) resolution regional climate model. It was generated by dynamically downscaling the CanESM2 SMILE (CanESM2-LE) over two computational domains covering Europe and northeastern North America. CanESM2 is the second version of the Canadian Centre for Climate Modeling and Analysis (CCCma) Earth system model, with a spatial resolution of 2.8° (Arora et al. 2011). CanESM2-LE is a 50-member large ensemble derived from random atmospheric perturbation applied on historical data (Sigmond and Fyfe 2016). As described in Leduc et al. (2019), the CanESM2-LE was generated using a two-step perturbation process. Starting with a 1000-yr equilibrium simulation (CMIP5 preindustrial Control run), five sets of random atmospheric perturbations were applied in 1850, and the five runs evolved independently for 100 years, resulting in five different ocean states in 1950. From 1950 onward, 10 sets of random atmospheric perturbations were added to each of the original five simulations, resulting in a total of 50 simulations. CanESM2-LE and ClimEx both provide 50 members over the 1950–2100 period under historical forcing, and following the RCP 8.5 scenario from 2005. Large ensembles were developed to better understand the impact of internal variability (Deser et al. 2012; Frankcombe et al. 2015; McKinnon and Deser 2018; Thompson et al. 2015), but they can also be used to robustly sample very rare events since they provide multiple realizations of the climate under identical forcing (Maher et al. 2021). Thus, for the 24-yr reference period (1980–2003), ClimEx provides 1200 equivalent years (50 × 24). Extreme precipitation with long return periods (e.g., 100 years) can therefore be empirically determined without necessarily having to extrapolate existing data using a fitted generalized extreme value distribution, as is generally done with samples of limited size. This advantage of large ensembles has been exploited in many recent studies on extremes (e.g., Ehmele et al. 2020; Martel et al. 2020; Zhao et al. 2020). For each catchment in the present study, ClimEx hourly precipitation and 3-h temperature data were extracted for all grid points within a catchment boundary and averaged over both the reference (1980–2003) and future (2075–99) periods. The 3-h averaged temperature was subsequently interpolated to the hourly time step using a piecewise cubic Hermite interpolating polynomial (Barker and McDougall 2020; Epstein 1976). Only one catchment (the smallest at 67 km2) did not contain a CLIMEX grid point within its boundary. The CLIMEX grid point closest to the catchment centroid was therefore used in this case.
e. Bias correction
Many studies have discussed the necessity of bias-correcting climate data to provide accurate streamflow representation when using a hydrological model (Crochemore et al. 2016; Hagemann et al. 2011; Tan et al. 2020; Teutschbein and Seibert 2012; Tiwari et al. 2022). In this study, the hourly ClimEx temperature and precipitation data were bias-corrected by using the N-dimension multivariate bias correction (MBC-n) method of Cannon (2018). MBC-n is a multivariate generalization of quantile mapping that considers the dependency among different variables (Cannon 2018). By applying MBC-n, all statistical characteristics of an observed continuous multivariate distribution are transferred to the corresponding multivariate distributions of simulated variables. MBC-n is arguably the most advanced quantile mapping algorithm available. Unlike many other multivariate methods, it is not limited to correcting a given measure of joint dependence (e.g., Pearson or Spearman rank correlation) (Cannon 2018).
MBC-n also possesses the highly desirable attribute of preserving the climate change signal from the parent climate model across all quantiles, which is a significant limitation in most other quantile mapping methods (Maraun et al. 2017). In this study, the bias correction factors were computed after pooling all 50 members of the ClimEx dataset (temperature and precipitation) together to preserve the underlying internal variability (Chen et al. 2019). The correction factors were computed on a monthly basis to account for the seasonality, and on an hourly basis to correct for biases in the model reproduction of the diurnal cycles (Faghih et al. 2022). More details can be found in the above references.
f. Streamflow scenarios and analysis of extremes
The bias-corrected temperature and precipitation data were used as inputs to the calibrated hydrological model. For both the reference and future periods, the hydrological model was run at the hourly step for the 50 members of the ClimEx ensemble, for a total of 1200 equivalent years for each period. For each of those years, the maximum accumulated rainfall over durations of 1, 2, 6, 12, 24, and 72 h was selected using a 1-h moving window. Since this paper focuses on rainfall-generated floods, accumulated rainfall was only computed over months during which the snow cover was deemed negligible. As this study covers a variety of catchments and climate zones, the selected period was catchment-specific and restricted to the months in which the mean average temperature was above 0°C. Other higher temperature thresholds were tested with no significant impact on the results. Annual maximum computed streamflow discharge and rainfall were taken over the same selected period. For example, for the 6-h duration, precipitation was accumulated over all possible 6-h continuous intervals, and the maximum value for each year was selected. The process was similar for streamflow, although the average streamflow (and not the accumulated value) was computed for all possible 6-h continuous intervals. This procedure resulted in 1200 values (24 years times 50 members) for the maximum yearly accumulated rainfall and maximum yearly discharge (for all 6 durations).
Finally, rainfall and streamflow values corresponding to return periods of 2, 10, 20, 50, 100, and 300 years were computed for each catchment using the unbiased Cunnane ranking formula.
3. Results
Figure 3 presents the calibration results over the 24-yr reference period. The first year is used for hydrological model spinup and is not otherwise used in any of the methodological steps. The median NSE value is equal to 0.78 and the range for the 133 catchments goes from 0.61 to 0.87. There is little difference across the three class sizes, with medians of 0.73, 0.76, and 0.79, respectively, for the small, medium, and large size classes. These values show that the GR4H-CemaNeige hydrological model performs very well over the reference period and that it is able to adequately represent the main hydrological processes over the study area.
NSE calibration value for all study catchments.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
NSE calibration value for all study catchments.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
NSE calibration value for all study catchments.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Figure 4 displays the projected change in extreme rainfall between the reference and future periods. Changes are all presented in terms of relative percentage increases. A value of 25% represents a 25% increase for a given rainfall duration and return period. Figure 4 is divided into six subplots corresponding to the six rainfall durations under study (1, 2, 6, 12, 24, and 72 h). For each of these subplots, six series of three boxplots are presented. The six series correspond to the six return periods (2, 10, 20, 50, 100, and 300 years) considered, and the three boxplots correspond to the three catchment size classes [small (S), medium (M), and large (L)]. Each boxplot represents the distribution of rainfall increases among all catchments within each size class (12, 25, and 96 catchments, respectively). The boxplots show the median (red line), 25th, and 75th quantiles of the distribution (blue box), whereas the upper and lower whiskers present the minimum and maximum range of values. The red crosses are considered statistical outliers. To better outline the results, Table 1 presents the median values of Fig. 4.
Extreme rainfall increase (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme rainfall increase (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme rainfall increase (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Figure 4 and Table 1 display four important features:
-
First, generalized increases in extreme precipitation are observed across all durations and return periods, with a few rare exceptions (mostly outlier catchments in the longer rainfall durations).
-
Second, with the exception of the 2-yr return period, the projected increases in extreme rainfall are strongly related to catchment size, with the smaller catchments seeing a much larger change compared to the medium and large ones. For example, the 100-yr 1-h rainfall sees a projected median increase twice as large as that of the large size class.
-
Third, the projected increases in extreme rainfall are clearly related to rainfall duration, with the shorter duration witnessing the largest increases.
-
Fourth, the projected increases in extreme rainfall are clearly related to the return period of the rainfall duration, with the longer return period witnessing the largest increases.
To sum up, Fig. 4 shows that projected increases in extreme rainfall become greater for smaller rainfall durations, longer return periods, and smaller areas. This behavior observed in the ClimEx rainfall data is consistent with what has been observed in recent works with other climate models, as previously discussed in the literature review.
Figure 5 presents the same results in a different format. The three subplots in the figure present the results for the three catchment size classes. For each of these subplots, the median relative increase in precipitation (boxplot red line in Fig. 4) is color plotted as a function of rainfall duration and return period. Figure 5 clearly shows the amplification of rainfall in a warmer climate as a function of decreasing duration and increasing return period for all three size classes. The impact of the catchment size is made quite clear by comparing all three subplots.
Extreme rainfall median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme rainfall median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme rainfall median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Figure 6 presents the future return period of the 100-yr reference period rainfall. A value below 100 indicates an increasing frequency in the future. For example, a future return period of 20 years indicates that the 100-yr rainfall of the reference period will occur every 20 years (on average) for the future period, a fivefold increase in frequency. A value above 100 indicates that the same 100-yr rainfall will become less frequent in the future.
(right) Future return period (x axis) of the 100-yr reference period rainfall, as a function of catchment size class and rainfall duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
(right) Future return period (x axis) of the 100-yr reference period rainfall, as a function of catchment size class and rainfall duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
(right) Future return period (x axis) of the 100-yr reference period rainfall, as a function of catchment size class and rainfall duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Results strongly emphasize the strong increases in the frequency of the reference period 100-yr rainfall, particularly for durations shorter than 12 h. For such durations, the median 100-yr rainfall becomes at least 6 times more frequent over the future period for all catchment size classes. Increases are larger and boxplots are tighter for the small catchments, but the latter may be related to the different number of catchments in each of the three size classes. The frequency increases become progressively smaller for the longer rainfall durations and the difference between size classes becomes larger. The spread of the small catchment boxplots is considerably tighter, but, once again, this may be due to the smaller number of catchments. There is a relatively small number of catchments for which there is no increase in the future frequency of the 100-yr rainfall (future return period longer than 100 years). These are all large and, to a lesser extent, medium size catchments. This occurs only for rainfall durations equal to or longer than 24 h (with a single exception for the 12-h duration). This behavior is not random and is only seen in catchments located at the northern end of the study location.
The results presented above are consistent with the recent body of literature on subdaily precipitation in a changed climate. However, how these changes will impact streamflow is not clear, considering the potential impact of increased evapotranspiration due to warmer temperatures. The next figures (Figs. 7, 9, and 10) follow the layout of Figs. 4–6, but for streamflow. To allow for a direct comparison, the corresponding figures are all plotted using the same scales.
Extreme streamflow increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and streamflow return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme streamflow increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and streamflow return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme streamflow increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (six main subplots) and streamflow return period (x axis). The series of three boxplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Figure 7 displays the projected changes in extreme streamflow between the reference and future periods. Changes are all presented in terms of relative percentage increases, just as was the case for Fig. 4. Figure 7 follows the same patterns observed for precipitation in Fig. 4, and the four main observed patterns also apply here. These patterns show larger relative increases for shorter durations, lower frequencies (longer return periods) and smaller catchments. However, the relative increases appear to be larger for extreme streamflow as compared to extreme rainfall. To better explore this apparent increase, Fig. 8 plots the ratio of the relative streamflow increase over that of precipitation. A value larger than 1 indicates that the relative streamflow increase is larger than that of precipitation, while a value smaller than 1 indicates the opposite.
Ratio of the relative increase of streamflow (Fig. 7) over that of precipitation (Fig. 4).
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Ratio of the relative increase of streamflow (Fig. 7) over that of precipitation (Fig. 4).
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Ratio of the relative increase of streamflow (Fig. 7) over that of precipitation (Fig. 4).
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Three main observations can be made from Fig. 8:
-
First, all boxplots are more or less lined up on horizontal lines, therefore showing little dependence on the return period.
-
Second, the median ratios are systematically larger than 1, indicating that relative increases in streamflow tend to be larger than those of extreme rainfall.
-
Third, the ratios get larger for longer duration rainfall. While the median values are only slightly above 1 for the 1- and 2-h durations, they rise progressively all the way to 72 h when the 25th quantile is above 1, in all cases. There is, however, a lot of variability, with a significant number of catchments having a ratio below 1.
Figure 9 presents the same results as Fig. 7, but in a different format. Just as was the case for Fig. 5 (sister figure for rainfall) the three subplots of Fig. 9 present the results for the three catchment size classes. In each subplot, the median relative streamflow increase (boxplot red line in Fig. 7) is color plotted as a function of rainfall duration and return period. The lighter the color, the larger the median relative increase. Figure 9 clearly shows the amplification of streamflow in a warmer climate as a function of decreasing duration and increasing return period for all three size classes. The changes are also strongly dependent on catchment size, with the smaller size class showing the largest increases. When compared to Fig. 5 (rainfall), the colors are noticeably lighter, indicating the comparatively larger relative increases of streamflow.
Extreme streamflow median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme streamflow median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Extreme streamflow median increases (%) between the reference (1980–2003) and future (2075–99) periods as a function of rainfall duration (x axis) and return period (y axis). The three subplots respectively represent the small (S), medium (M), and large (L) catchment size classes.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
Figure 10 presents the future return period of the 100-yr reference period streamflow. Just as was the case for Fig. 6 (sister figure for rainfall), a value below 100 indicates an increasing frequency in the future. The results are similar to those for rainfall (Fig. 6), and particularly so for the 72-h duration. However, for the 1-h (and 24-h to a lesser extent) durations, the colors are darker, indicating a smaller increase in frequency (return period) as compared to rainfall. Therefore, despite the larger increase in relative streamflow (compared to rainfall), it nonetheless translates into a smaller decrease in future return period. These increases are significant, however, with future median return periods ranging between 20 and 45 years (2.2–5 times increase in frequency). Just as was the case for precipitation, the largest changes are observed in the Great Lakes region.
(right) Future return period (x axis) of the 100-yr reference period streamflow, as a function of catchment size class and streamflow duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
(right) Future return period (x axis) of the 100-yr reference period streamflow, as a function of catchment size class and streamflow duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
(right) Future return period (x axis) of the 100-yr reference period streamflow, as a function of catchment size class and streamflow duration (y axis). (left) Geographical distribution of future return period for 1, 24, and 72 h.
Citation: Journal of Hydrometeorology 24, 8; 10.1175/JHM-D-22-0224.1
4. Discussion
This work investigated the amplification of extreme rainfall and extreme streamflows in a warmer climate, and how catchment size impacts this amplification. It made use of the high temporal and spatial resolutions of the ClimEx SMILE coupled with a hydrological model to simulate future extreme flows. Results demonstrate quite convincingly the amplification of future rainfall as a function of both duration and return period, with low frequency and shorter-duration extreme rainfall seeing the largest relative increases in a warmer climate. The increases are especially notable over the smaller catchments for all rainfall durations, but especially so for the shorter durations. These results are consistent with recent climate modeling studies (Cannon and Innocenti 2019; Hosseinzadehtalaei et al. 2020; Westra et al. 2014) and observations over the historical data record (Lenderink and Van Meijgaard 2008, 2010; Panthou et al. 2014). In addition to duration and return period, these results suggest that future extreme rainfall increases are also constrained with respect to storm size, with smaller storm cells becoming more intense as compared to larger ones.
A similar pattern of future change is observed with modeled streamflows, but with the notable difference that the relative amplification of extreme flows tends to be larger. This is in line with the finding of Dougherty and Rasmussen (2020) who observed larger increases in future runoff than rainfall in flood-producing storms. However, we find that the amplification factors of Fig. 8 (ratio of the relative increase of streamflow over that of precipitation) depend mostly on extreme rainfall duration, with durations longer than 8 h showing systematically larger values (compared to shorter durations), and above 1 for almost all catchments. The amplification factors do not appear to be closely related to catchment size. The complex relationship between rainfall characteristics and runoff has been explored in a few studies (e.g., Reaney et al. 2007; Wainwright and Parsons 2002), which showed that the temporal variability of precipitation interacts with the catchment scale to reduce the apparent runoff coefficients as catchments get larger. We do not clearly observe this in our study. This could be due to the fact that a series of maximum annual rainfall and streamflows were sampled independently, meaning that for any given year, the maximum annual streamflow may be independent from its precipitation counterpart. Antecedent soil moisture conditions are important for the generation of extreme streamflows and would typically explain this. While amplification factors tend to be larger than 1 for relative changes, our results show that changes expressed in terms of changes in return period tend to be larger for rainfall than for streamflow, and particularly so for durations below 12 h. Results show that the current 100-yr rainfall may become 3–20 times more frequent in the future (depending on catchment), versus 2–6 times more frequent for streamflows. This is likely a consequence of streamflow distributions having a heavier tail (Basso et al. 2015; Bernardara et al. 2008) than for rainfall.
Overall, the results presented in this work demonstrate the importance of catchment size on future flooding changes, with smaller size catchments being significantly more impacted by projected increases in extreme rainfall in a warmer climate. The projected changes in extreme floods indicate that storm drainage infrastructures are particularly at risk, especially those in urban areas, which typically serve small catchments, and those in small and steep rural catchments, which are already susceptible to flash floods (Zhang et al. 2019, 2021).
There are several limitations associated with this study, the most important being the use of a single climate model under a single greenhouse gas emission scenario (RCP8.5). Several studies have clearly outlined the uncertainty related to the choice of climate models (Chen et al. 2011; Giuntoli et al. 2015, 2018) with respect to future impacts of climate change. There is a large consensus on the benefits of using multimodel GCM ensembles when performing impact studies to adequately frame the uncertainty associated with GCM climate sensitivity. The methodological choice of using the ClimEx SMILE was based on its high temporal and spatial resolutions. This allowed long streamflow and precipitation return periods (up to 300 years) to be robustly sampled over a wide range of catchment sizes, without the need for any additional downscaling step. In fact, ClimEx data had to be upscaled at the catchment scale prior to the hydrological modeling. Upscaling is considered robust, while statistical downscaling of climate model data is considered hazardous, and especially so for the commonly used model output statistics methods (Maraun et al. 2017). It would not be possible to perform this work in the framework of a multimodel ensemble as no such ensemble exists with the appropriate spatial and temporal resolutions allowing to study rainfall and streamflow amplification on catchments with a clear subdaily response. Another advantage of using a large ensemble is that it controls for model uncertainty and captures the uncertainty related to internal climate variability. The RCP 8.5 scenario is no longer generally considered as a realistic scenario (e.g., Hausfather and Peters 2020) since current fossil fuel consumption no longer tracks well with the scenario. However, this high-emission scenario is still useful even though a +5°C world by the end of this century appears less and less likely. Climate simulations rarely run past 2100, and therefore, a +5°C future world may still be reached, albeit at a slower pace, than according to RCP8.5. A high-emission scenario results in larger impacts to be modeled, which allows for a clearer vision through the fog of climate internal variability, which is particularly important for precipitation (Chen et al. 2021; Deser et al. 2020). It would be worthwhile to redo this experiment with other climate models and emission scenarios when outputs are more commonly available at the proper spatial and temporal resolutions.
A single hydrological model was used in this study. The choice of a hydrological model is now known to have a potentially large impact on uncertainty (Giuntoli et al. 2018; Krysanova et al. 2018), although this has been shown mostly for low flow metrics. This is likely because most rainfall–runoff hydrological models are ill suited to modeling low flows. During droughts, streams are mostly fed from the water table, and groundwater models are best suited to this task. The hydrological model used in this study is a simple lumped conceptual model. Although this class of models has been shown to perform as well as more complex physically based models for streamflow simulation at a catchment outlet (e.g., Reed et al. 2004), its empirical nature may make it less suited for climate change impact studies, when the climate may have drifted from that of the reference period over which the hydrological model was calibrated. Of particular interest is the potential evapotranspiration formula, which is modeled as a function of temperature. The temperature sensitivity of evapotranspiration (ETP) formulations has raised concerns with respect to their suitability in climate change impact studies (e.g., Dallaire et al. 2021; Wang et al. 2017). However, the recent works of Lemaitre-Basset et al. (2021) and Seiller and Anctil (2016) point to ETP formulations not being a major source of uncertainty. In light of this, and since this study is concerned with high flows, for which rainfall–runoff models are considered robust, it is unlikely that using another hydrological model would lead to different conclusions.
The ClimEx SMILE has a high spatial resolution of 11 km. However, this high resolution remains too coarse to physically resolve deep convections resulting from subgrid processes. Convection is therefore parameterized in ClimEx data. It is now generally accepted that convection-permitting models are better able to simulate extreme rainfall (Lucas‐Picher et al. 2021). However, the resolution needed to run this class of model makes it expensive to run and most current studies are centered on relatively small computational domains and short time horizons. Convection-permitting models will eventually allow the study of future subhourly rainfall data to better frame the impacts on streamflow on small catchments. Many infrastructures require rainfall information at the subhourly time scale, and these models will permit a better understanding of rainfall and streamflow amplification at even finer scales than done here. Ultimately, streamflow amplification will need to be studied up to the acre/hectare scale, which is typical for urbanized subcatchments. This is roughly two orders of magnitude finer than the smallest catchments included in this study.
Finally, our results strongly reflect the climate of our study area (northeastern United States), which is essentially composed of two climate zones (from Köppen’s classification): the humid continental zone (Dfb) and the humid subtropical zone (Cfa). Therefore, any extension of the results of our study to other climatic zones must be done with caution. This is particularly the case for arid and semiarid zones where mean annual precipitation is expected to decrease in a warmer climate. However, recent observation-based work (e.g., Sun et al. 2021; Kirchmeier-Young and Zhang 2020) shows that extreme precipitation could increase in large parts of the world, and even in some regions where mean annual precipitation is decreasing. This suggests that the findings of our work may be applicable to many other climate zones around the world. Overall, our results point to increases in extreme rainfall and extreme streamflow across all durations, return periods and catchment sizes. However, the larger increases are systematically skewed toward shorter durations, longer return periods and smaller catchment sizes. It is clear that smaller rural and urban catchments will be significantly more impacted by the expected changes in future extreme rainfall as compared to larger catchments. It is therefore imperative to rapidly reviewed design rainfall and streamflow values (e.g., 100-yr flood) (e.g., Martel et al. 2021) since such values are ubiquitous in engineering design, and drainage infrastructures have typical lifespans exceeding 50 years. This is especially critical for impervious urbanized catchments, which are particularly vulnerable to rainfall increases.
5. Conclusions
This study assessed how future extreme rainfall and floods are impacted by catchment size, over a sample of 133 North American catchments ranging from 66.5 to 9886 km2. The ClimEx Single Model Initial-Condition Large Ensemble experiment was utilized to examine the intensification of extreme rainfall and floods in a warmer climate. This experiment offers 50 sets of climate variables at a 0.11° spatial resolution and time steps as frequent as hourly. We analyzed extreme rainfall and floods with durations ranging from 1 to 72 h and return periods spanning from 2 to 300 years across all catchments.
The main conclusions of this study are as follows:
-
An increase in extreme rainfall between the reference and future periods is observed in all catchments, for all durations and return periods. The increase gets progressively larger for the shorter duration and longer return periods for all catchments, which confirms the results of recent studies based on regional climate models and on observations.
-
The increase in extreme rainfall is largest in the smaller catchments, indicating that future changes in extreme rainfall are also strongly dependent on the spatial scale of future storms.
-
The pattern of increases in future extreme streamflow is very similar to that of future extreme rainfall. The largest increases for future extreme floods are observed in the smaller catchments, over shorter durations, and longest return periods. The relative increases are larger for streamflow than for rainfall. Extreme rainfall duration appears to be the most important factor in the amplification of future extreme events. Overall, the results presented in this study indicate that the projected future changes in extreme rainfall will disproportionately affect smaller catchments, especially urban areas, where flood management design criteria are based on short-duration long-return-period rainfall.
Acknowledgments.
This work was partly financed through the ClimEx project funded by the Bavarian State Ministry for the Environment and Consumer Protection. The authors acknowledge the contributions from the Canadian Centre for Climate Modelling and Analysis [Environment and Climate Change Canada (ECCC)] for simulating and making available the CanESM2-LE used in this study, and the Canadian Sea Ice and Snow Evolution Network for proposing the simulations. The authors would also like to thank the Ouranos Consortium for helping with data transfer. The CanESM2-LE dataset is now available on the ECCC website: http://collaboration.cmc.ec.gc.ca/cmc/cccma/CanSISE/output/CCCma/CanESM2/. The CESM1 ensemble was downloaded from the Large Ensemble Community Project (Schwarzwald and Lenssen 2022) website (http://www.cesm.ucar.edu/projects/community-projects/LENS/). The CRCM5 was developed by the ESCER Centre at Université du Québec à Montréal (UQAM; www.escer.uqam.ca) in collaboration with ECCC. Computations with the CRCM5 for the ClimEx project were made on the SuperMUC supercomputer at the Leibniz Supercomputing Centre (LRZ) of the Bavarian Academy of Sciences and Humanities. The operation of this supercomputer is funded via the Gauss Centre for Supercomputing (GCS) by the German Federal Ministry of Education and Research and the Bavarian State Ministry of Education, Science and the Arts.
Data availability statement.
All data and models used in this study can be found using the links in Table 2. The list of MOPEX catchments selected for this study is presented in the appendix (Table A1).
Data and models availability.
APPENDIX
USGS IDs of the Selected MOPEX Catchments
This appendix contains Table A1, which presents the list of MOPEX catchments selected for this study.
USGS ID of the selected MOPEX catchments.
REFERENCES
Arnbjerg-Nielsen, K., 2006: Significant climate change of extreme rainfall in Denmark. Water Sci. Technol., 54, 1–8, https://doi.org/10.2166/wst.2006.572.
Arora, V. K., and Coauthors, 2011: Carbon emission limits required to satisfy future representative concentration pathways of greenhouse gases. Geophys. Res. Lett., 38, L05805, https://doi.org/10.1029/2010GL046270.
Arsenault, R., A. Poulin, P. Côté, and F. Brissette, 2014: Comparison of stochastic optimization algorithms in hydrological model calibration. J. Hydrol. Eng., 19, 1374–1384, https://doi.org/10.1061/(ASCE)HE.1943-5584.0000938.
Arsenault, R., F. Brissette, and J.-L. Martel, 2018: The hazards of split-sample validation in hydrological model calibration. J. Hydrol., 566, 346–362, https://doi.org/10.1016/j.jhydrol.2018.09.027.
Ball, J. E., M. K. Babister, R. Nathan, P. E. Weinmann, W. Weeks, M. Retallick, and I. Testoni, 2016: Australian Rainfall and Runoff-A Guide to Flood Estimation. Commonwealth of Australia, 1526 pp.
Barker, P. M., and T. J. McDougall, 2020: Two interpolation methods using multiply-rotated piecewise cubic hermite interpolating polynomials. J. Atmos. Oceanic Technol., 37, 605–619, https://doi.org/10.1175/JTECH-D-19-0211.1.
Basso, S., M. Schirmer, and G. Botter, 2015: On the emergence of heavy-tailed streamflow distributions. Adv. Water Resour., 82, 98–105, https://doi.org/10.1016/j.advwatres.2015.04.013.
Bennett, J. C., D. E. Robertson, D. L. Shrestha, Q. J. Wang, D. Enever, P. Hapuarachchi, and N. K. Tuteja, 2014: A System for Continuous Hydrological Ensemble Forecasting (SCHEF) to lead times of 9 days. J. Hydrol., 519, 2832–2846, https://doi.org/10.1016/j.jhydrol.2014.08.010.
Berg, P., C. Moseley, and J. O. Haerter, 2013: Strong increase in convective precipitation in response to higher temperatures. Nat. Geosci., 6, 181–185, https://doi.org/10.1038/ngeo1731.
Bernardara, P., D. Schertzer, E. Sauquet, I. Tchiguirinskaia, and M. Lang, 2008: The flood probability distribution tail: How heavy is it? Stochastic Environ. Res. Risk Assess., 22, 107–122, https://doi.org/10.1007/s00477-006-0101-2.
Bertola, M., A. Viglione, D. Lun, J. Hall, and G. Blöschl, 2020: Flood trends in Europe: Are changes in small and big floods different? Hydrol. Earth Syst. Sci., 24, 1805–1822, https://doi.org/10.5194/hess-24-1805-2020.
Bormann, H., 2010: Runoff regime changes in German rivers due to climate change. Erdkunde, 64, 257–279, https://doi.org/10.3112/erdkunde.2010.03.04.
Brigode, P., L. Oudin, and C. Perrin, 2013: Hydrological model parameter instability: A source of additional uncertainty in estimating the hydrological impacts of climate change? J. Hydrol., 476, 410–425, https://doi.org/10.1016/j.jhydrol.2012.11.012.
Brooks, H. E., 2013: Severe thunderstorms and climate change. Atmos. Res., 123, 129–138, https://doi.org/10.1016/j.atmosres.2012.04.002.
Cannon, A. J., 2018: Multivariate quantile mapping bias correction: An N-dimensional probability density function transform for climate model simulations of multiple variables. Climate Dyn., 50, 31–49, https://doi.org/10.1007/s00382-017-3580-6.
Cannon, A. J., and S. Innocenti, 2019: Projected intensification of sub-daily and daily rainfall extremes in convection-permitting climate model simulations over North America: Implications for future intensity–duration–frequency curves. Nat. Hazards Earth Syst. Sci., 19, 421–440, https://doi.org/10.5194/nhess-19-421-2019.
Capek, T. J., 2021: Understanding the effects of water vapor and temperature on aerosol using novel measurement methods. Ph.D. dissertation, Michigan Technological University, 152 pp.
Chen, H., C.-Y. Xu, and S. Guo, 2012: Comparison and evaluation of multiple GCMs, statistical downscaling and hydrological models in the study of climate change impacts on runoff. J. Hydrol., 434–435, 36–45, https://doi.org/10.1016/j.jhydrol.2012.02.040.
Chen, J., F. P. Brissette, A. Poulin, and R. Leconte, 2011: Overall uncertainty study of the hydrological impacts of climate change for a Canadian watershed. Water Resour. Res., 47, W12509, https://doi.org/10.1029/2011WR010602.
Chen, J., F. P. Brissette, X. J. Zhang, H. Chen, S. Guo, and Y. Zhao, 2019: Bias correcting climate model multi-member ensembles to assess climate change impacts on hydrology. Climatic Change, 153, 361–377, https://doi.org/10.1007/s10584-019-02393-x.
Chen, J., X. Li, J.-L. Martel, F. P. Brissette, X. J. Zhang, and A. Frei, 2021: Relative importance of internal climate variability versus anthropogenic climate change in global climate change. J. Climate, 34, 465–478, https://doi.org/10.1175/JCLI-D-20-0424.1.
Cheng, L., and A. AghaKouchak, 2014: Nonstationary precipitation intensity-duration-frequency curves for infrastructure design in a changing climate. Sci. Rep., 4, 7093, https://doi.org/10.1038/srep07093.
Crochemore, L., M.-H. Ramos, and F. Pappenberger, 2016: Bias correcting precipitation forecasts to improve the skill of seasonal streamflow forecasts. Hydrol. Earth Syst. Sci., 20, 3601–3618, https://doi.org/10.5194/hess-20-3601-2016.
Croitoru, A.-E., and I. Minea, 2015: The impact of climate changes on rivers discharge in eastern Romania. Theor. Appl. Climatol., 120, 563–573, https://doi.org/10.1007/s00704-014-1194-z.
Cuo, L., T. C. Pagano, and Q. J. Wang, 2011: A review of quantitative precipitation forecasts and their use in short-to medium-range streamflow forecasting. J. Hydrometeor., 12, 713–728, https://doi.org/10.1175/2011JHM1347.1.
Dallaire, G., A. Poulin, R. Arsenault, and F. Brissette, 2021: Uncertainty of potential evapotranspiration modelling in climate change impact studies on low flows in North America. Hydrol. Sci. J., 66, 689–702, https://doi.org/10.1080/02626667.2021.1888955.
Deser, C., A. Phillips, V. Bourdette, and H. Teng, 2012: Uncertainty in climate change projections: The role of internal variability. Climate Dyn., 38, 527–546, https://doi.org/10.1007/s00382-010-0977-x.
Deser, C., and Coauthors, 2020: Insights from Earth system model initial-condition large ensembles and future prospects. Nat. Climate Change, 10, 277–286, https://doi.org/10.1038/s41558-020-0731-2.
Diffenbaugh, N. S., M. Scherer, and R. J. Trapp, 2013: Robust increases in severe thunderstorm environments in response to greenhouse forcing. Proc. Natl. Acad. Sci. USA, 110, 16 361–16 366, https://doi.org/10.1073/pnas.1307758110.
Donat, M. G., and Coauthors, 2013: Updated analyses of temperature and precipitation extreme indices since the beginning of the twentieth century: The HadEX2 dataset. J. Geophys. Res. Atmos., 118, 2098–2118, https://doi.org/10.1002/jgrd.50150.
Dougherty, E., and K. L. Rasmussen, 2020: Changes in future flash flood–Producing storms in the United States. J. Hydrometeor., 21, 2221–2236, https://doi.org/10.1175/JHM-D-20-0014.1.
Dougherty, E., E. Sherman, and K. L. Rasmussen, 2020: Future changes in the hydrologic cycle associated with flood-producing storms in California. J. Hydrometeor., 21, 2607–2621, https://doi.org/10.1175/JHM-D-20-0067.1.
Duan, Q., S. Sorooshian, and V. K. Gupta, 1994: Optimal use of the SCE-UA global optimization method for calibrating watershed models. J. Hydrol., 158, 265–284, https://doi.org/10.1016/0022-1694(94)90057-4.
Duan, Q., and Coauthors, 2006: Model Parameter Estimation Experiment (MOPEX): An overview of science strategy and major results from the second and third workshops. J. Hydrol., 320, 3–17, https://doi.org/10.1016/j.jhydrol.2005.07.031.
Ehmele, F., L.-A. Kautz, H. Feldmann, and J. G. Pinto, 2020: Long-term variance of heavy precipitation across central Europe using a large ensemble of regional climate model simulations. Earth Syst. Dyn., 11, 469–490, https://doi.org/10.5194/esd-11-469-2020.
Epstein, M. P., 1976: On the influence of parametrization in parametric interpolation. SIAM J. Numer. Anal., 13, 261–268, https://doi.org/10.1137/0713025.
Faghih, M., F. Brissette, and P. Sabeti, 2022: Impact of correcting sub-daily climate model biases for hydrological studies. Hydrol. Earth Syst. Sci., 26, 1545–1563, https://doi.org/10.5194/hess-26-1545-2022.
Feng, Z., and Coauthors, 2016: More frequent intense and long-lived storms dominate the springtime trend in central US rainfall. Nat. Commun., 7, 13429, https://doi.org/10.1038/ncomms13429.
Fildier, B., H. Parishani, and W. D. Collins, 2017: Simultaneous characterization of mesoscale and convective‐scale tropical rainfall extremes and their dynamical and thermodynamic modes of change. J. Adv. Model. Earth Syst., 9, 2103–2119, https://doi.org/10.1002/2017MS001033.
Forestieri, A., E. Arnone, S. Blenkinsop, A. Candela, H. Fowler, and L. V. Noto, 2018: The impact of climate change on extreme precipitation in Sicily, Italy. Hydrol. Processes, 32, 332–348, https://doi.org/10.1002/hyp.11421.
Fowler, H. J., C. Wasko, and A. F. Prein, 2021: Intensification of short-duration rainfall extremes and implications for flood risk: Current state of the art and future directions. Philos. Trans. Roy. Soc., A379, 20190541, https://doi.org/10.1098/rsta.2019.0541.
Frankcombe, L. M., M. H. England, M. E. Mann, and B. A. Steinman, 2015: Separating internal variability from the externally forced climate response. J. Climate, 28, 8184–8202, https://doi.org/10.1175/JCLI-D-15-0069.1.
Ganguli, P., and P. Coulibaly, 2019: Assessment of future changes in intensity-duration-frequency curves for Southern Ontario using North American (NA)-CORDEX models with nonstationary methods. J. Hydrol., 22, 100587, https://doi.org/10.1016/j.ejrh.2018.12.007.
Gensini, V. A., and T. L. Mote, 2015: Downscaled estimates of late 21st century severe weather from CCSM3. Climatic Change, 129, 307–321, https://doi.org/10.1007/s10584-014-1320-z.
Giuntoli, I., J.-P. Vidal, C. Prudhomme, and D. M. Hannah, 2015: Future hydrological extremes: The uncertainty from multiple global climate and global hydrological models. Earth Syst. Dyn., 6, 267–285, https://doi.org/10.5194/esd-6-267-2015.
Giuntoli, I., G. Villarini, C. Prudhomme, and D. M. Hannah, 2018: Uncertainties in projected runoff over the conterminous United States. Climatic Change, 150, 149–162, https://doi.org/10.1007/s10584-018-2280-5.
Gobiet, A., S. Kotlarski, M. Beniston, G. Heinrich, J. Rajczak, and M. Stoffel, 2014: 21st century climate change in the European Alps—A review. Sci. Total Environ., 493, 1138–1151, https://doi.org/10.1016/j.scitotenv.2013.07.050.
Guerreiro, S. B., H. J. Fowler, R. Barbero, S. Westra, G. Lenderink, S. Blenkinsop, E. Lewis, and X.-F. Li, 2018: Detection of continental-scale intensification of hourly rainfall extremes. Nat. Climate Change, 8, 803–807, https://doi.org/10.1038/s41558-018-0245-3.
Hagemann, S., C. Chen, J. O. Haerter, J. Heinke, D. Gerten, and C. Piani, 2011: Impact of a statistical bias correction on the projected hydrological changes obtained from three GCMs and two hydrology models. J. Hydrometeor., 12, 556–578, https://doi.org/10.1175/2011JHM1336.1.
Harlan, D., M. Wangsadipura, and C. M. Munajat, 2010: Rainfall-Runoff Modeling of Citarum Hulu river basin by using GR4j. Proc. World Congress on Engineering 2010, Vol. II, London, United Kingdom, World Congress on Engineering, 1–5, https://www.iaeng.org/publication/WCE2010/WCE2010_pp1607-1611.pdf.
Hausfather, Z., and G. P. Peters, 2020: RCP8.5 is a problematic scenario for near-term emissions. Proc. Natl. Acad. Sci. USA, 117, 27 791–27 792, https://doi.org/10.1073/pnas.2017124117.
Hersbach, H., and Coauthors, 2018: ERA5 hourly data on single levels from 1979 to present. Copernicus Climate Change Service (C3S), Climate Data Store (CDS), accessed 10 May 2020, https://doi.org/10.24381/cds.adbb2d47.
Hosseinzadehtalaei, P., H. Tabari, and P. Willems, 2020: Climate change impact on short-duration extreme precipitation and intensity–duration–frequency curves over Europe. J. Hydrol., 590, 125249, https://doi.org/10.1016/j.jhydrol.2020.125249.
IPCC, 2007: Climate Change 2007: The Physical Science Basis. Cambridge University Press, 996 pp.
Kirchmeier-Young, M. C., and X. Zhang, 2020: Human influence has intensified extreme precipitation in North America. Proc. Natl. Acad. Sci. USA, 117, 13 308–13 313, https://doi.org/10.1073/pnas.1921628117.
Krysanova, V., C. Donnelly, A. Gelfan, D. Gerten, B. Arheimer, F. Hattermann, and Z. W. Kundzewicz, 2018: How the performance of hydrological models relates to credibility of projections under climate change. Hydrol. Sci. J., 63, 696–720, https://doi.org/10.1080/02626667.2018.1446214.
Kunnath-Poovakka, A., and T. I. Eldho, 2019: A comparative study of conceptual rainfall-runoff models GR4J, AWBM and Sacramento at catchments in the upper Godavari River basin, India. J. Earth Syst. Sci., 128, 33, https://doi.org/10.1007/s12040-018-1055-8.
Leduc, M., and Coauthors, 2019: The ClimEx project: A 50-member ensemble of climate change projections at 12-km resolution over Europe and northeastern North America with the Canadian Regional Climate Model (CRCM5). J. Appl. Meteor. Climatol., 58, 663–693, https://doi.org/10.1175/JAMC-D-18-0021.1.
Lemaitre-Basset, T., L. Oudin, G. Thirel, and L. Collet, 2021: Uncertainty on evapotranspiration formulation and its hydrological implication under climate change over France. EGU General Assembly 2021, Online, Abstract EGU21-9946, https://doi.org/10.5194/egusphere-egu21-9946.
Lenderink, G., and E. van Meijgaard, 2008: Increase in hourly precipitation extremes beyond expectations from temperature changes. Nat. Geosci., 1, 511–514, https://doi.org/10.1038/ngeo262.
Lenderink, G., and E. Van Meijgaard, 2010: Linking increases in hourly precipitation extremes to atmospheric temperature and moisture changes. Environ. Res. Lett., 5, 025208, https://doi.org/10.1088/1748-9326/5/2/025208.
Li, F., Y. Zhang, Z. Xu, J. Teng, C. Liu, W. Liu, and F. Mpelasoka, 2013: The impact of climate change on runoff in the southeastern Tibetan Plateau. J. Hydrol., 505, 188–201, https://doi.org/10.1016/j.jhydrol.2013.09.052.
Lucas‐Picher, P., D. Argüeso, E. Brisson, Y. Tramblay, P. Berg, A. Lemonsu, S. Kotlarski, and C. Caillaud, 2021: Convection‐permitting modeling with regional climate models: Latest developments and next steps. Wiley Interdiscip. Rev.: Climate Change, 12, e731, https://doi.org/10.1002/wcc.731.
Maher, N., S. Milinski, and R. Ludwig, 2021: Large ensemble climate model simulations: Introduction, overview, and future prospects for utilising multiple types of large ensemble. Earth Syst. Dyn., 12, 401–418, https://doi.org/10.5194/esd-12-401-2021.
Maraun, D., and Coauthors, 2017: Towards process-informed bias correction of climate change simulations. Nat. Climate Change, 7, 764–773, https://doi.org/10.1038/nclimate3418.
Markonis, Y., S. Papalexiou, M. Martinkova, and M. Hanel, 2019: Assessment of water cycle intensification over land using a multisource global gridded precipitation dataset. J. Geophys. Res. Atmos., 124, 11 175–11 187, https://doi.org/10.1029/2019JD030855.
Martel, J.-L., A. Mailhot, and F. Brissette, 2020: Global and regional projected changes in 100-yr subdaily, daily, and multiday precipitation extremes estimated from three large ensembles of climate simulations. J. Climate, 33, 1089–1103, https://doi.org/10.1175/JCLI-D-18-0764.1.
Martel, J.-L., F. P. Brissette, P. Lucas-Picher, M. Troin, and R. Arsenault, 2021: Climate change and rainfall intensity–duration–frequency curves: Overview of science and guidelines for adaptation. J. Hydrol. Eng., 26, 03121001, https://doi.org/10.1061/(ASCE)HE.1943-5584.0002122.
Mathevet, T., 2005: Quels modèles pluie-débit globaux au pas de temps horaire? Développements empiriques et comparaison de modèles sur un large échantillon de bassins versants. Ph.D. thesis, Ecole Nationale du Génie Rural, 463 pp.
McKinnon, K. A., and C. Deser, 2018: Internal variability and regional climate trends in an observational large ensemble. J. Climate, 31, 6783–6802, https://doi.org/10.1175/JCLI-D-17-0901.1.
Milly, P. C. D., J. Betancourt, M. Falkenmark, R. M. Hirsch, Z. W. Kundzewicz, D. P. Lettenmaier, and R. J. Stouffer, 2008: Stationarity is dead: Whither water management? Science, 319, 573–574, https://doi.org/10.1126/science.1151915.
Mineo, C., E. Ridolfi, F. Napolitano, and F. Russo, 2018: The areal reduction factor: A new analytical expression for the Lazio region in central Italy. J. Hydrol., 560, 471–479, https://doi.org/10.1016/j.jhydrol.2018.03.033.
Minville, M., F. Brissette, and R. Leconte, 2008: Uncertainty of the impact of climate change on the hydrology of a Nordic watershed. J. Hydrol., 358, 70–83, https://doi.org/10.1016/j.jhydrol.2008.05.033.
Moustakis, Y., S. M. Papalexiou, C. J. Onof, and A. Paschalis, 2021: Seasonality, intensity, and duration of rainfall extremes change in a warmer climate. Earth’s Future, 9, e2020EF001824, https://doi.org/10.1029/2020EF001824.
Nash, J. E., and J. V. Sutcliffe, 1970: River flow forecasting through conceptual models Part I—A discussion of principles. J. Hydrol., 10, 282–290, https://doi.org/10.1016/0022-1694(70)90255-6.
O’Brien, T. A., W. D. Collins, K. Kashinath, O. Rübel, S. Byna, J. Gu, H. Krishnan, and P. A. Ullrich, 2016: Resolution dependence of precipitation statistical fidelity in hindcast simulations. J. Adv. Model. Earth Syst., 8, 976–990, https://doi.org/10.1002/2016MS000671.