Several extensive and persistent droughts have affected the Canadian Prairies, most notably in the 1910s, 1930s, 1960s and 1980s (Nkemdirm and Weber, 1999; Chipanshi et al., 2006), with the most severe, single year drought occurring in 1961 (Maybank et al., 1995). The multi-year drought conditions of 1999–2002 were the most severe in 100 years with below normal precipitation for eight consecutive seasons from September 2000 to August 2002 (Bonsal and Wheaton, 2005; Bonsal and Regier, 2007). According to third assessment report of the Intergovernmental Panel on Climate Change (IPCC, 2001), the Canadian Prairies will possibly face a further increase in drought frequency in future climate, which would have significant impacts on a wide range of sectors including agriculture, forestry, industry, ecosystem and health; it is therefore desirable to perform thorough assessments of future changes to various drought characteristics to enable improved adaptation strategies for the region.
The primary tools used to study anticipated climate change are Coupled Global Climate Models (CGCMs) and Regional Climate Models (RCMs) integrated from the recent past to some time in the future with a prescribed, time-evolving emission scenario for anthropogenic greenhouse gases and aerosols (GHGA) (IPCC, 2007). At this point, RCMs offer higher spatial resolution than GCMs, allowing for greater topographic complexity and finer scale atmospheric dynamics to be simulated, and thereby representing a more adequate tool for generating climate change information required for many impact and adaptation studies. GCM projections were used by Bonsal and Regier (2006) to assess future changes to drought characteristics in Canada for the period 2041–2070; their study suggests an increase in the frequency and severity of droughts in southern Canada. More recently, Sushama et al. (2010) projected changes to dry spell characteristics over Canada using the Canadian RCM (CRCM); their results suggest increases in the number of dry days and 10- and 30 year return levels of maximum dry spell duration for the Canadian Prairies, particularly the southern parts of the region, for 2041–2070 and 2071–2100 future periods.
Several indices have been developed to study the frequency, severity and persistence of droughts (Palmer, 1965; van Rooy, 1965; Bhalme and Mooley, 1980; McKee et al., 1993; Meyer et al., 1993; Phillips and McGregor, 1998), with the majority being based on meteorological and hydrological parameters, i.e. precipitation, evapotranspiration, temperature, soil moisture, runoff, etc. However, except for precipitation, long-term values of other hydroclimatic parameters are not available for the Canadian Prairies (Bonsal and Regier, 2007); therefore indices solely based on precipitation are commonly used to assess drought characteristics in this region, such as the Standardized Precipitation Index (SPI), originally defined in McKee et al. (1993). The SPI was used by Quiring and Papakryiakou (2003) and Bonsal and Regier (2007) to study drought characteristics over Canadian Prairies. Over Europe, however, many RCM-based studies focusing on future changes to drought characteristics have used other indices based on cumulative monthly precipitation anomalies (Fowler and Kilsby, 2002; Blenkinshop and Fowler, 2007a, 2007b). These studies have investigated the frequency and severity of short-term droughts of 3–6 months duration, and long-term droughts of at least 6 and 10 months duration. The above cumulative monthly precipitation anomalies-based drought indices and events have been the most commonly used approaches due to their simplicity, small data requirements and easy interpretation (Phillips and McGregor, 1998; Fowler and Kilsby, 2002; Blenkinshop and Fowler, 2007a, 2007b). We therefore in this study analyse cumulative monthly precipitation anomaly-based drought indices and events derived from the CRCM simulated precipitation series for the current (1971–2000) and future (2041–2070) periods to assess projected changes to the severity, frequency and duration of short- and long-term droughts, over the Canadian Prairies.
The paper is organised as follows. Section 2 describes the CRCM and its simulations and observation dataset. A detailed methodology for the estimation of precipitation-based drought indices and drought events is described in Section 3. In Section 4, suitability of various precipitation-based drought events for the Canadian Prairies, validation of CRCM simulated short- and long-term drought characteristics and their projected changes are presented. Discussion and conclusions of the study are given in section 5.
2. Model and data
2.1. Model description
The precipitation outputs used in this study to define droughts are generated by the fourth generation of the CRCM (Music and Caya, 2007), which is the current operational version. The CRCM is a fully elastic non-hydrostatic limited-area nested model. It uses a semi-implicit and semi-Lagrangian numerical scheme to solve the basic non-hydrostatic Euler equations (Caya, 1996; Laprise et al., 1998; Caya and Laprise, 1999). The model's horizontal grid is uniform in polar stereographic projection with 45 km horizontal resolution and uses 15 min time step. The model uses Gal-Chen scaled height terrain-following vertical coordinates (Gal-Chen and Somerville, 1975) with 29 levels in the vertical and the top of the domain is located at 29 km altitude.
Spectral nudging technique was implemented to large-scale winds within the regional domain for all CRCM simulations used in this study in order to achieve a better representation of large-scale flow of the driving data (Frigon et al., 2010). The spectral nudging technique applied for CRCM, developed by Riette and Caya (2002), uses the spectral decomposition from Denis et al. (2002). The spectral nudging applied is considered relatively weak, where horizontal winds with wavelengths greater than 1400 km are nudged with varying intensity at the vertical, starting just above 500 hPa and reaching a characteristic relaxation time of 10 h at the model top (near 10 hPa).
The CRCM's lateral boundary conditions are provided through one-way nesting method over a regional domain inspired by Davies (1976) and redefined by Yakimiw and Robert (1990). Therefore, the CRCM receives atmospheric nesting information from its driving data, but does not influence the driving data in return. The CRCM is driven by time-dependent vertical profiles from the driving data's wind, air temperature, humidity and pressure imposed at the lateral boundaries exactly, as interpolated onto the CRCM's atmospheric levels. The simulated horizontal winds are relaxed toward values of the driving data over the sponge zone. In addition, spectral large-scale nudging is also imposed to force coherence of the CRCM large-scale winds with the driving data (Biner et al., 2000).
The CRCM generally uses most of the sub-grid scale physical parameterization packages of the Coupled General Circulation Model (CGCM3.1; Flato and Boer, 2001), except for moist convection. Cloud cover is parameterized in terms of local relative humidity assuming maximum (random) overlap, depending on presence (or absence) of clouds in adjacent layers as in CGCM3.1 and precipitation is parameterized in terms of a simple super saturation-based condensation scheme as in CGCM3.1 (Laprise et al., 2003). Mesoscale convection follows the parameterization scheme of Kain and Fritsch (1990) and Bechtold et al. (2001).
An ensemble of ten 30 year simulations are analysed in this paper, of which five correspond to current 1971–2000 period and the other five are the matching simulations for the future 2041–2070 period; these five pairs of current and future CRCM simulations were driven by five different members of a CGCM3.1 initial condition ensemble. RCM simulations in general are associated with several uncertainties including structural uncertainties associated with regional model formulation, uncertainties associated with the lateral boundary conditions from the driving GCM, emission scenarios, as well as the RCM's own internal variability (de-Elia et al., 2008). The ensemble of CRCM used in this study would be helpful in addressing the combined uncertainty associated with the natural variability of the driving GCM and the internal variability of CRCM. The future simulations are affected by changes in GHGA, following IPCC's (2001) Special Report on Emissions Scenarios (SRES) A2 scenario. Though in this study we focus only on the Canadian Prairies, the model simulations were performed over a 200 × 192 point grid covering the whole of North America (see inset in Figure 1). Figure 1 shows the study region, and the spatial distribution of major/minor watersheds within the region; the names of the various watersheds are provided in Table I. There are 47 watersheds and some of these extend into the United States, e.g. Red River and Winnipeg River. In the rest of the paper, results will only be displayed for the subset of the domain corresponding to the 47 Prairie watersheds.
Table I. Identification numbers (IDs) and names of 47 watersheds, located in the Canadian Prairies. Their distribution is shown in Figure 1
Red Deer River
Beaver Stone River
Big Muddy Creek
Seven Persons Creek
North Saskatchewan River
Old Wives Lake
South Saskatchewan River
Swift Current Creek
The Climate Research Unit (CRU) dataset (CRU TS 2.1; Mitchell et al., 2004), which consists of gridded global land surface observations of monthly precipitation at 0.5° × 0.5° resolution, is used in this study to validate model simulated precipitation and drought characteristics because of its high-spatial resolution. The intercomparison of CRU data, WM global precipitation dataset (Willmott and Matsuura, 2001), available at 0.5° × 0.5°, and the Global Precipitation Climatology Centre Dataset (GPCC; Rudolf et al., 1994), available on a 1° × 1° grid, shows substantial numerical differences with largest absolute differences in the wet Tropics and largest relative difference in the dry regions. Despite these differences, the overall precipitation pattern in CRU data is fairly similar to that of WM and GPCC global precipitation dataset (Fekete and Vörömarty, 2004). The correlation between CRU and GPCC (WM) annual precipitation is about 0.92 (0.98) globally and 0.76 (0.83) for Canadian Prairies. The differences between the CRU and GPCC (WM) annual precipitation generally lie within ± 0.06 ( ± 0.05) mm d−1 range for the Canadian Prairies. As the precipitation patterns for the three observed datasets for the Canadian Prairies are fairly similar, we have decided to use only CRU as the observed dataset in this study. The CRU data is interpolated via inverse distance weighting (IDW) method to the CRCM grid prior to comparisons and computation of drought characteristics.
In this study, three drought severity indices, DSI3, DSI6 and DSI10 based on the accumulated monthly precipitation deficit concept of Bryant et al. (1992) are applied to the Canadian Prairies. The DSI3 index uses a 3 month initiation and termination rule, after Bryant et al. (1992, 1994). The calculation of DSI3 is briefly explained below. Similarly, the DSI6 (DSI10) uses a six-monthly (ten-monthly) rule, after Marsh et al. (1994) and further elaborated in the work of Phillips and McGregor (1998) and Blenkinsop and Fowler (2007a, 2007b).
The calculations of DSIs are based on monthly precipitation anomalies with respect to the 1971–2000 reference climatology. If xt represents precipitation anomaly in month t, then DSI3 is triggered at month t whenever the mean three-monthly precipitation in the preceding three months (i.e. t − 1, t − 2, and t − 3) is less than the mean three-monthly climatologic precipitation for the same months for the reference 1971–2000 period and the month t is associated with a precipitation deficit (i.e. xt < 0; DSI3 is then initialised to xt. The DSI3s for the following months are computed by adding respective monthly anomalies to previous month's DSI3. However, if for any month, the mean three-monthly precipitation received in the preceding three months is above the mean three-monthly climatologic precipitation, drought termination occurs and DSI3 is assigned a zero value. The computed DSI3s are then expressed as a percentage of the annual mean precipitation. The DSI6 and DSI10 are calculated in a similar manner except that six- and ten-monthly periods are used in the calculations. The DSI6 and DSI10 are indicative of more severe drought than DSI3 because longer lasting precipitation deficits will be required over the preceding two or three seasons to initiate such severe droughts for a particular season.
The short- and long-term drought events DRO3, DRO6 and DRO10, defined as droughts lasting between 3 to 6 months, and at least 6 and 10 months, respectively, are based on DSI3, DSI6 and DSI10, with accumulated deficits exceeding 10% of mean annual precipitation. The DRO3 events are likely to influence surface water resources and represent only weak drought conditions, while the DRO6 and DRO10 events are likely to affect groundwater resources and hence represent strong drought conditions. Additional details and discussion about these indicators of drought conditions can also be found in Fowler and Kilsby (2002, 2004) and Blenkinsop and Fowler (2007a, 2007b). Every occurrence of DRO3, DRO6 and DRO10 event is associated with a maximum precipitation deficit and duration, defined as the time (in months) between drought initiation and termination. In this study, severity is defined as the average of absolute maximum precipitation deficits for each category of DRO events, while frequency is taken as the number of occurrences over a given period of time. The DSIs and DROs are computed for the observed data using monthly precipitation anomalies with respect to the 1971–2000 observed climatology and for five current and five future CRCM simulations using monthly precipitation anomalies with respect to the CRCM climatology for the 1971–2000 period. It must be noted that all figures presented in this article correspond to ensemble means, unless indicated otherwise.
4.1. Evaluation of current climate simulation (1971–2000)
As the drought indices considered in this study are based on precipitation, we start by validating model-simulated precipitation. The errors in the CRCM simulated precipitation over the region corresponding to the 47 watersheds are assessed first by comparing simulated average seasonal precipitation with gridded seasonal climatology of precipitation from CRU for the period 1971–2000. Figure 2 reveals that the model captures the observed distribution of winter (DJF), spring (MAM) and fall (SON) precipitation across a large portion of the Canadian Prairies, with errors within the ± 40% range, with the exception of south-western Alberta where the model overestimates precipitation by up to 50% in winter (DJF) along the eastern slopes of the Rocky Mountains (Figure 2(a), (b)). The large errors over south-western Alberta can be explained by the underestimation of the orographic effect in the model, yielding an underestimation of precipitation in the high-elevated region and a corresponding overestimation in the downwind region. Summer precipitation is overestimated across a substantial part of the Canadian Prairies (on average up to 60%), except western central Alberta where there is a slight underestimation (Figure 2(e), (f)); summer errors reach up to 100% over much of Alberta and Saskatchewan. Since summer precipitation is mostly convective, it is believed that the summer precipitation error is mostly associated with the inadequacies of the convective parameterizations in the model (Caya and Biner, 2004; Plummer et al., 2006).
In addition to the seasonal averages, simulated and observed interannual precipitation variability is compared. Interannual standard deviation of seasonal precipitation is normalized by corresponding seasonal average to compute coefficient of variation. The observed precipitation coefficient of variation varies between 0.15 and 0.53 throughout the year, with largest coefficient of variation over south-western Prairies in winter (DJF) (Figure 3(a)) and southern Prairies in fall (SON) (Figure 3(g)), while the CRCM simulated coefficient of variation varies between 0.06 and 0.20 with the highest coefficient of variation over Saskatchewan and Manitoba in spring (MAM) (Figure 3(d)) and southern Saskatchewan in fall (SON) (Figure 3(h)). Overall, the model generally underestimates observed interannual variability, with relatively better agreement in spring. In winter (DJF) and fall (SON), the model underestimates the observed coefficient of variation by up to 75% over southern and central Alberta and Saskatchewan and the eastern slopes of the Rocky Mountains. The evaluation of interannual variability for summer (JJA) is influenced by the precipitation errors for this season with highest underestimation of observed variability over northern and north-western Alberta (Figures 2(e), (f), and 3(e), (f)).
4.1.2. Drought characteristics
Before comparing the model simulated characteristics of DRO events with those derived from CRU data, the usefulness of the DRO events as indicators of drought conditions for the Canadian Prairies is investigated. Severity of DRO events, derived from the CRU dataset, for the 1910s, 1930s, 1980, 1984, 1989 and 1999–2002 drought events are shown in Figure 4. Note that for multi-year droughts, i.e. 1910s, 1930s and 1999–2002 periods, DRO10 events are considered, while in the case of single year droughts, i.e. 1980, 1984 and 1989, DRO3 events are considered. The spatial distribution of drought affected areas in Figure 4 is in good agreement with the findings of Nkemdirim and Weber (1999), Chipanshi et al. (2006), Bonsal and Regier (2007); the most severe drought occurred generally over western Alberta, Saskatchewan and south-eastern Manitoba in the 1910s (Figure 4(a)), over southern and central Alberta and Saskatchewan and eastern and south-eastern Manitoba in the 1930s (Figure 4(c)), over northern and southern Saskatchewan in 1980 (Figure 4(b)), over the southern Prairies in 1984 (Figure 4(d)), over northern Alberta, south-eastern Saskatchewan and central and southern Manitoba in 1989 (Figure 4(f)) and over south-western and south-eastern Alberta, south-western Saskatchewan during the period 1999–2002 (Figure 4(e)). The above confirms the suitability of DROs for studying droughts over the Canadian Prairies.
We now compare the CRCM simulated short- and long-term drought characteristics (i.e. severity and frequency of DROs) with those derived from the CRU dataset for the current climate (1971–2000) in Figures 5 and 6. The model has difficulties in reproducing the observed severity of long-term droughts, i.e. DRO6 and DRO10 events, over northern Alberta and Manitoba, central Alberta and the southern Prairies (Figure 5(c)–(f)). However, the model captures the severity of the shorter droughts (i.e. DRO3 events) (Figure 5(a), (b)). Severity of DRO6 events is consistently underestimated for a substantial part of the Canadian Prairies and the same conclusion holds for DRO10 events. In general, the model has difficulties in reproducing severities of drought events over the north western and southern Prairies, while it reasonably reproduces severities elsewhere (Figure 5).
The model also has some difficulties in reproducing the observed patterns of frequency of occurrence of DRO3, DRO6 and DRO10 events, especially over regions that experience highest and lowest frequencies of drought events (Figure 6). It underestimates the occurrences of 3- to 6 month long droughts by up to eight events over southern Alberta and Saskatchewan, northern and north-western Alberta and north-eastern Manitoba and overestimates by up to three events over the southern Manitoba (Figure 6(a), (b)). For droughts lasting at least 6 months, the model produces negative biases over southern Prairies, south-western Alberta and northern Alberta and Manitoba (Figure 6(c), (d)). The model also underestimates the frequency of DRO10 events by up to five events over south-western and northern Alberta and overestimates by up to three events over north-eastern Saskatchewan and south-western Manitoba (Figure 6(e), (f)). However, the underestimations of drought occurrence are largest for DRO6 events and are believed to be related to the excessive CRCM summer precipitation, which reduces DSI6 values based on 6 month initiation and termination rules and thus reduces the frequency of occurrence of DRO6 events. Similarly, the overestimations of drought occurrence of DRO10 events are believed to be due to the underestimation of fall and winter precipitation.
The model underestimates the maximum duration of DRO6 and DRO10 events over a greater part of Alberta and Manitoba as shown in Figure 7, while it is overestimated over parts of Saskatchewan.
4.2. Future projections
Projected changes to average seasonal precipitation in the future 2041–2070 period with respect to the current 1971–2000 period are shown in Figure 8. The model projects widespread precipitation increases across the Canadian Prairies for all seasons except summer (JJA). In winter (DJF), large relative increases in precipitation (up to 50%) are projected for northern and north-eastern Manitoba, and a secondary increase in precipitation is projected to occur along the eastern slopes of the Rocky Mountains (Figure 8(a), (b)). In spring (MAM), the model projects up to a 30% increase in precipitation over the eastern Prairies (Manitoba) (Figure 8(c), (d)). The projected change for the summer (JJA) season is more mixed, with the largest relative decrease (up to 10%) in precipitation over the western and southern Prairies (Figure 8(e), (f)). The maxima in the summer (JJA) precipitation are likely to decrease from the current climate to future climate over the Rocky Mountains in western Alberta. For fall (SON), the model projects increased precipitation along the eastern slopes of the Rocky Mountains (Figure 8(g), (h)).
4.2.2. Drought characteristics
Future changes in the characteristics of drought events for the period 2041–2070 exhibit an increase in the severity of short and long drought events across the southern parts of the Canadian Prairies and along the eastern slopes of the Rocky Mountains (Figure 9). However, increases in the severity of longer drought events (i.e. DRO10s and DRO6s) are larger in comparison to shorter drought events (i.e. DRO3s). Thus, in general, future projections of DRO3 type drought events exhibit the smallest change in severity across the Canadian Prairies.
For DRO3 type drought events, a large portion of southern Saskatchewan and Manitoba and eastern slopes of the Rocky Mountains is projected to experience more frequent droughts (Figure 10(a), (b)), while results suggest a decrease for the northern Prairies. The spatial patterns of projected changes associated with DRO6 events are similar to those of DRO3 events, however with generally larger magnitudes (Figure 10(c), (d)). The frequency of DRO10 type events is projected to increase by up to four events during 2041–2070 over southern Manitoba and Saskatchewan and eastern slopes of the Rocky Mountains (Figure 10(e), (f)). Similar to DRO6 events, the projected increases are generally larger in magnitude than those of shorter drought events resulting in more frequent and more severe longer drought events. The largest projected increase in the frequency of both short- and long-term drought events over southern Prairies is believed to be due to the projected decrease in summer precipitation (Figures 8(e), (f) and 10(e), (f)), while projected decreases in the drought events can be attributed to the large projected increases in winter (DJF), spring (MAM) and fall (SON) precipitation over northern, eastern and north-eastern Prairies which interrupt drought sequences.
The maximum duration of DRO6 events is projected to increase by up to 10 months over a few grid cells in the southern Prairies and along the eastern slopes of the Rocky Mountains and decrease by up to 10 months elsewhere (Figure 11(a), (b)). An increase (by more than 20 months) in the maximum duration of DRO10 type events is also projected across a larger proportion of grid cells over the southern Prairies and along the eastern slopes of the Rocky Mountains (Figure 11(c), (d)). This study thus indicates that in the future, the severity, frequency and maximum duration of all drought events are likely to increase over southern Prairies, with the larger projected changes associated with longer drought events and hence, increased pressure on water resources, especially groundwater, due to decreases in recharge during cooler and wetter seasons.
4.2.3. Uncertainty in drought characteristics at watershed scale
The projected changes to both short- and long-term drought characteristics differ among simulation pairs due to the uncertainty introduced by the different CGCM3.1 members used to drive CRCM. To quantify uncertainty, the projected changes to the severity, frequency and maximum duration of DRO10 events, for the five simulation pairs, are discussed in this section. Results are very similar for DRO3 and DRO6 events and therefore they not discussed.
Figure 12(a)–(c) illustrates respectively projected changes to the severity, frequency and maximum duration of DRO10 events for the five simulation pairs, at watershed scale, for the 47 Prairie watersheds. Three out of the five pairs, i.e. Figure 12(a) (ii–iv), suggest increases in the severity of DRO10 events for the southern watersheds, while the remaining two pairs (Figure 12(a) (i, v)) suggest both increase/decrease in severity for the various southern watersheds. Similar behaviour is noticed for projected changes to the frequency and maximum duration of DRO10 events (Figure 12(b), (c)).
Figure 12(d) classifies watersheds based on projected changes to severity and frequency of DRO10 events, for the five pairs of simulations. Once again, differences between the different pairs are evident in the subplots. The most vulnerable watersheds, i.e. those with projected increase in both severity and frequency of DRO10 events, clearly lie in the southern Prairies, as can be seen in Figure 12(d) (i–iv). Note that the spread of highly vulnerable watersheds is limited in the fifth pair of simulations shown in Figure 12(d) (v). However, in general, the ensemble mean, as already discussed, identifies the southern Prairies as a hot spot, with high likeliness of severe, more frequent drought events.
5. Discussion and conclusions
This study evaluated the drought characteristics of severity, frequency and maximum duration in the current climate (1971–2000) and projected changes for the future 2041–2070 period as simulated across the Canadian Prairies by the CRCM and SRES A2 scenario. The short- and long-term droughts are defined based on cumulative monthly precipitation anomalies.
Results suggest that the model consistently underestimates the severity and frequency of both short- and long-term drought events over the Canadian Prairies in current climate. The performance of the model is more mixed for long-term drought events, overestimating the observed maximum duration of these events over Saskatchewan and underestimating elsewhere. The CRCM biases in mean seasonal precipitation and the underestimation of interannual variability of seasonal precipitation probably influence the model performance in simulating drought events.
For future climate, the results of the study suggest an increase in severity of both short- and long-term droughts over a large portion of the southern Prairies and along the eastern slopes of the Rocky Mountains, with the largest increase occurring with long-term events. Projected increases in shorter and longer drought frequency will affect the surface and groundwater resources over most of the southern Prairies, respectively. Maximum durations of long-term drought events are also projected to increase across a large portion of the southern Prairies with the largest increase associated with droughts lasting at least 10 months.
In general, this study indicates that the southern Canadian Prairies are projected to experience droughts of even greater severity and duration in the future than those of the 20th century, which could significantly impact environmental and socio-economic conditions in this region that contains more than 80% of Canada's cropland and rangeland (Chipanshi et al., 2006). Prolonged droughts are also likely to become a threat to conservation of biodiversity by damaging the domesticated forage plants for livestock, natural grasslands, savannas, many wetlands, tundra, and certain forb and shrub communities of the Canadian Prairies (Tilman and Haddi, 1992; Chipanshi et al., 2006). Consequently, more efficient adaptation strategies are essential to address the impact of these short- and long-term droughts in future.
The indices used in this study to determine severity, frequency and maximum duration of droughts have some limitations in spite of their usefulness. These indices capture only the effect of precipitation change on the occurrence and severity of droughts, but the severity also depends on other factors, such as evapotranspiration, soil moisture, temperature, etc., which were not taken into account. Therefore, in future, other drought indices will also be explored to study future changes in drought characteristics over the Canadian Prairies. In addition, multivariate copula-based approaches for modelling drought frequency, duration and severity will also be explored to characterize drought features in a multidimensional space. The results presented in this article are based on an ensemble of simulations from one model; in future, to further quantify the uncertainties, multi-model, multi-scenario simulations will need to be considered.
The authors would like to thank the Ouranos Climate Simulations Team for supplying the CRCM generated daily precipitation outputs used in this study. This research was carried out within the Canadian Regional Climate Modelling and Diagnostics (CRCMD) Network funded by the Canadian Foundation for Climate and Atmospheric Sciences (CFCAS), the Ouranos Consortium and the Pacific Climate Impacts Consortium (PCIC). Anonymous review comments are gratefully acknowledged.