sign in sign up
<<
Home , Atmospheric river

FEBRUARY 2019 Y A N G E T A L . 275

Spatiotemporal Changes in Extremes over Canada and Their Teleconnections to Large-Scale Patterns

YANG YANG AND THIAN YEW GAN Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada

XUEZHI TAN Department of Civil and Environmental Engineering, University of Alberta, Edmonton, Alberta, Canada, and Department of Water Resources and Environment, Sun Yat-sen University, Guangzhou, China

(Manuscript received 12 January 2018, in final form 8 November 2018)

ABSTRACT

In the past few decades, there have been more extreme climate events occurring worldwide, including Canada, which has also suffered from many extreme precipitation events. In this paper, trend analysis, probability distribution functions, principal component analysis, and wavelet analysis were used to investigate the spatial and temporal patterns of extreme precipitation events of Canada. Ten extreme precipitation in- dices were calculated using long-term daily precipitation data (1950–2012) from 164 Canadian gauging sta- tions. Several large-scale climate patterns such as El Niño–Southern Oscillation (ENSO), Pacific decadal oscillation (PDO), Pacific–North American (PNA), and North Atlantic Oscillation (NAO) were selected to analyze the relationships between extreme precipitation and climate indices. Convective available potential energy (CAPE), specific humidity, and surface temperature were employed to investigate potential causes of trends in extreme precipitation. The results reveal statistically significant positive trends for most extreme precipitation indices, which means that extreme precipitation of Canada has generally become more severe since the mid-twentieth century. The majority of indices display more increasing trends along the southern border of Canada while decreasing trends dominated the central Canadian Prairies. In addition, strong teleconnections are found between extreme precipitation and climate indices, but the effects of climate patterns differ from region to region. Furthermore, complex interactions of climate patterns with synoptic atmospheric circulations can also affect precipitation variability, and changes to the and extreme precipitation could be explained more by the thermodynamic impact and the combined thermo- dynamic and dynamic effects, respectively. The seasonal CAPE, specific humidity, and temperature are correlated to Canadian extreme precipitation, but the correlations are dependent, which could be positive or negative.

1. Introduction severe hardship to our society and natural systems (Gao et al. 2017; Zhang et al. 2011). Conversely, because In recent decades, hydrologic extremes such as floods of global warming impacts, droughts in semiarid/arid and droughts have caused more public attention be- regions of Africa have worsened in recent years (Gan cause they have been occurring more frequently and in et al. 2016). A brief review of some related studies is greater severity worldwide (Easterling et al. 2000; presented below. Costa and Soares 2009; Wang et al. 2014; Chen et al. Climatic extremes can have devastating impacts on 2015; Elewa et al. 2016; Zilli et al. 2017). As Earth our societies (Hales et al. 2003; Mass et al. 2011). For warms, a higher temperature likely means that more example, Canada has experienced severe floods that precipitation will fall over shorter time intervals, thus resulted in billions of dollars of damage, such as the increasing the frequency and severity of extreme flood events of Calgary and Toronto in 2013 which, as events, which could incur significant damage and the respective worst natural disaster of Alberta and Ontario, are also ranked the first and the third larg- Corresponding author: Thian Yew Gan, [email protected] est natural insured disasters in Canada, respectively

DOI: 10.1175/JHM-D-18-0004.1 Ó 2019 American Meteorological Society. For information regarding reuse of this content and general copyright information, consult the AMS Copyright Policy (www.ametsoc.org/PUBSReuseLicenses). Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 276 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

(Milrad et al. 2015; Wang et al. 2014). In southern Al- scientifically robust measures of the characteristics of berta, Canada, Gizaw and Gan (2016) projected an precipitation and temperature, the two most important overall increase in its future extreme precipitation in daily climate variables (Zhang et al. 2011); for example, the mid- and late twenty-first century, while Kunkel Miao et al. (2015) and Jiang et al. (2017) used ETCCDI et al. (1999) detected increasing trends in extreme indices to study changing behaviors of precipitation precipitation of 1–7-day duration over 1951–93 in extremes in China. Canada. Meanwhile, the increase in the frequency or Past studies on the extreme precipitation in Canada intensity of precipitation extremes has also been ob- have either used probability distributions such as served in southern China (Zhai et al. 2005), Japan the generalized extreme value (GEV) distribution (Duan et al. 2015), Denmark (Gregersen et al. 2013), (Simonovic et al. 2016; Tan and Gan 2017), or they have Sweden (Gregersen et al. 2015), the United States only focused on extreme precipitation of some parts of (Dhakal and Tharu 2018; Huang et al. 2017), and Brazil Canada (Benyahya et al. 2014; Wang et al. 2015). Even (Zilli et al. 2017). though large-scale climate patterns have been tele- Flooding may cause an outbreak of cholera, typhoid, connected to precipitation in some parts of Canada and diarrheal disease because of environmental pollu- (e.g., Gan et al. 2007), their influences on precipitation tion resulting from floodwater mixed with human and extremes have not received much attention and thus animal waste. On the other hand, droughts that reduce require further investigation (Xi et al. 2018; Zhang et al. the amount of water available for sanitation can increase 2001). Using the Pacific decadal oscillation (PDO), the risk of diseases such as malaria and dengue fever Pacific–North American (PNA), and Atlantic Oscilla- (Hales et al. 2003). Other than damage and hardships tion (AO) teleconnection patterns as covariates, Asong caused by floods and famines by droughts, extreme cli- et al. (2016) built a generalized linear model (GLM) to mate events can also have significant impacts on human model seasonal precipitation, temperature, and their health. Li et al. (2018) have shown that human perceived extremes in the Canadian Prairies (CP). Based on the temperature or apparent temperature (AP) has in- Bayesian spatiotemporal quantile (BSTQR) model, creased faster than air temperature over land, and the Tan et al. (2018a) examined effects of large-scale cli- summer increase in AP-based thermal discomfort is mate patterns on Canadian winter precipitation at expected to outpace the winter decrease in AP-based different quantile levels. They found that the tele- thermal discomfort. connections of Canadian winter precipitation to cli- Global warming could increase occurrences of pre- mate patterns are stronger at higher than at medium cipitation extremes (Allan and Soden 2008)since quantiles, implying that large-scale climate pat- according to the Clausius–Clapeyron equation, the terns likely exert stronger influence on precipitation water-holding capacity of the will in- extremes. 2 crease at about 7% 8C 1 in temperature. According to The primary objectives of this study are 1) to analyze the Synthesis Report of the IPCC Fifth Assessment spatiotemporal changes of historical extreme precipi- Report, the surface temperature is projected to in- tation over Canada using 10 extreme precipitation in- crease over the twenty-first century under all represen- dices and 2) to find the influence of large-scale climate tative concentration pathway (RCP) emission scenarios patterns on precipitation extremes of Canada. The re- of IPCC (2013), and extreme precipitation events are sults of this study will provide a more comprehensive projected to become more intensive and frequent in understanding of precipitation extremes and their ob- many regions across the world (IPCC 2014). Located in served changes in Canada. The datasets are described high-latitude areas, warming and extreme precipita- in section 2, the research methodology in section 3, tion in Canada are expected to be more pronounced results and discussion in section 4, and conclusions in (Lemmen and Warren 2004; Bush et al. 2014; Fischer section 5. and Knutti 2016). From 1948 to 2012, air temperature has increased in most parts of Canada, with the largest 2. Datasets warming in winter and , and precipitation has also a. Precipitation time series increased, especially in northern Canada (Jiang et al. 2015; Vincent et al. 2015). Among 464 stations of daily precipitation data given Among various climate indices used for research in in the second generation of the Adjusted Daily Pre- extreme climate, 27 indices developed by the Expert cipitation (ACP2) dataset for Canada (Mekis and Team on Climate Change Detection and Indices Vincent 2011), 164 stations across Canada (Fig. 1) (ETCCDI; http://etccdi.pacificclimate.org/)havebeen that met the following requirements were selected in relatively popular, for they are designed to provide this study. The station should have data over the

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 277

FIG. 1. Study area and meteorological stations. Abbreviations: Alberta (AB), Saskatchewan (SK), Manitoba (MB), Newfoundland and Labrador (NL), Prince Edward Island (PE), Nova Scotia (NS), Northwest Territories (NT), Nunavut (NU), Ontario (ON), New Brunswick (NB), Yukon Territory (YT), British Columbia (BC), Quebec (QC).

1950–2012 period, and it should have no more than c. Climate indices two consecutive years of missing values. ACP2 is We have selected certain large-scale climate pat- part of the Adjusted and Homogenized Canadian terns that have contributed to the precipitation vari- Climate Dataset (AHCCD), and it is the most ho- ability over Canada (Coulibaly 2006), such as the North mogeneous long-term observed daily precipitation Atlantic Oscillation (NAO), PNA, North Pacific Gyre data currently available for Canada (Tan and Gan Oscillation (NPGO), PDO, and El Niño–Southern 2017). Among the 164 stations selected, only a few Oscillation (ENSO). ENSO likely has the most signif- stations are located in northern Canada. More de- icant interannual climate variability and influence on tails about the adjustments and quality control of this dataset have been extensively discussed by Mekis the climate of Northern Hemisphere (Rasmusson and and Vincent (2011). Wallace 1983), including Africa (e.g., Ntale and Gan 2004). The Multivariate ENSO Index (MEI) is selected b. Extreme precipitation indices to represent ENSO. Out of 27 indices recommended by ETCCDI, we have The PDO is the leading principal component (PC1) of adopted 10 extreme indices grouped into two types to North Pacific (poleward of 208N) monthly sea surface analyze extreme precipitation data of Canada and make temperature anomalies since 1900 (Mantua and Hare the results comparable internationally (Fu et al. 2015; 2002). The warm and cold phases of PDO have an Lovino et al. 2018; Wang et al. 2013): 1) precipitation influence on the precipitation of North America. For amount or intensity, which includes PRCPTOT, SDII, example, winter precipitation in the western United R95p, R99p, Rx1day, and Rx5day, and 2) number of States (Brown and Comrie 2004) and western Canada days exceeding certain thresholds in rainfall depth, has been found to be affected by the PDO (Gan et al. which are CDD, CWD, R10mm, and R20mm. The 2007). The PNA depicts a quadripole of 500-hPa geo- definitions and units of these indices are described in potential height anomalies, with opposite anomalies Table 1. centered over Hawaii and central Canada, and with

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 278 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

TABLE 1. Definitions of extreme precipitation indices.

Index Definition Unit Rx1day Monthly maximum 1-day precipitation mm Rx5day Monthly maximum consecutive 5-day precipitation mm R95p Annual total precipitation when daily precipitation . 95th percentile mm R99p Annual total precipitation when daily precipitation . 99th percentile mm PRCPTOT Annual total precipitation in wet days (daily precipitation $ 1 mm) mm 2 SDII Simple precipitation intensity index mm day 1 R10mm Annual count of days when daily precipitation $ 10 mm day R20mm Annual count of days when daily precipitation $ 20 mm day CDD Max number of consecutive days with daily precipitation , 1 mm day CWD Max number of consecutive days with daily precipitation $ 1 mm day similar signals south of the Aleutian Islands and severe convective and events (Brooks over the southeastern United States (https://ncdc. et al. 2007; Dong et al. 2018; Kishtawal et al. 2010; noaa.gov/teleconnections/pna/). It is one of the most Monkam 2002; Seeley and Romps 2015). We used significant modes of low-frequency variability in the monthly CAPE data of version 2 of the Twentieth extratropics of the Northern Hemisphere throughout Century Reanalysis (20CR) of the National Oceanic the year except June and July. and Atmospheric Administration (https://www.esrl.noaa. The NAO represents the climate variability from the gov/psd/data/gridded/data.20thC_ReanV2.html), which East Coast of the United States to Siberia and from the is an international project aimed at producing a high- Arctic to the subtropical Atlantic (Hurrell et al. 2001). quality global atmospheric circulation dataset (Compo It is characterized by changes in surface pressure, and et al. 2011).Thedatasetcoversfrom1871to2012and it is one of the dominant and prevailing modes of at- is available at 28 spatial resolution. Comparisons with mospheric behavior in the North Atlantic (Hurrell et al. satellite data and other reanalysis data show that 2001; Coulibaly 2006). The NPGO is the second domi- 20CR is generally of high quality (Compo et al. 2011). nant mode of sea surface height (SSH) variability in the Using the CAPE data from 20CR, Krichak et al. northeast Pacific correlated with fluctuations of salinity, (2015) found that heavy precipitating events are as- nutrients, and chlorophyll in the California Current and sociated with an intense intrusion of humid tropical air Gulf of Alaska. It indicates changes in the intensity of and the presence of high CAPE values in the Medi- central and eastern North Pacific gyre circulations, and terranean region. Further, given that precipitation is it is driven by upwelling and horizontal advection (Di also related to air temperature and humidity, monthly Lorenzo et al. 2008). temperature and specific humidity were also analyzed The MEI is the first unrotated principal component in this study. of six atmosphere–ocean variables over the tropical Pacific: sea level pressure, zonal and meridional com- ponents of the surface wind, sea surface temperature, 3. Research methodology surface air temperature, and total cloudiness fraction a. Trend analysis of the sky over the tropical Pacific. It gives a more comprehensive description of ENSO events than the The nonparametric Mann–Kendall (MK) trend test traditional ENSO indices, the Niño-3 and Southern (Mann 1945; Kendall 1955) recommended by the World Oscillation index (SOI; Wolter and Timlin 1993, Meteorological Organization (WMO) was used for the 1998, 2011). trend analysis of the precipitation data. The null hy- pothesis H is that the data are independent and ran- d. CAPE and specific humidity 0 domly distributed, while the alternative hypothesis H1 Convective available potential energy (CAPE), a is that a monotonic trend exists (Song et al. 2015). The proxy for conditions amenable for the occur- MK method has been widely used in the trend analysis rence of extreme precipitation events, is the vertical of hydrologic and climate data (Shadmani et al. 2012; integral of parcel buoyancy between the level of free Frazier and Giambelluca 2016; Pedron et al. 2017). convection and the level of neutral buoyancy (Ye et al. However, the presence of autocorrelation in a time se- 2 1998). CAPE (J kg 1) has been widely used to measure ries can affect the detection of trends in the time series. the onset of convection given that high CAPE values Hamed and Rao (1998) proposed subtracting a non- represent favorable conditions for the occurrence of parametric trend estimator from the original time series

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 279 to account for the autocorrelation, and a detailed cal- and Compo 1998). Wavelet analysis has been exten- culation procedure was given by Daufresne et al. (2009). sively used in climate research (Gan et al. 2007; Mwale The modified MK test was then applied to the time se- et al. 2009; Jiang et al. 2014; Okonkwo 2014). Herein we ries of all indices listed in Table 1, and trend magnitudes chose the Morlet wavelet as the mother wavelet because were estimated using the nonparametric Sen’s slope it finds a delicate balance between time and frequency estimator (Sen 1968). localizations (Grinsted et al. 2004). It should be noted that since we are dealing with finite-length time series, b. Field significance and false discovery rate zeros are padded at the ends of the time series to reduce Statistically, an individual test performed at a given the edge effects (Torrence and Compo 1998), and so the significance level a has an a chance of falsely rejecting wavelet power beyond the cone of influence (COI) the null hypothesis. When conducting multiple tests si- should be explained with caution (Gan et al. 2007). multaneously on data that are spatially correlated, it is necessary to adjust a to avoid falsely rejecting a large number of the null hypotheses (Ventura et al. 2004). In 4. Results and discussion view of a large number of stations selected in the study, a. Changes in extreme precipitation amount/intensity it is necessary to consider the field significance statisti- cally in relation to multiple tests. Here the false dis- To present overall changes to extreme precipitation of covery rate (FDR) method which controls the expected Canada in 1950–2012, Fig. 2 shows the average annual percentage of falsely rejected null hypotheses is em- time series of Canada for all indices, while Table 2 shows ployed to identify the significance of tests conducted in the number of individual stations detected with positive this study (Benjamini and Hochberg 1995). We applied or negative trends for the indices analyzed in this study. the FDR method by using the ‘‘p.adjust’’ function in the Overall, in these indices, positive trends dominate over R language (R Core Team 2017). negative trends, which means that extreme precipitation over Canada shows more increasing than decreasing c. Principal component analysis trends in the study period. Principal component analysis (PCA) is a statistical Over 1950–2012, both the average time series of method to reduce the dimensionality of a multivariate RX1day (Fig. 2a) and RX5day (Fig. 2b) of Canada show time series to several orthogonal principal components statistically significant increasing trends at 0.34 and 21 (PCs) that explain a large percentage of the variability in 0.90 mm decade , respectively. Figures 2c and 2d show the time series (Jolliffe 2002). For the R95p and R99p that the average time series of R95p and R99p also show 21 indices analyzed in this study, the first two leading PCs significant positive trends at 3.13 and 1.52 mm decade , account for more than 30% of the total variance. Fol- respectively. However, for individual stations, R99p 2 lowing Cioffi et al. (2015) and for brevity, the leading trends ranged from 28.77 to 21.06 mm decade 1 (Table 2). PC1s are used for subsequent analysis of the time series As for the average time series of PRCPTOT, the trend 2 of R95p and R99p. of 9.44 mm decade 1 is statistically significant, while the average SDII time series of Canada has a nonsignificant d. Wavelet analysis negative trend. The wavelet transform is an effective tool designed to There are more stations showing positive than nega- transform a time series into time and frequency domains tive trends for RX1day and RX5day (respectively 102 simultaneously, revealing temporal and frequency and 107 vs 60 and 55), but only about 10% of positive changes of the dominant oscillations of the time series trends are statistically significant, which agrees with (Torrence and Compo 1998). Compared to the tradi- Shephard et al. (2014), who also reported a general lack tional Fourier transform, the wavelet transform is well of significant trend signals. By contrast, no station has known for its ability to analyze nonstationary time series shown a significant negative trend in either index. A (Cazelles et al. 2008), which is particularly useful for comparable number of stations show positive and neg- climate series that exhibit nonstationary behaviors ative trends in R95p and R99p (respectively 108 and (Chang et al. 2015; Tan and Gan 2017). 103 vs 54 and 60), but for R99p only two stations show Another important application of the wavelet trans- statistically significant positive trends. Additionally, a form is the wavelet transform coherence (WTC) be- comparable number of stations show positive and neg- tween two time series, defined as the square of their ative trends in SDII (88 vs 76), but more positive than cross-spectrum normalized by the power spectrum of negative trends are statistically significant (20 vs 9). each time series, which gives us their cross-correlation Spatial patterns of these results are presented in Fig. 3. (between 0 and 1) as a function of frequency (Torrence Generally, stations with a positive extreme precipitation

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 280 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

FIG. 2. Annual series of extreme precipitation indices from 1950 to 2012. The red line is the linear trend, the blue dashed line is the mean, and S is the trend per decade by Sen’s slope.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 281

TABLE 2. Trends per decade and number of stations showing positive or negative trends of extreme precipitation indices during 1950– 2012. Annual trends that are statistically significant at the 0.05 level are in bold.

Regional trends Positive Significant Negative Significant 2 Index (unit decade 1) Range trend positive trend trend negative trend Rx1day 0.34 23.19 to 3.46 102 9 60 0 Rx5day 0.90 23.93 to 8.95 107 10 55 0 R95p 3.13 238.25 to 50.91 108 10 54 3 R99p 1.52 28.77 to 21.06 103 2 60 0 PRCPTOT 9.44 265.92 to 89.47 117 32 47 3 SDII 20.005 20.54 to 0.34 88 20 76 9 R10mm 0.27 22.67 to 2.63 107 27 56 9 R20mm 0.08 21.71 to 1.32 101 10 63 7 CDD 20.33 25.00 to 2.08 62 0 101 3 CWD 0.08 20.77 to 1.25 114 16 50 2

amount/intensity trend were located along the southern discussed in section 4a. However, CDD displays a dif- border of Canada while trends that are more negative ferent spatial distribution: negative trends dominated in are found in the central CP: British Columbia (BC), the north and a mixed pattern in the south, which again Alberta (AB), Saskatchewan (SK), and Manitoba (MB). shows that Canada had generally become wetter since Meanwhile, positive trends dominate the northern part the 1950s, especially in the north. of Canada. We have detected decreasing trends in the annual SDII and CDD over Canada and more increasing trends b. Changes in number of days with extreme in days with heavy precipitation (R10mm) in 1950–2003, precipitation similar to the results of Vincent and Mekis (2006). One About twice as many stations show positive than consequence of this change could be an increased fre- negative trends in R10mm and R20mm (number of quency and severity of flash floods (Limsakul and 2 days $ 10 and 20 mm day 1 of precipitation, respec- Singhruck 2016). From analyzing changes in global pre- tively), and more positive trends are statistically signif- cipitation extremes using ETCCDI indices, Alexander icant especially for R10mm (27 vs 9). Therefore, out of et al. (2006) also found a widespread and significant in- 164 stations of R10mm data of Canada analyzed, the crease in precipitation extremes. Wang et al. (2015) 2 average trend estimated was about 0.27 days decade 1. detected decreasing heavy rainfall intensities in eastern For R20mm, a comparable number of negative and and southern Ontario, which is similar to the decreasing positive trends are significant, which over 1950–2012 SDII of southern Ontario in Fig. 3f. However, using a 2 range from 21.71 to 1.32 days decade 1. It is noted that regional climate model, Wang et al. (2014) projected more stations show negative trends in consecutive dry both the intensity and frequency of extreme rainfall of days (CDD) than in consecutive wet days (CWD) (101 Ontario would likely increase in the future. Even though vs 50), but the reverse number of positive trends be- hydrologic extremes are generally expected to become tween CDD and CWD (62 vs 114). However, most of more severe in North America (Kunkel 2003), locally the detected trends are not statistically significant ex- possible changes to future extreme precipitation remain cept for positive trends in CWD. The trend magnitude uncertain because of many possible factors involved, 2 for the average CDD time series (20.33 days decade 1) such as moisture availability, thermodynamic instability, is much higher than that of the average CWD time series effects of large-scale atmospheric circulations, terrain 2 (0.08 days decade 1). The overall results obtained for features, and others. the above 10 indices show that from 1950 to 2012, Spatially, these extreme precipitation indices exhibit Canada has generally become wetter, which is what we a mixture of increasing and decreasing trends across would expect from increasing atmospheric moisture as Canada in 1950–2012, which is consistent with the rela- the global climate has become warmer since the mid- tively low spatial coherence of extreme precipitation twentieth century. shown by Vincent and Mekis (2006). However, other Figure 4 shows spatial distributions of trends for than CDD, all indices exhibit more increasing than de- extreme precipitation indices, R10mm, R20mm, CWD, creasing trends in southern Canada, except in the central and CDD across Canada, of which the first three indices CP, where decreasing trends are more dominant. Zhang display analogous spatial distributions with the indices et al. (2001) found that although the temporal distribu- Rx1day, Rx5day, R95p, R99p, SDII, and PRCPTOP tion of the number of heavy precipitation events in

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 282 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

FIG. 3. Spatial patterns of trends for extreme precipitation amount/intensity indices from 1950 to 2012. The abbreviations are P, positive trend; SP, significant positive trend; N, negative trend; SN, significant negative trend; and NT, no trend.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 283

FIG. 4. Spatial patterns of trends for extreme precipitation day indices from 1950 to 2012. The abbreviations are as in Fig. 3.

Canada is spatially coherent, it varies considerably be- annual precipitation, which is expected because CDD tween and regions. Tan and Gan (2017), who represents the number of consecutive dry days. The re- analyzed the nonstationarity of heavy precipitation sult demonstrates that these 10 extreme precipitation events of Canada, concluded that stations with in- indices can adequately reflect changes in the annual creasing trends are mostly located in the southwest and precipitation of Canada. in Quebec while more decreasing trends are detected d. Probability distribution functions in the CP. To investigate the temporal variation in extreme pre- c. Relationship between extreme precipitation indices cipitation further, all indices were divided into three 20-yr and annual precipitation subperiods: 1950–70, 1971–91, and 1992–2012 (Fig. 5). To examine if extreme precipitation indices are good Based on results obtained from the Kolmogorov– indicators of the annual precipitation, their relationships Smirnov test, it is noted that distributions of RX5day, are estimated using the Spearman’s rank correlation PRCPTOT, R10mm, CDD, and CWD of subperiod 1 rho (Spearman 1904). As Table 3 shows, all indices are are statistically different from that of subperiod 3, significantly correlated to the annual precipitation, es- suggesting a shift in these extreme precipitation indices. pecially for R95p, PRCPTOT, R10mm, and R20mm, Generally, all indices except CDD tend to shift to the whose Spearman rank correlation coefficients exceeded right from subperiods 1 to 2 and to 3, which demon- 0.80. CDD is the only index negatively correlated to the strates that probabilistically, extreme precipitation of

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 284 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

TABLE 3. Correlation coefficients between extreme precipitation indices and annual precipitation. AP stands for annual precipitation. Coefficients that are statistically significant at the 0.05 level are in bold.

Rx1day Rx5day R95p R99p PRCPTOT SDII R10mm R20mm CDD CWD AP 0.571 0.657 0.847 0.723 0.989 0.583 0.946 0.832 20.448 0.398

Canada has been increasing in severity and frequency the 2000s, respectively. A small band of 1 year in the since the mid-twentieth century. For example, R10mm 1980s, two bands of 2–4-yr and 6–8-yr periodicities in was less than 32 days in 1950–70, but it exceeded 32 days 2000s are found in eastern R95p and R99p power in 1971–91 and 1992–2012. spectrum plots, respectively. Moreover, from probability distribution functions In investigating heavy precipitation of Canada for (PDFs) of RX1day, RX5day, R99p, PRCPTOT, and 1900–98, Zhang et al. (2001) also found decadal oscil- R10mm, it seems that the probability of occurrence of lation as a dominant feature in precipitation extremes, moderate extreme precipitation events has decreased while wavelet analysis of R95p and R99p indices in Fig. 6 since the 1950s. For example, the probability of getting show a combination of significant interannual and de- RX1day at 46 mm was about 35% in 1950–70, but the cadal oscillations that appeared and disappeared over probability decreased to just over 10% in 1992–2012. In 1950–2012 in eastern, central, and western Canada contrast, the probability of getting moderate CDD has without any consistent pattern. increased in the second and third subperiods, while 2) WAVELET TRANSFORM COHERENCE CWD has shifted so that the duration of CWD tends to increase compared to the past. Overall, it seems that The WTC plots between extreme precipitation indices extreme precipitation of Canada will tend to occur more (PC1 of R95p and R99p) and selected climate indices frequently in the future, probably leading to more are shown in Figs. 7–9. The arrows indicate the phase flooding events. difference: arrows pointing to the right (left) mean that two time series are in phase (antiphase) while arrows e. Relationship between R95p/R99p and climate pointing up (down) mean that one time series leads indices (lags) the other by 908. Analogous to Fig. 6, WTC be- tween R95p and climate indices are similar to WTC 1) WAVELET ANALYSIS OF R95P AND R99P between R99p and climate anomalies in the same re- Given the vast landmass and various climatic zones gion, and so we only present results of R99p. The co- of Canada, we have divided Canada into three regions, herence spectrum plots between the MEI and R99p eastern, central, and western Canada. A similar regionali- show statistically significant power at 2–4-yr bands in zation approach has been adopted to analyze synoptic some years over western Canada (Fig. 7a) and at 2–6-yr circulation patterns related to heavy precipitation (Tan bands over eastern Canada, respectively. After 1970s, and Gan 2017) and seasonal precipitation (Coulibaly there was a consistently strong coherence at 8–16-yr 2006) in Canada. Figure 6 shows the wavelet power bands between R99p and MEI in western and eastern spectra of PC1 for R95p and R99p of western, central, Canada, although parts of the bands are outside the and eastern Canada. The thick black contours represent COI. For central Canada, the power of the wavelet co- statistically significant power at the 95% significance herence between MEI and R99p (Fig. 7c) was relatively level against red noise, and the white sag line is the COI, weak, which is expected because of the blocking effect of outside of which results may be affected by edge effects the Canadian Rockies. of zero paddings. There was a strong 8–12-yr wavelet coherence since In general, R95p and R99p showed somewhat similar the 1980s between PDO and R99p in western Canada oscillation patterns in the same regions. However, each (Fig. 7b), but a weak coherence between PDO and R99p region exhibited different oscillation patterns for R95p in eastern and central Canada. There was a high 2–8-yr and R99p. As shown in Figs. 6a and 6b, the power coherence between the PNA and R99p from the 1950s spectrum plot of western R95p reveals two distinct to 1990s in western Canada (Fig. 8a) and a strong bands at 4–8-yr and 2–4-yr periodicities from 1960 to 8–14-yr coherence between the PNA and R99p during 1975 and in the 1980s, respectively. Two bands of 2–6-yr the 1950s and after 1995 in eastern Canada (Fig. 8e). For and 12–16-yr periodicities are found in the 1980s for central Canada, other than a strong 1–4-yr coherence the power spectrum plot of western R99p. In contrast, that appeared since the 1990s, the PNA did not seem an 8–16-yr band and a 1–4-yr band dominate the central to have much effect on R99p. Figure 8b shows that there R95p and R99p power spectrum plots in 1950–70 and is scattered strong interannual (1–4-yr) coherence

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 285

FIG. 5. Annual probability density functions for extreme precipitation indices from 1950 to 2012 for three time periods: 1950–70, 1971–91, and 1992–2012.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 286 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

FIG. 6. Wavelet power spectra of PC1 for R95p and R99p in eastern, central, and western Canada. between the NAO and R99p in western Canada and GEV distribution, Tan and Gan (2017) found a strong a significant coherence band of 6–8-yr periodicity in nonstationary relationship between heavy precipitation central Canada. The band of high 12–16-yr coherence and large-scale climate patterns. For example, annual mainly occurred after the 1990s, but it is outside the COI maximum daily precipitation in southwestern coastal of the wavelet coherence plot (Fig. 8f). The wavelet regions and the southern CP tend to be larger in El Niño coherence between the NPGO and R99p was strong in than in La Niña years. In our study, results of the WTC western Canada, of interannual 2–8-yr cycle and 8–16-yr analysis further show that extreme precipitation of cycle for R99p (Fig. 9a). Some scattered coherence be- western Canada based on R95p/R99p indices are tween the NPGO and R99p of 4–6-yr cycle can be found strongly correlated with the MEI, PDO, PNA, and in central Canada, and a strong 8–16-yr coherence from NPGO. For extreme precipitation of central Canada, the 1990s is detected in eastern Canada (Fig. 9c). the NAO and NPGO exerted more influence than other From a nonstationary analysis of the frequency and climate anomalies, while for eastern Canada, extreme intensity of heavy precipitation over Canada using a precipitation primarily comes under the influence of the

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 287

FIG. 7. Wavelet coherence between the MEI/PDO and R99p for eastern, central, and western Canada.

MEI, PNA, and NPGO. Jiang et al. (2014) also found Singhruck 2016; Xi et al. 2018). Using storm back- the seasonal precipitation of Alberta (central Canada) trajectory analyses, Tan et al. (2018b) showed that ex- to be strongly influenced by ENSO, PDO, and NPGO, treme precipitation events in southwestern Canada are while Gan et al. (2007) showed the influence of ENSO generally associated with the atmospheric river over the and PDO on the winter precipitation of southwestern North Pacific, while moisture pathways for central and Canada. eastern Canada follow the westerlies in the midlatitudes coming from the North Pacific Ocean or the northern 3) RELATIONS OF ATMOSPHERIC CIRCULATIONS polar over high-latitude regions. WITH CLIMATE PATTERNS AND ANOMALOUS Using boosted regression tree analysis, Theobald et al. PRECIPITATION (2018) demonstrated that teleconnections do not act in Oceanic and atmospheric circulation patterns are isolation, and their complex interactions with synoptic potential drivers behind climate extremes worldwide atmospheric circulation can affect precipitation vari- through their impacts on sea surface temperatures, ability. Furthermore, Tan et al. (2018c) applied the self- surface winds, and sea level pressure (Limsakul and organizing map (SOM) algorithm on vertically integrated

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 288 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

FIG.8.AsinFig. 7, but for the PNA and NAO. transport (IVT) data to analyze large-scale associated with the NAO in western Canada is larger meteorological patterns (LSMPs) associated with pre- than that in eastern Canada (Tan et al. 2018c), which cipitation extremes for summer and fall in Canada. again agrees with the more extensive wavelet coherence During summer, the occurrence of LSMPs associated of NAO–R99p in Fig. 8b than that in Fig. 8f.Addi- with ENSO in western Canada is greater than that in tionally, LSMPs patterns associated with less frequent eastern Canada (Tan et al. 2018c), as shown by more summer extreme precipitation events over western extensive statistically significant wavelet coherence be- Canada tend to occur during the negative phase of the tween the MEI and R99p in Fig. 7a than that in Fig. 7e. NAO, which has also been shown by the coherence In addition, Tan et al. (2018c) further show that LSMPs phase in Fig. 8b, indicating that the NAO also has sig- associated with more frequent occurrence of extreme nificant influence on certain LSMPs patterns they have precipitation over western Canada are more likely to identified. occur during positive phases of ENSO, as shown by the Furthermore, it has been shown that regional positive coherence phase in Fig. 7a, implying that ENSO has large (negative) precipitation anomalies are generally asso- influence on certain LSMP patterns identified by Tan ciated with midtropospheric convergence (divergence), et al. (2018c). Similarly, the occurrence of LSMPs located to the right (left) of the ridge axis and left (right)

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 289

TABLE 4. Months from January (1) to December (12) are ranked in terms of the number of extreme precipitation events that oc- curred in each month.

Province Month West BC 1 11 12 YK 6 7 8 Central AB 6 7 8 MB 7 6 8 NU 8 9 7 SK 6 7 8 NT 7 8 6 East NB 1 11 2 NL 1 3 2 NS 1 12 2 ON 9 7 6 PE 2 8 12 QC 8 9 7

precipitating. In comparison, anomalously low IVT values are associated with more frequent extreme pre- cipitation events, partly caused by an extremely low ground surface temperature that facilitates the moisture flux to precipitate (Tan et al. 2018c). f. Correlation between extreme precipitation and seasonal CAPE, specific humidity, and temperature Table 4 shows that extreme precipitation tends to occur more frequently in summer and winter for eastern Canada, in summer for central Canada, and in summer and winter for western Canada. To investigate the pos- sible impacts of CAPE, specific humidity, and temper- ature on extreme precipitation of Canada, Fig. 10 shows the spatial distribution of trends in seasonal CAPE, specific humidity, and surface temperature, respectively. In summer, both CAPE and specific humidity show increasing trends in central and eastern Canada, but decreasing trends along the southern border and in parts of northern Canada in 1950–2012. For seasonal surface temperature, negative trends were widespread in cen- tral Canada while positive trends dominated northern and southern Canada. In winter, there had been more increasing trends in CAPE in central and southeastern FIG.9.AsinFig. 7, but for the NPGO. Canada, and more increasing trends in specific humidity over central and western Canada. However, widespread of the trough. Extremely large IVT values over western increasing trends in the winter surface temperature were Canada are related to the Aleutian low and Gulf cy- detected across Canada except for the easternmost part. clone, which force moisture from the North Pacific to Using the Spearman rank correlation, the influence of western Canada; however, it is not necessarily associ- CAPE, specific humidity, and surface temperature on ated with positive precipitation anomalies or frequent extreme precipitation (Rx1day) is further investigated extreme precipitation events, because intensive mois- (Fig. 11). Given that CAPE values are much higher in ture fluxes sometimes just pass over a region without summer than in other seasons, the correlation analysis is

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 290 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

FIG. 10. Spatial distribution of the seasonal CAPE, specific humidity, and surface temperature trend from 1985 to 2012. The abbreviations are as in Fig. 3. only conducted for the summer. Figures 11a and 11b increasing trends of extreme precipitation detected in show that summer CAPE and specific humidity values this region are at least partly attributed to increasing are positively correlated with extreme precipitation in CAPE and specific humidity in the summer. Our results southern Canada, respectively. The annual cycle of agree with Lepore et al. (2015), who investigated the CAPE generally reaches its maximum during summer in dependency of rainfall extremes on temperature and the Northern Hemisphere (Riemann-Campe et al. 2009), CAPE in the United States. They found that rainfall and summer precipitation is often of convective origin, intensity quantiles are related to CAPE by a power-law for which latent heat release is the primary source of relationship. In Turkey, Lolis and Türkes¸ (2016) also energy driving the upward air motions (Guichard et al. found that the occurrence of extreme precipitation events 2004; Lenderink and Van Meijgaard 2008). Apparently, in summer is linked to low upper-air temperatures and

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 291

FIG. 11. Spatial distribution of the Spearman rank correlation between extreme precipitation and (a) summer CAPE, (b) summer specific humidity, (c) winter specific humidity, (d) summer temperature, and (e) winter temperature. high static instability (through CAPE) associated with d’Orgeville et al. (2014) projected the future moisture the summer heating of land and upper-air disturbances. availability and the rainfall of the Great Lakes basin to Similarly, Murugavel et al. (2012) also found that in- increase because of climate change. Besides, extreme creasing CAPE over India compensates the weakening rainfall is projected to increase more than the mean of circulation and is responsible for the in- annual rainfall because extreme rainfall is projected to crease in the frequency of extreme events over the 1984– originate from the midtroposphere, where warming is 2008 period. From simulating the future regional climate projected to be higher than surface warming. of the Great Lakes using a regional climate model called For the CP, negative correlations between extreme the Weather Research and Forecasting (WRF) Model, precipitation and specific humidity (Fig. 11c), and

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 292 JOURNAL OF HYDROMETEOROLOGY VOLUME 20 between extreme precipitation and temperature (Fig. 11e), changes projected in six climate model experiments. have been detected in winter. This means that winter Their climate model results consistently show that extreme precipitation has decreased even though both over the subtropics, the thermodynamic change for specific humidity and surface temperature have in- extreme precipitation is an overall increase as a result creased under climate warming. Zhang et al. (2001) of increased atmospheric moisture, while for mean detected both positive and negative trends in summer precipitation the thermodynamic change is small or it heavy rainfall, but negative trends in winter heavy decreases. Tandon et al. (2018) also noted that large- snowfall in central CP, respectively. Vincent and Mekis scale upward motion of air during extreme precipitation (2006) confirmed that annual total snowfall in southern events could be related to changes in vertical stability in Canada has decreased significantly in the second half of subtropical areas, or changes in the seasonal mean cir- the twentieth century, and annual maximum depth culation near the equator. Therefore, while changes to in Canada has also decreased (Kunkel et al. 2016; the summer extreme precipitation could be attributed to Vincent et al. 2015), which is likely related to climate the thermodynamic impact, winter extreme precipita- warming impacts. Wang and Zhang (2008) used statis- tion changes are more complicated and are likely at- tically downscaled principal components of sea level tributed to the combined thermodynamic and dynamic pressure and specific humidity as covariates to a GEV to effects (Ban et al. 2018; Tandon et al. 2018; Zhou derive winter maximum daily precipitation over North et al. 2018). America. From projected changes in covariates ob- tained from climate change simulations of the CCCma 5. Summary and conclusions Coupled Global Climate Model, version 3.1 (CGCM3.1) forced by the IPCC Special Report Emissions Scenario Because of effects of climate warming, hydrologic A2, they projected the maximum daily precipitation extremes such as droughts and floods have occurred over the CP to decrease. This is because higher humidity more frequently and in greater severity globally in re- values in the CP are associated with smaller GEV lo- cent decades, resulting in severe environmental and cation parameters, which means that higher humidity societal impacts (Beniston and Stephenson 2004). This is associated with large-scale circulations unfavorable study analyzed monotonic trends of 10 extreme pre- for precipitation in the CP, where circulations exert a cipitation indices and possible relationships between stronger influence over humidity. Additionally, climate R95p/R99p and several climate patterns over Canada. models projected a smaller decrease in the intensity of Statistically significant trends have been detected for all the daily snowfall extreme than in the mean snowfall indices except for SDII and R20mm. Except for CDD, over many terrestrial regions of Northern Hemisphere most indices analyzed exhibit positive trends, which due to global warming (O’Gorman 2014). This could be demonstrate that increasing extreme precipitation related to the theory of –snow phase transition that events have occurred in Canada over 1950–2012, con- snowfall extremes occur in a range near the optimal stituting a growing risk for urban settlements and temperature that is insensitive to warming (Dai 2008). infrastructure losses (Lovino et al. 2018). In addition, These together may contribute to the negative correla- probability distribution functions of these indices plot- tion with specific humidity and the general decline of ted over three subsequent subperiods further confirm extreme precipitation in the central CP. this conclusion. Spatially, even though these extreme From simulations of the Canadian Regional Climate precipitation indices exhibit a mixture of increasing and Model, Mladjic et al. (2011) projected that 1–7-day decreasing trends, positive trends dominate over nega- precipitation extremes of 20-, 50-, and 100-yr return tive trends across the southern and northern part of levels will increase over most of Canada, especially for Canada, except for CDD. While in the central CP, more northern regions, which would have severe implications decreasing trends were detected. for managing water resources of Canada such as sewer From wavelet analysis and wavelet transform co- systems, flood control, and water storage systems. herence plots, extreme precipitation of western Canada However, precipitation extremes of some regions in was strongly teleconnected to the MEI, PDO, PNA, and southern Canada (mainly the central CP) are projected NPGO, that of central Canada was highly associated to decrease. Results of Mladjic et al. (2011) and Zhou with the NAO and NPGO, and for eastern Canada by et al. (2018) generally agree with our results based on the MEI, PNA, and NPGO. However, teleconnections historical precipitation data of Canada. In addition, do not act in isolation, and their complex interactions Emori and Brown (2005) separated the dynamic (atmo- with synoptic atmospheric circulation can affect pre- spheric motion) change and atmospheric moisture con- cipitation variability. For example, LSMPs related to tent (thermodynamic) change of extreme precipitation more frequent extreme precipitation in western Canada

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 293 tend to occur during the positive phase of the PNA, models. Climate Dyn., 47, 2901–2921, https://doi.org/10.1007/ ENSO, and the PDO. s00382-016-3004-z. ä Detected increasing trends in extreme precipitation Ban, N., J. Rajczak, J. Schmidli, and C. Sch r, 2018: Analysis of Alpine precipitation extremes using generalized extreme in southern Canada are related to the increasing CAPE value theory in convection-resolving climate simulations. Cli- and specific humidity in summer, while decreasing mate Dyn., https://doi.org/10.1007/s00382-018-4339-4. extreme precipitation over central CP is associated Beniston, M., and D. B. Stephenson, 2004: Extreme climatic with the decreasing snowfall and the influence of circu- events and their evolution under changing climatic condi- lation in winter. It is noted that while changes to the tions. Global Planet. Change, 44,1–9,https://doi.org/10.1016/ j.gloplacha.2004.06.001. summer extreme precipitation could be explained by the Benjamini, Y., and Y. Hochberg, 1995: Controlling the false dis- thermodynamic impact, winter extreme precipitation covery rate: A practical and powerful approach to multiple changes are more complicated and are likely attributed testing. J. Roy. Stat. Soc., 57B, 289–300, https://www.jstor.org/ to the combined thermodynamic and dynamic effects. stable/2346101. In addition to the influence of large-scale climate Benyahya, L., P. Gachon, A. St-Hilaire, and R. Laprise, 2014: Frequency analysis of seasonal extreme precipitation in patterns, land reclamation, urbanization, and other southern Quebec (Canada): An evaluation of regional climate human activities can also cause changes in climate model simulation with respect to two gridded datasets. Hy- extremes (Xi et al. 2018). For example, as terrestrial drol. Res., 45, 115–133, https://doi.org/10.2166/nh.2013.066. evapotranspiration is the major moisture source for Brooks, H. E., A. R. Anderson, K. Riemann, I. Ebbers, and many extreme precipitation events in Canada (Tan H. Flachs, 2007: Climatological aspects of convective param- eters from the NCAR/NCEP reanalysis. Atmos. Res., 83, 294– et al. 2018b), urbanization and reclamation that change 305, https://doi.org/10.1016/j.atmosres.2005.08.005. the land surface could also have impacts on extreme Brown, D. P., and A. C. Comrie, 2004: A winter precipitation precipitation events. Therefore, a comprehensive ‘‘dipole’’ in the western United States associated with multi- analysis of the influence of climatic and nonclimatic decadal ENSO variability. Geophys. Res. Lett., 31, L09203, factors on precipitation extremes will require further https://doi.org/10.1029/2003GL018726. research. Bush, E., J. Loder, T. James, L. Mortsch, and S. Cohen, 2014: An overview of Canada’s changing climate. Canada in a changing climate: Sector perspectives on impacts and adaptation, Acknowledgments. This project was partly funded by F. J. Warren and D. S. Lemmen, Eds., Government of Canada the Floodnet, a Strategic Network of Natural Science Rep., 292 pp., https://www.nrcan.gc.ca/environment/resources/ and Engineering Research Council (NSERC) of Canada. publications/impacts-adaptation/reports/assessments/2014/16309. The first author was also partly funded by the Chinese Cazelles, B., M. Chavez, D. Berteaux, F. Menard, J. O. Vik, S. Jenouvrier, and N. C. Stenseth, 2008: Wavelet analysis of Scholarship Council of China and by the University of ecological time series. Oecologia, 156, 287–304, https://doi.org/ Alberta. CAPE, specific humidity, and surface temper- 10.1007/s00442-008-0993-2. ature data were taken from the 20th Century Reanalysis Chang, N.-B., M. V. Vasquez, C.-F. Chen, S. Imen, and L. Mullon, (V2) of National Oceanic and Atmospheric Administration 2015: Global nonlinear and nonstationary climate change (https://www.esrl.noaa.gov/psd/data/gridded/data.20thC_ effects on regional precipitation and forest phenology in Panama, Central America. Hydrol. Processes, 29, 339–355, ReanV2.html). PNA and NAO were collected from the https://doi.org/10.1002/hyp.10151. NOAA Climate Prediction Center (CPC), PDO from Chen, L., V. P. Singh, S. Guo, J. Zhou, J. Zhang, and P. Liu, 2015: the Joint Institute for the Study of the Atmosphere and An objective method for partitioning the entire flood season into Ocean (JISAO), NPGO from http://www.o3d.org/npgo/ multiple sub-seasons. J. Hydrol., 528, 621–630, https://doi.org/ index.html, and NOI from the Pacific Fisheries Envi- 10.1016/j.jhydrol.2015.07.003. Cioffi, F., U. Lall, E. Rus, and C. K. B. Krishnamurthy, 2015: Space- ronmental Laboratory (PFEL). For ENSO, the MEI data time structure of extreme precipitation in Europe over the were taken from the NOAA Earth System Research Lab- last century. Int. J. Climatol., 35, 1749–1760, https://doi.org/ oratory (http://www.esrl.noaa.gov/psd/enso/mei/#ref_w). 10.1002/joc.4116. Compo, G. P., and Coauthors, 2011: The Twentieth Century Re- REFERENCES analysis Project. Quart. J. Roy. Meteor. Soc., 137, 1–28, https:// doi.org/10.1002/qj.776. Alexander, L. V., and Coauthors, 2006: Global observed Costa, A. C., and A. Soares, 2009: Trends in extreme precipitation changes in daily climate extremes of temperature and pre- indices derived from a daily rainfall database for the South cipitation. J. Geophys. Res., 111, D05109, https://doi.org/10.1029/ of Portugal. Int. J. Climatol., 29, 1956–1975, https://doi.org/ 2005JD006290. 10.1002/joc.1834. Allan, R. P., and B. J. Soden, 2008: Atmospheric warming and the Coulibaly, P., 2006: Spatial and temporal variability of Canadian amplification of precipitation extremes. Science, 321, 1481– seasonal precipitation (1900-2000). Adv. Water Resour., 29, 1484, https://doi.org/10.1126/science.1160787. 1846–1865, https://doi.org/10.1016/j.advwatres.2005.12.013. Asong, Z. E., M. N. Khaliq, and H. S. Wheater, 2016: Multisite Dai, A., 2008: Temperature and pressure dependence of the rain- multivariate modeling of daily precipitation and temperature snow phase transition over land and ocean. Geophys. Res. in the Canadian Prairie Provinces using generalized linear Lett., 35, L12802, https://doi.org/10.1029/2008GL033295.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 294 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

Daufresne, M., K. Lengfellner, and U. Sommer, 2009. Global Gregersen, I. B., H. J. D. Sørup, H. Madsen, D. Rosbjerg, P. S. warming benefits the small in aquatic ecosystems. Proc. Natl. Mikkelsen, and K. Arnbjerg-Nielsen, 2013: Assessing future Acad. Sci. USA, 106, 12 788–12 793, https://doi.org/10.1073/ climatic changes of rainfall extremes at small spatio-temporal pnas.0902080106. scales. Climatic Change, 118, 783–797, https://doi.org/10.1007/ Dhakal, N., and B. Tharu, 2018: Spatio-temporal trends in daily s10584-012-0669-0. precipitation extremes and their connection with North ——, H. Madsen, D. Rosbjerg, and K. Arnbjerg-Nielsen, 2015: Atlantic tropical for the southeastern United States. Long term variations of extreme rainfall in Denmark and Int. J. Climatol., 38, 3822–3831, https://doi.org/10.1002/joc.5535. southern Sweden. Climate Dyn., 44, 3155–3169, https://doi.org/ Di Lorenzo, E., and Coauthors, 2008: North Pacific Gyre Oscilla- 10.1007/s00382-014-2276-4. tion links ocean climate and ecosystem change. Geophys. Res. Grinsted, A., J. C. Moore, and S. Jevrejeva, 2004: Application of Lett., 35, L08607, https://doi.org/10.1029/2007GL032838. the cross wavelet transform and wavelet coherence to geo- Dong, W., Y. Lin, J. S. Wright, Y. Xie, X. Yin, and J. Guo, physical time series. Nonlinear Processes Geophys., 11, 561– 2018: Precipitable water and CAPE dependence of rainfall 566, https://doi.org/10.5194/npg-11-561-2004. intensities in China. Climate Dyn., https://doi.org/10.1007/ Guichard, F., and Coauthors, 2004: Modelling the diurnal cycle of s00382-018-4327-8. deep precipitating convection over land with -resolving d’Orgeville, M., W. R. Peltier, A. R. Erler, and J. Gula, 2014: models and single-column models. Quart. J. Roy. Meteor. Soc., Climate change impacts on Great Lakes Basin precipitation 130, 3139–3172, https://doi.org/10.1256/qj.03.145. extremes. J. Geophys. Res. Atmos., 119, 10 799–10 812, https:// Hales, S., S. J. Edwards, and R. S. Kovats, 2003: Impacts on health doi.org/10.1002/2014JD021855. of climate extremes. Climate Change and Human Health: Duan, W., B. He, K. Takara, P. Luo, M. Hu, N. E. Alias, and Risks and Responses,WMO,79–102,http://www.who.int/ D. Nover, 2015: Changes of precipitation amounts and ex- globalchange/publications/climatechangechap5.pdf. tremes over Japan between 1901 and 2012 and their connec- Hamed, K. H., and A. R. Rao, 1998: A modified Mann-Kendall tion to climate indices. Climate Dyn., 45, 2273–2292, https:// trend test for autocorrelated data. J. Hydrol., 204, 182–196, doi.org/10.1007/s00382-015-2778-8. https://doi.org/10.1016/S0022-1694(97)00125-X. Easterling, D. R., J. L. Evans, P. Ya. Groisman, T. R. Karl, K. E. Huang, H., J. M. Winter, E. C. Osterberg, R. M. Horton, and Kunkel, and P. Ambenje, 2000: Observed variability and trends B. Beckage, 2017: Total and extreme precipitation changes in extreme climate events: A brief review. Bull. Amer. Meteor. over the northeastern United States. J. Hydrometeor., 18, Soc., 81, 417–425, https://doi.org/10.1175/1520-0477(2000) 1783–1798, https://doi.org/10.1175/JHM-D-16-0195.1. 081,0417:OVATIE.2.3.CO;2. Hurrell, J. W., Y. Kushnir, and M. H. Visbeck, 2001: The North Elewa, H. H., E.-S. M. Ramadan, and A. M. Nosair, 2016: Spatial- Atlantic Oscillation. Science, 291, 603–605, https://doi.org/ based hydro-morphometric watershed modeling for the as- 10.1126/science.1058761. sessment of flooding potentialities. Environ. Earth Sci., 75, IPCC, 2013: Climate Change 2013: The Physical Science Basis. 927, https://doi.org/10.1007/s12665-016-5692-4. Cambridge University Press, 1535 pp., https://doi.org/10.1017/ Emori, S., and S. J. Brown, 2005: Dynamic and thermodynamic CBO9781107415324. changes in mean and extreme precipitation under changed ——, 2014: Climate Change 2014: Synthesis Report. IPCC, 151 pp., climate. Geophys. Res. Lett., 32,L17706,https://doi.org/10.1029/ http://www.ipcc.ch/report/ar5/syr/. 2005GL023272. Jiang, R., T. Y. Gan, J. Xie, and N. Wang, 2014: Spatiotemporal Fischer, E. M., and R. Knutti, 2016: Observed heavy precipitation variability of Alberta’s seasonal precipitation, their telecon- increase confirms theory and early models. Nat. Climate nection with large-scale climate anomalies and sea surface Change, 6, 986–991, https://doi.org/10.1038/nclimate3110. temperature. Int. J. Climatol., 34, 2899–2917, https://doi.org/ Frazier, A. G., and T. W. Giambelluca, 2016: Spatial trend analysis 10.1002/joc.3883. of Hawaiian rainfall from 1920 to 2012. Int. J. Climatol., 37, ——, ——, ——, ——, and C. C. Kuo, 2015: Historical and 2522–2531, https://doi.org/10.1002/joc.4862. potential changes of precipitation and temperature of Fu, X., C. C. Kuo, and T. Y. Gan, 2015: Change point analysis of Alberta subjected to climate change impact: 1900–2100. precipitation indices of western Canada. Int. J. Climatol., 35, Theor. Appl. Climatol., 127, 725–739, https://doi.org/10.1007/ 2592–2607, https://doi.org/10.1002/joc.4144. s00704-015-1664-y. Gan, T. Y., A. K. Gobena, and Q. Wang, 2007: Precipitation of ——, X. Yu, J. Xie, Y. Zhao, F. Li, and M. Yang, 2017: Recent southwestern Canada: Wavelet, scaling, multifractal analysis, changes in daily climate extremes in a serious water shortage and teleconnection to climate anomalies. J. Geophys. Res., metropolitan region, a case study in Jing-Jin-Ji of China. 112, D10110, https://doi.org/10.1029/2006JD007157. Theor. Appl. Climatol., 134, 565–584, https://doi.org/10.1007/ ——, and Coauthors, 2016: Possible climate change/variability and s00704-017-2293-4. human impacts, vulnerability of African drought prone regions, Jolliffe, I. T., 2002: Principal Component Analysis. Springer, its water resources and capacity building. Hydrol. Sci. J., 61, 488 pp. 1209–1226, https://doi.org/10.1080/02626667.2015.1057143. Kendall, M. G., 1955: Rank Correlation Methods. C. Griffin, 196 pp. Gao, L., J. Huang, X. Chen, Y. Chen, and M. Liu, 2017: Risk of Kishtawal, C. M., D. Niyogi, M. Tewari, R. A. Pielke, and J. M. extreme precipitation under nonstationarity conditions during Shepherd, 2010: Urbanization signature in the observed heavy the second flood season in the southeastern coastal region of rainfall climatology over India. Int. J. Climatol., 30, 1908–1916, China. J. Hydrometeor., 18, 669–681, https://doi.org/10.1175/ https://doi.org/10.1002/joc.2044. JHM-D-16-0119.1. Krichak, S. O., J. Barkan, J. S. Breitgand, S. Gualdi, and S. B. Gizaw, M. S., and T. Y. Gan, 2016: Possible impact of climate Feldstein, 2015: The role of the export of tropical moisture change on future extreme precipitation of the Oldman, Bow into midlatitudes for extreme precipitation events in the and Red Deer River Basins of Alberta. Int. J. Climatol., 36, Mediterranean region. Theor. Appl. Climatol., 121, 499–515, 208–224, https://doi.org/10.1002/joc.4338. https://doi.org/10.1007/s00704-014-1244-6.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC FEBRUARY 2019 Y A N G E T A L . 295

Kunkel, K. E., 2003: North American trends in extreme precipi- Monkam, D., 2002: Convective available potential energy (CAPE) tation. Nat. Hazards, 29, 291–305, https://doi.org/10.1023/ in Northern Africa and tropical Atlantic and study of its A:1023694115864. connections with rainfall in Central and West Africa during ——, K. Andsager, and D. D. R. Easterling, 1999: Long-term trends summer 1985. Atmos. Res., 62, 125–147, https://doi.org/10.1016/ in extreme precipitation events over the conterminous United S0169-8095(02)00006-6. States and Canada. J. Climate, 12, 2515–2527, https://doi.org/ Murugavel, P., S. D. Pawar, and V. Gopalakrishnan, 2012: Trends 10.1175/1520-0442(1999)012,2515:LTTIEP.2.0.CO;2. of convective available potential energy over the Indian re- ——, D. A. Robinson, S. Champion, X. Yin, T. Estilow, and R. M. gion and its effect on rainfall. Int. J. Climatol., 32, 1362–1372, Frankson, 2016: Trends and extremes in Northern Hemi- https://doi.org/10.1002/joc.2359. sphere snow characteristics. Curr. Climate Change Rep., 2, 65– Mwale, D., T. Y. Gan, K. Devito, C. Mendoza, U. Silins, and 73, https://doi.org/10.1007/s40641-016-0036-8. R. Petrone, 2009: Precipitation variability and its relationship Lemmen, D. S., and F. J. Warren, 2004: Climate change impacts to hydrologic variability in Alberta. Hydrol. Processes, 23, and adaptation: A Canadian perspective. Natural Resources 3040–3056, https://doi.org/10.1002/hyp.7415. Canada, 174 pp., https://cfs.nrcan.gc.ca/publications?id527428. Ntale, H. K., and T. Y. Gan, 2004: East African rainfall anomaly Lenderink, G., and E. Van Meijgaard, 2008: Increase in hourly patterns in association with El Niño/Southern Oscillation. precipitation extremes beyond expectations from tempera- J. Hydrol. Eng., 9, 257–268, https://doi.org/10.1061/(ASCE) ture changes. Nat. Geosci., 1, 511–514, https://doi.org/10.1038/ 1084-0699(2004)9:4(257). ngeo262. O’Gorman, P. A., 2014: Contrasting responses of mean and ex- Lepore, C., D. Veneziano, and A. Molini, 2015: Temperature and treme snowfall to climate change. Nature, 512, 416–418, https:// CAPE dependence of rainfall extremes in the eastern United doi.org/10.1038/nature13625. States. Geophys. Res. Lett., 42, 74–83, https://doi.org/10.1002/ Okonkwo, C., 2014: An advanced review of the relationships 2014GL062247. between Sahel precipitation and climate indices: A wavelet Li, J., Y. D. Chen, T. Y. Gan, and N.-C. Lau, 2018: Elevated in- approach. Int. J. Atmos. Sci., 2014, 759067, https://doi.org/ creases in human-perceived temperature under climate warm- 10.1155/2014/759067. ing. Nat. Climate Change, 8, 43–47, https://doi.org/10.1038/ Pedron, I. T., M. A. Silva Dias, S. de Paula Dias, L. M. Carvalho, s41558-017-0036-2. and E. D. Freitas, 2017: Trends and variability in extremes of Limsakul,A.,andP.Singhruck,2016: Long-term trends and vari- precipitation in Curitiba – Southern Brazil. Int. J. Climatol., ability of total and extreme precipitation in Thailand. Atmos. 37, 1250–1264, https://doi.org/10.1002/joc.4773. Res., 169, 301–317, https://doi.org/10.1016/j.atmosres.2015.10.015. Rasmusson, E. M., and J. M. Wallace, 1983: Meteorological aspects Lolis, C., and M. Türkes¸, 2016: Atmospheric circulation charac- of the El Nino/southern oscillation. Science, 222, 1195–1202, teristics favouring extreme precipitation in Turkey. Climate https://doi.org/10.1126/science.222.4629.1195. Res., 71, 139–153, https://doi.org/10.3354/cr01433. R Core Team, 2017: R: A language and environment for statistical Lovino, M. A., O. V. Müller, E. H. Berbery, and G. V. Müller, 2018: computing. R Foundation for Statistical Computing, https:// How have daily climate extremes changed in the recent past www.R-project.org/. over northeastern Argentina? Global Planet. Change, 168, 78– Riemann-Campe, K., K. Fraedrich, and F. Lunkeit, 2009: Global cli- 97, https://doi.org/10.1016/j.gloplacha.2018.06.008. matology of convective available potential energy (CAPE) and Mann, H. B., 1945: Nonparametric tests against trend. Econo- convective inhibition (CIN) in ERA-40 reanalysis. Atmos. Res., metrica, 13, 245–259, https://doi.org/10.2307/1907187. 93, 534–545, https://doi.org/10.1016/j.atmosres.2008.09.037. Mantua, N. J., and S. R. Hare, 2002: The Pacific Decadal Oscil- Seeley, J. T., and D. M. Romps, 2015: Why does tropical convective lation. J. Oceanogr., 58, 35–44, https://doi.org/10.1023/A: available potential energy (CAPE) increase with warming? 1015820616384. Geophys. Res. Lett., 42, 10 429–10 437, https://doi.org/10.1002/ Mass, C., A. Skalenakis, and M. Warner, 2011: Extreme pre- 2015GL066199. cipitation over the west coast of North America: Is there a Sen, P. K., 1968: Estimates of the regression coefficient based trend? J. Hydrometeor., 12, 310–318, https://doi.org/10.1175/ on Kendall’s Tau. J. Amer. Stat. Assoc., 63, 1379–1389, https:// 2010JHM1341.1. doi.org/10.1080/01621459.1968.10480934. Mekis, É., and L. A. Vincent, 2011: An overview of the Second Shadmani, M., S. Marofi, and M. Roknian, 2012: Trend analysis Generation Adjusted Daily Precipitation Dataset for trend in reference evapotranspiration using Mann-Kendall and analysis in Canada. Atmos.–Ocean, 49, 163–177, https://doi.org/ Spearman’s Rho tests in arid regions of Iran. Water Resour. 10.1080/07055900.2011.583910. Manage., 26, 211–224, https://doi.org/10.1007/s11269-011-9913-z. Miao, C., H. Ashouri, K.-L. Hsu, S. Sorooshian, and Q. Duan, 2015: Shephard, M. W., E. Mekis, R. J. Morris, Y. Feng, X. Zhang, Evaluation of the PERSIANN-CDR daily rainfall estimates in K. Kilcup, and R. Fleetwood, 2014: Trends in Canadian short- capturing the behavior of extreme precipitation events over duration extreme rainfall: Including an intensity-duration- China. J. Hydrometeor., 16, 1387–1396, https://doi.org/10.1175/ frequency perspective. Atmos.–Ocean, 52, 398–417, https:// JHM-D-14-0174.1. doi.org/10.1080/07055900.2014.969677. Milrad, S. M., J. R. Gyakum, and E. H. Atallah, 2015: A mete- Simonovic, S. P., A. Schardong, and D. Sandink, 2016: Mapping orological analysis of the 2013 Alberta flood: Antecedent extreme rainfall statistics for Canada under climate change large-scale flow pattern and synoptic–dynamic characteris- using updated intensity-duration-frequency curves. J. Water tics. Mon. Wea. Rev., 143, 2817–2841, https://doi.org/10.1175/ Resour. Plan. Manage., 143, https://doi.org/10.1061/(ASCE) MWR-D-14-00236.1. WR.1943-5452.0000725. Mladjic, B., L. Sushama, M. N. Khaliq, R. Laprise, D. Caya, and Song, X., S. Song, W. Sun, X. Mu, S. Wang, J. Li, and Y. Li, 2015: R. Roy, 2011: Canadian RCM projected changes to extreme Recent changes in extreme precipitation and drought over the precipitation characteristics over Canada. J. Climate, 24, 2565– Songhua River Basin, China, during 1960–2013. Atmos. Res., 2584, https://doi.org/10.1175/2010JCLI3937.1. 157, 137–152, https://doi.org/10.1016/j.atmosres.2015.01.022.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC 296 JOURNAL OF HYDROMETEOROLOGY VOLUME 20

Spearman, C., 1904: The proof and measurement of association Wang, X., G. Huang, and J. Liu, 2014: Projected increases in in- between two things. Amer.J.Psychol., 15, 72–101, https:// tensity and frequency of rainfall extremes through a regional doi.org/10.2307/1412159. climate modeling approach. J. Geophys. Res. Atmos., 119, Tan, X., and T. Y. Gan, 2017: Non-stationary analysis of the fre- 13 271–13 286, https://doi.org/10.1002/2014JD022564. quency and intensity of heavy precipitation over Canada and Wang, Y., E. A. McBean, and P. Jarrett, 2015: Identification of their relations to large-scale climate patterns. Climate Dyn., changes in heavy rainfall events in Ontario, Canada. Stochastic 48, 2983–3001, https://doi.org/10.1007/s00382-016-3246-9. Environ. Res. Risk Assess., 29, 1949–1962, https://doi.org/10.1007/ ——, ——, S. Chen, and B. Liu, 2018a: Modeling distributional s00477-015-1085-6. changes in winter precipitation of Canada using Bayesian Wolter, K., and M. S. Timlin, 1993: Monitoring ENSO in COADS spatiotemporal quantile regression subjected to different tel- with a seasonally adjusted principal component index. Proc. econnections. Climate Dyn., https://doi.org/10.1007/s00382-018- 17th Climate Diagnostics Workshop, Norman, OK, NOAA, 4241-0. 52–57. ——, ——, and Y. D. Chen, 2018b: Moisture sources and pathways ——, and ——, 1998: Measuring the strength of ENSO events: associated with the spatial variability of seasonal extreme How does 1997/98 rank? Weather, 53, 315–324, https://doi.org/ precipitation over Canada. Climate Dyn., 50, 629–640, https:// 10.1002/j.1477-8696.1998.tb06408.x. doi.org/10.1007/s00382-017-3630-0. ——, and ——, 2011: El Niño/Southern Oscillation behaviour ——, ——, and ——, 2018c: Synoptic moisture pathways associ- since 1871 as diagnosed in an extended multivariate ENSO ated with mean and extreme precipitation over Canada index (MEI.ext). Int. J. Climatol., 31, 1074–1087, https://doi.org/ for summer and fall. Climate Dyn., https://doi.org/10.1007/ 10.1002/joc.2336. s00382-018-4300-6. Xi, Y., C. Miao, J. Wu, Q. Duan, X. Lei, and H. Li, 2018: Spatio- Tandon, N. F., X. Zhang, and A. H. Sobel, 2018: Understanding the temporal changes in extreme temperature and precipitation dynamics of future changes in extreme precipitation intensity. events in the Three-Rivers Headwater region, China. Geophys. Res. Lett., 45, 2870–2878, https://doi.org/10.1002/ J. Geophys. Res. Atmos., 123, 5827–5844, https://doi.org/10.1029/ 2017GL076361. 2017JD028226. Theobald, A., H. Mcgowan, and J. Speirs, 2018: Teleconnection in- Ye, B., A. D. Del Genio, and K. K. W. Lo, 1998: CAPE vari- fluence of precipitation-bearing synoptic types over the Snowy ations in the current climate and in a climate change. Mountains region of south-east Australia. Int. J. Climatol., 38, J. Climate, 11, 1997–2015, https://doi.org/10.1175/1520- 2743–2759, https://doi.org/10.1002/joc.5457. 0442-11.8.1997. Torrence, C., and G. P. Compo, 1998: A practical guide to wavelet Zhai, P., X. Zhang, H. Wan, and X. Pan, 2005: Trends in total analysis. Bull. Amer. Meteor. Soc., 79, 61–78, https://doi.org/ precipitation and frequency of daily precipitation extremes 10.1175/1520-0477(1998)079,0061:APGTWA.2.0.CO;2. over China. J. Climate, 18, 1096–1108, https://doi.org/10.1175/ Ventura, V., C. J. Paciorek, and J. S. Risbey, 2004: Controlling the JCLI-3318.1. proportion of falsely rejected hypotheses when conducting Zhang, X., W. D. Hogg, and É. Mekis, 2001: Spatial and temporal multiple tests with climatological data. J. Climate, 17, 4343– characteristics of heavy precipitation events over Canada. 4356, https://doi.org/10.1175/3199.1. J. Climate, 14, 1923–1936, https://doi.org/10.1175/1520-0442(2001) Vincent, L. A., and É. Mekis, 2006: Changes in daily and extreme 014,1923:SATCOH.2.0.CO;2. temperature and precipitation indices for Canada over the ——, L. Alexander, G. C. Hegerl, P. Jones, A. K. Tank, T. C. twentieth century. Atmos.–Ocean, 44, 177–193, https://doi.org/ Peterson, B. Trewin, and F. W. Zwiers, 2011: Indices for 10.3137/ao.440205. monitoring changes in extremes based on daily temperature ——,X.Zhang,R.D.Brown,Y.Feng,E.Mekis,E.J.Milewska, and precipitation data. Wiley Interdiscip. Rev.: Climate Change, H. Wan, and X. L. Wang, 2015: Observed trends in Canada’s cli- 2, 851–870, https://doi.org/10.1002/wcc.147. mate and influence of low-frequency variability modes. J. Cli- Zhou, X., G. Huang, X. Wang, and G. Cheng, 2018: Future changes mate, 28, 4545–4560, https://doi.org/10.1175/JCLI-D-14-00697.1. in precipitation extremes over Canada: Driving factors and Wang,B.,M.Zhang,J.Wei,S.Wang,S.Li,Q.Ma,X.Li,andS.Pan, inherent mechanism. J. Geophys. Res. Atmos., 123, 5783–5803, 2013: Changes in extreme events of temperature and precipi- https://doi.org/10.1029/2017JD027735. tation over Xinjiang, northwest China, during 1960–2009. Quat. Zilli,M.T.,L.M.V.Carvalho,B.Liebmann,andM.A.Silva Int., 298, 141–151, https://doi.org/10.1016/j.quaint.2012.09.010. Dias, 2017: A comprehensive analysis of trends in ex- Wang, J., and X. Zhang, 2008: Downscaling and projection of treme precipitation over southeastern coast of Brazil. winter extreme daily precipitation over North America. Int. J. Climatol., 37, 2269–2279, https://doi.org/10.1002/ J. Climate, 21, 923–937, https://doi.org/10.1175/2007JCLI1671.1. joc.4840.

Unauthenticated | Downloaded 10/07/21 09:59 PM UTC