The Canadian Prairies depend on summer precipitation especially during the early to mid-growing season (May through August) when the majority of annual precipitation normally occurs (e.g., Bonsal et al., 1993). High natural variability in growing season precipitation causes periodic occurrences of extreme precipitation (Li et al., 2017; Liu et al., 2016) and droughts that are often associated with reduced agriculture yields, low streamflow, and increased occurrence of forest fires (Wheaton et al., 2005; Bonsal and Regier, 2007). Drought events with great environmental and economic impacts on the Canadian Prairies have occurred in 1961, 1988, 2001–2002, and as recent as 2015 (Dey, 1982; Liu et al., 2004; Bonsal et al., 1999; Wheaton et al., 2005; Shabbar et al., 2011; Bonsal et al., 2013; Szeto et al., 2016). The sub-seasonal forecast of precipitation for the growing season is crucial for the agriculture, water resource management, and the economy of the region. Therefore, an investigation into the causes of inter-annual variability in the growing season precipitation of the Canadian Prairies is needed.

Low precipitation and extended dry periods on the Canadian Prairies are often associated with an upper-level ridge and a persistent high pressure centred over the region (Dey, 1982; Liu et al., 2004). These prolonged atmospheric anomalies often concurred with abnormal boundary layer conditions such as a large-scale sea surface temperature (SST) anomalies in the Pacific Ocean (Shabbar and Skinner, 2004). Large-scale oscillation in the SST anomalies in the Pacific Ocean, namely El Niño and the Pacific Decadal Oscillation (PDO), can affect the hydroclimatic pattern in summer over North America, although the strongest impacts of these boundary conditions occur during the boreal winter. Inter-annual variability such as El Niño–Southern Oscillation (ENSO) is linked with extended droughts in the Prairies (Bonsal et al., 1999; Shabbar and Skinner, 2004). Interdecadal oscillations such as the PDO and the Atlantic Multi-decadal Oscillation (AMO) also affect the seasonal temperature and precipitation in the Canadian Prairies (Shabbar et al., 2011).

ENSO's relationship with the Canadian Prairies' precipitation has been studied extensively. Previous investigations (e.g., Shabbar et al., 2011) have found that El Niño events are associated with a summer moisture deficit in western Canada while La Niña events cause an abundance of moisture in far western Canada (British Columbia and Yukon). However, they also noted that, although tropical SST variability accounted for some aspects of the large-scale circulation anomalies that influence the Canadian Prairies meteorological drought, a consistent and clear-cut relationship was not found. The warm phase of ENSO often favours drought in this region, especially during the growing season after the mature phase of El Niño (Bonsal and Lawford, 1999; Shabbar and Skinner, 2004). The positive North Pacific Mode (NPM, Hartmann et al., 2015) like the North Pacific SST anomaly pattern often follows a matured El Niño, and the accompanying atmospheric ridging leads to extended dry spells over the Prairies during the growing season (Bonsal and Lawford, 1999). Furthermore, in association with the recent North Pacific SST anomaly from 2013 to 2014, researchers have attributed the precipitation deficit in California during 2013 to the anomalous upper-level ridge over western North America (Wang et al., 2014; Szeto et al., 2016).

The aforementioned SST variations mostly vary on inter-annual and decadal scales. Another important factor that affects the weather patterns in North America is the Madden–Julian Oscillation (MJO), an intra-seasonal (40–90 days) oscillation in convection and precipitation pattern over the tropics (Madden and Julian, 1971; Zhang, 2005; Riddle et al., 2013; Carbone and Li, 2015). MJO is a coupled atmosphere–ocean oscillation involving convection and large-scale equatorial waves, which produces an eastward propagation of the tropical convection anomaly (Madden and Julian, 1971). The MJO affects the winter temperature and precipitation in North America and Europe through its impact on moisture transport associated with the “Pineapple Express” and its effects on the North Atlantic Oscillation and stratospheric polar vortex (Cassou, 2008; Garfinkel et al., 2012; Rodney et al., 2013). MJO is also connected to the summer precipitation anomalies in the southwestern United States (Lorenz and Hartmann, 2006). During the warm season, MJO's impact on the Canadian Prairies' precipitation has not been thoroughly investigated as MJO's amplitude is weak during spring and early summer. The amplitude of MJO in spring and early summer is related to the inter-annual variation of the tropical SST, especially the SST in the central Pacific (Hendon et al., 2007; Marshall et al., 2016). MJO in terms of the Real-time Multivariate MJO index (RMM, Wheeler and Hendon, 2004) was extremely strong in the early spring of 2015 with a positive PDO-like SST anomaly in the central Pacific and, at the same time, El Niño started to strengthen.

MJO activities in the western Pacific under the modulation of inter-annual SST variability have the potential to act together with ENSO and impact mid-tropospheric circulation over western Canada and, thus, warm-season precipitation over the Canadian Prairies. The goal of this study is to demonstrate that MJO has contributed to the 2015 growing season drought in the Canadian Prairies through the propagation of stationary Rossby waves. Subsequently, further investigations are carried out to determine if similar relationships exist in association with other summer extreme precipitation events during the instrumental record (1979–2016). Section 2 provides the datasets and methodology used in this paper while Sect. 3 presents the analysis of the upper-level circulation anomaly and SST pattern associated with the 2015 drought. This is followed by the examination of the effects of central Pacific SST anomalies and MJO on the summer precipitation in the Canadian Prairies. The mechanism by which MJO affects summer precipitation when equatorial central Pacific SST is warmer than normal is discussed in Sect. 4, followed by the summary and concluding remarks in Sect. 5.

Multiple observation and reanalysis datasets are used to investigate the circulation anomalies associated with the Canadian Prairies' growing season (May–August) precipitation. The observed precipitation is taken from the Climate Prediction Center (CPC) Merged Analysis of Precipitation (CMAP) dataset (Xie and Arkin, 1997; NOAA CPC, 2017a). Geopotential height (GPH) fields from the National Centers for Environmental Predictions (NCEP) Reanalysis (Kalnay et al., 1996; NOAA NCEP, 2017) and the European Centre for Medium-Range Weather Forecast's (ECMWF) ERA-Interim reanalysis (Dee et al., 2011; ECMWF, 2017) are used to analyze the mid- and upper-level (500 and 200 hPa) atmospheric circulation patterns.

To represent the central Pacific SST anomaly, the NINO4 SST index (Rayner et al., 2003; NOAA, 2017b) from the CPC of the National Oceanic and Atmospheric Administration (NOAA) is used since the NINO4 region is near the central Pacific and spans over the dateline (5∘ S–5∘ N, 160∘ E–150∘ W). Multivariate ENSO Index (NOAA, 2017c) data are retrieved from NOAA's Climate Data Center (CDC) website and are used to determine the ENSO phase (Wolter, 1987; Wolter and Timlin, 1993). In particular, the El Niño condition is defined when the monthly mean index of MEI is larger than 0.5 (Andrews et al., 2004).

The Real-time Multivariate MJO series (RMM1 and RMM2; Bureau of Meteorology of Australia, 2017) developed by Wheeler and Hendon (2004) are used to identify periods of strong MJO activity as the MJO amplitudes are directly calculated by the square root of RMM1 + RMM2. For MJO intensities over the investigated regions, we used the monthly averaged pentad MJO indices from the NOAA CPC's MJO index (Xue et al., 2002; NOAA CPC, 2017b), which have 10 indices representing locations around the globe. The CPC's MJO index is based on extended empirical orthogonal function (EEOF) analysis on pentad velocity potential at 200 hPa. A total of 10 MJO indices on a daily scale are constructed by projecting the daily (00:00 UTC) velocity potential anomalies at 200 hPa (CHI200) onto the 10 time-lagged patterns of the first EEOF of pentad CHI200 anomalies (Xue et al., 2002). Negative values of 10 MJO indices correspond to enhanced convection in the 10 regions centred on 20∘ E, 70∘ E, 80∘ E, 100∘ E, 120∘ E, 140∘ E, 160∘ E, 120∘ W, 40∘ W, and 10∘ W in the tropics. MJO indices usually vary between -2 and 2, with negative values indicating above-average convective activities in the corresponding region. Because boreal summer usually corresponds to a period of a weaker amplitude of MJO than the winter, we chose the monthly mean value of -0.3 as the criterion of strong convection which is connected to MJO as the indices generally vary between -1 and 1. An MJO-4 index (centred on 140∘ E) of less than -0.3 was considered a relatively strong convection in the western Pacific, which has been found to be a source region of stationary Rossby waves (Simmons and Hoskins, 1980). SST observations include the Extended Reconstructed Sea Surface Temperature (ERSST) v4 (NOAA, 2017a; Huang et al., 2015). Outgoing Longwave Radiation (NOAA, 2017d) data from NOAA Interpolated Outgoing Longwave Radiation are used to derive the composite of anomalies of OLR for a certain phase of MJO.

Our study focuses on the growing season precipitation in the provinces of Alberta and Saskatchewan in the Canadian Prairies, where the largest deficits were observed in 2015. Specifically, the regional mean precipitation over 115–102.5∘ W, 50–57.5∘ N is used (boxed area in Fig. 1, top panel) to represent the Canadian Prairies east of the Rocky Mountains and south of the boreal forest. The chosen region also covers most of the arable land in the Canadian Prairies. Considering the unique MJO-4 and NINO4 indices for 2015, the relationship between the Prairies' warm-season (May–August) precipitation with MJO-4 and ENSO during the instrumental records is investigated using correlation and regression. Though the dry months of the 2015 growing season are May and June when MJO-4 was in negative phase, we want to study the statistical relationship between MJO-4 and the Prairies' precipitation in the whole period of the growing season (May–August). The possible mechanism behind the correlation between MJO-4 and the Prairies' warm-season precipitation under El Niño condition is further investigated by analyzing the upper-level circulation associated with convection in the tropical Pacific and stationary Rossby waves in mid-latitudes.

The summer of 2015 is the first summer after the development of El Niño during the 2014–2015 winter. Though the upper-level GPH pattern, seen in summer 2015, can be attributed to the SST modes in the Pacific, namely ENSO and NPM, the precipitation in the western Canadian Prairies is not strongly correlated with either. Bonsal and Lawford (1999) found that more extended dry spells tend to occur in the Canadian Prairies during the second summer following the mature stage of the El Niño events. The winter precipitation in Canada has a strong connection to ENSO (Shabbar et al., 1997), whereas summer precipitation, in most regions of western Canada (except the coast of British Columbia and Southern Alberta), does not have a significant correlation with ENSO. This is consistent with our investigation using instrumental records from 1948 to 2016.

Growing season precipitation in the Canadian Prairies is affected by many factors. Precipitation deficits can occur under various circulation and lower boundary conditions. Thus, it is not expected that a universal condition for all the significant droughts in the region can be identified. In fact, extreme drought events have been found in both El Niño and La Niña years. A previous study by Bonsal and Lawford (1999) indicates the meteorological drought often occurs after the mature phase of El Niño, which is not the case for 2015. The associated anomaly in the North Pacific represented by the NPM positive phase is consistent with their results. The direct linkage between ENSO and the summer precipitation in the Canadian Prairies is not clear. In fact, the correlation between MEI and the precipitation in the investigated region is -0.096 (p=0.239, sample size = 152). The investigated region's growing season precipitation does not possess a significant correlation with ENSO, which is consistent with other researchers' findings (Dai and Wigley, 2000).

The regression pattern is consistent with stationary Rossby wave theory as shown in a hierarchy of theoretical and modelling studies (Karoly et al., 1989; Simmons et al., 1983; Hoskins and Ambrizzi, 1993; Ambrizzi and Hoskins, 1997; Held et al., 2002). A similar wave train extends from the western Pacific toward extratropical South America but at lower latitudes compared to its counterpart in the Northern Hemisphere (not shown). The node of the wave train in western Canada and the Pacific Northwest of the US corresponds to an anomalous ridge, which is in phase of El Niño forcing. When the convection in the region associated with MJO-4 is weaker than normal (MJO-4 > 0), a wave train with the opposite sign will reach western Canada, which then counteracts the El Niño forcing. Thus, the weak correlation between Canadian Prairies precipitation and ENSO is understandable as MJO plays an additional role that enhances or cancels out the GPH anomaly caused by El Niño.

In the mid-latitude North America, the atmospheric response to the tropical forcing in the western Pacific depends on the mean circulation condition associated with tropical SST. Intraseasonal tropical convection oscillation in the western Pacific associated with the MJO-4 index cannot determine the sign of the precipitation anomaly in the Prairies alone. Both warm SST in the central Pacific and strong tropical convection in the western Pacific and Maritime Continent are essential to cause a significant precipitation deficit in the western Canadian Prairies. Warm SST in the central Pacific causes an eastward expansion of Pacific warm pool that favours enhanced MJO activity in the western-central Pacific (Hendon et al., 1999; Marshall et al., 2016). As shown by the ray-tracing result, the NINO4 also affects the wavenumber of the quasi-stationary Rossby waves over the source region in the western Pacific. Under warm NINO4, the wavenumbers tend to be smaller due to stronger westerly winds in the source region and these waves can propagate northeastward into western Canada. Conversely, from May to August under cold NINO4, the westerly flow is weaker, and the meridional vorticity gradient is stronger in the subtropics near the source region. These mean flow conditions correspond to waves with lager wavenumbers that cannot propagate across the dateline.

In the year 2015, the SST anomaly in the Pacific (e.g., ENSO, NPM) coincided with the anomalous ridge on the west coast of Canada. This positive GPH anomaly was associated with the strong negative MJO4 indices, it then caused a blocking pattern, and suppressed precipitation in the Canadian Prairies in the early summer through the mechanism discussed above. Although the El Niño continued to strengthen in July and August 2015, the active convection associated with MJO in the western Pacific propagated eastward into the central Pacific. As the convection in the western Pacific/Maritime Continent waned, the precipitation in the Canadian Prairies returned to slightly above normal in July.

The cause of the 2015 summer precipitation deficit in the western Canadian Prairies is investigated in relation to atmospheric circulation anomalies, SST, and the intraseasonal tropical convection oscillation, MJO. The drought in western Canada is immediately related to an anomalous upper-level ridge that persisted over the west coast of Canada and Alaska since fall 2014. This ridge was likely associated with a developing El Niño that was enhanced by the MJO.

In general, MJO-4 indices demonstrated significant correlation with the meteorological drought over the Canadian Prairies from May to August when the SST in the central Pacific is warm (NINO4 > 0), which also corresponds to a stronger MJO amplitude in boreal summer. Our study discovered that MJO phase/strength is connected to the anomalous ridge over western Canada through the propagation of stationary Rossby waves from the western Pacific when NINO4 is positive. Though seasonally MJO is weaker in summer, the spring and early summer MJO amplitudes are larger than normal when the central Pacific SST is warmer than normal (NINO4 > 0). The teleconnection between the Canadian Prairies precipitation deficit and MJO is stronger when NINO4 is positive. The underlying cause of this significant correlation between MJO-4 indices and the prairie precipitation in May–August is a stationary Rossby wave train originating from the Maritime Continent and western Pacific which propagates into Canada. The ray-tracing experiments show the main difference between a warm phase of NINO4 and a cold phase is the changes in stationary Rossby wave wavenumber over the source region. Under NINO4 > 0.5 May–August conditions, the total wavenumber is about 4 and can propagate into western Canada if the waves in the experiments are oriented relatively zonally. Compared to NINO4 > 0.5, NINO4 < -0.5 corresponds to a weaker zonal wind and stronger meridional gradient of absolute vorticity in the subtropics of the source region (140–150∘ E). Hence, the wavenumbers of stationary Rossby waves from the source region under NINO4 < -0.5 are larger (about 6), and these waves fail to reach the western hemisphere. The intra-seasonal predictability of the growing season precipitation in the Canadian Prairies can be potentially improved by including the MJO amplitude and phase factors for medium-range and intra-seasonal projection in addition to the ENSO effect especially when the central-Pacific SST is warm.