Extreme precipitation events impact the Pacific Northwest during winter months, causing flooding, landslides, extensive property damage, and loss of life. Outstanding questions about such events include whether there are a range of associated synoptic evolutions, whether such evolutions vary along the coast, and the associated rainfall duration and variability. To answer these questions, this study uses 60 years of National Climatic Data Center (NCDC) daily precipitation observations to identify the top 50 events in two-day precipitation at six coastal stations from northern California to northwest Washington. NCEP–NCAR reanalysis data were used to construct synoptic composite evolutions of these events for each coastal location. Most regional flooding events are associated with precipitation periods of 24 h or less, and two-day precipitation totals identify nearly all major events. Precipitation areas of major events are generally narrow, roughly 200 km in width, and most are associated with atmospheric rivers. Composite evolutions indicate negative anomalies in sea level pressure and upper-level height in the central Pacific, high pressure anomalies over the southwest United States, large positive 850-hPa temperature anomalies along the coast and offshore, and enhanced precipitable water and integrated water vapor fluxes over southwest to northeast swaths. A small subset of extreme precipitation events over the southern portion of the domain is associated with a very different synoptic evolution: a sharp trough in northwesterly flow and post-cold-frontal convection. High precipitable water values are more frequent during the summer, but are not associated with heavy precipitation due to upper-level ridging over the eastern Pacific and weak onshore flow that limit upward vertical velocities.
The northwest United States frequently experiences extreme precipitation events during the winter months, resulting in billions of dollars of damage as well as loss of life. Since 1955, these storms were responsible for roughly two-thirds of the presidential disaster declarations in Washington State and Oregon, and nearly one-quarter of the declarations in California (see http://www.fema.gov/news/disasters.fema). For example, during December 1996 and January 1997, heavy rain (25–100 cm in two weeks) produced severe flooding over portions of California (CA), Washington (WA), and Oregon (OR), causing 3.9 billion dollars (all amounts are in U.S. dollars) in losses and 36 deaths (see http://www.ncdc.noaa.gov/oa/reports/billionz.html). More recently, heavy rainfall and resulting flooding during January 2009 closed Interstate 5 and other major routes in Washington State, flooded major river drainages throughout the Northwest, and heavily damaged the Howard Hanson Dam in the Washington Cascades, putting 10–20 billion dollars of assets and infrastructure, as well as tens of thousands of people, at risk (Mastin et al. 2010; Neiman et al. 2011).
Most extreme precipitation events along the North American west coast are associated with narrow plumes of above-normal water vapor that stretch from the subtropics or tropics to the coast and are often referred to as atmospheric rivers (ARs). In a seminal paper on the topic, Zhu and Newell (1998) found that greater than 90% of the hemispheric meridional water vapor flux is transported by such features, with four to five evident at any given time. AR water vapor is often quickly converted to heavy precipitation upon landfall on coastal and inland mountain ranges, inundating local watersheds and causing severe flooding in low-lying areas (Ralph et al. 2006; Stohl et al. 2008; Viale and Nuñez 2011; Dettinger et al. 2011; Neiman et al. 2011).
Some studies have suggested that anthropogenic global warming could impact the severity and frequency of extreme precipitation events in the future (Groisman et al. 2005; Held and Soden 2006; Trenberth et al. 2007). It has been theorized that changes in equator-to-pole temperature gradients (Yin 2005) and Hadley cell expansion could push the storm track poleward in the Northern Hemisphere (McCabe et al. 2001; Hu and Fu 2007; Lu et al. 2007; Meehl et al. 2007). Such changes could potentially have a large impact on precipitation in the Pacific Northwest and California (Salathé 2006; Dettinger 2011; Mass et al. 2011) since ARs are generally on the southern flank of the jet stream. A thorough understanding of the synoptic conditions conducive to extreme precipitation events along the west coast of North America is essential for making future projections.
A series of major floods over the western United States during the 1980s and 1990s stimulated research on the nature of heavy precipitation over the region. Lackmann and Gyakum (1999) examined the composite synoptic evolution associated with 46 heavy precipitation events over Washington State in which each of four observing sites received at least 12.5 mm (~0.5 in.) of daily precipitation. Composites of 500-hPa geopotential heights indicated a positive anomaly over the Bering Sea prior to such events, with a negative height anomaly over the Gulf of Alaska and ridging over the southwest United States, resulting in strong southwesterly flow approaching the Pacific Northwest. Using a piecewise potential vorticity (PV) analysis for one event, this paper found a dominant contribution to water vapor transport by mobile cyclonic disturbances rather than planetary-scale flow. Ralph et al. (2004) used aircraft dropsonde observations during the California Land-Falling Jets (CALJET) Experiment to demonstrate the relatively narrow, shallow nature of AR water vapor plumes and their close association with the low-level jets in the warm sector preceding cold fronts. These findings were confirmed by compositing satellite-observed microwave data. Dettinger (2004) used National Centers for Environmental Prediction–National Center for Atmospheric Research (NCEP–NCAR) reanalysis grids from 1948 to 1996 to identify 206 days with water vapor flux signatures similar to those found during AR events. He found that all events occurred between October and April, with a peak in January and February, and were associated with warmer and wetter conditions than normal. Neiman et al. (2002) studied the relationship of the speed and height of the AR-related low-level jet and coastal mountain precipitation, while Neiman et al. (2005) examined the relative importance of brightband and non-brightband precipitation on coastal terrain. Ralph et al. (2005) used dropsonde observations from CALJET-1998 and the Pacific Land-Falling Jets Experiment (PACJET-2001) to study the structure of eastern Pacific AR events and their interannual variability for two different phases of ENSO.
More recently, Ralph et al. (2006) found that ARs were associated with seven floods on California’s Russian River. They suggested that not all ARs produce flooding, with flooding requiring antecedent soil saturation, intense precipitation, and ARs remaining over a watershed for an extended period. Bao et al. (2006) found that local water vapor convergence was a key process in the formation of simulated water vapor plumes, although long-distance transport of tropical water vapor was possible in some cases. Neiman et al. (2008b) established a climatology of West Coast ARs using satellite-based integrated water vapor (IWV) for 1997 through 2005. In that paper they examined the synoptic conditions of north coast [WA, OR, and British Columbia (BC)] and south coast (CA) ARs. They found ARs in all seasons, with warm season events bringing little precipitation. AR water vapor signatures greatly outnumbered heavy precipitation and flooding events, suggesting that ARs are necessary, but not sufficient, for extreme precipitation situations. Neiman et al. (2008a) used the Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) to examine the November 2006 AR event that devastated areas of Washington and Oregon, revealing similar structures to past events studied using dropsonde and reanalysis data. Roberge et al. (2009), studying large integrated water vapor events over the west coast of Canada, found a range of possible air trajectories, with southwest trajectories associated with the largest precipitation events. Neiman et al. (2011) used several decades of daily stream flow data from unregulated rivers in Washington State and North American Regional Reanalysis (NARR) grids to examine synoptic conditions leading to flooding in those watersheds. Most flooding events were found to be associated with AR conditions and either westerly or southwesterly flow. The top 10 annual peak daily flows on all these watersheds were accompanied by heavy precipitation, low-level moist neutral static stability, and elevated freezing levels.
Although much work has been done to understand West Coast extreme precipitation events, many important questions still remain and are considered in this paper: 1) How long do extreme precipitation events that cause high river flows last and what are the characteristics of their temporal evolution? 2) How does the large-scale synoptic structural evolution associated with extreme precipitation events change with latitude along the coast? 3) Are there synoptic conditions other than atmospheric rivers that can cause extreme coastal precipitation, and if so, what are these alternate evolutions and how frequently do they occur? 4) What is the typical horizontal scale associated with extreme precipitation events?
To evaluate the temporal evolution of such heavy precipitation events, 60 years of hourly National ClimaticData Center (NCDC) precipitation records at two stations along the Oregon coast are examined. To define extreme precipitation events and to establish the climatology of these events over the last 60 years, NCDC daily precipitation observations from six coastal stations from northwest Washington to northern California are used. Synoptic composites and composite evolutions are created for the top 50 events at each station to determine differences along the coast, and the individual evolutions are examined to determine their variability. Finally, the synoptic-scale evolutions for heavy precipitation events that occurred in the absence of an AR are examined.
2. Event definition and temporal variation
This study is based on a series of relatively evenly spaced coastal stations in the Pacific Northwest. Specifically, NCDC Global Historical Climatology Network (GHCN) daily precipitation observations from 1950 to 2009 for six coastal stations from northern Washington State to northern California (Fig. 1) were used to identify extreme precipitation events. The northernmost station, Forks (FORW1), is located in the northwest corner of Washington State, west of the Olympic Mountains. Four of the stations are located in Oregon, including Astoria (KAST), Newport (KONP), North Bend (KOTH), and Brookings (K4BK). Astoria and Newport have relatively low coastal terrain to their east, while North Bend and Brookings are west of the higher Klamath Mountains. Finally, Eureka (KEKA) is located along the northern California coast, with the relatively high terrain of the South Fork Mountains to the east. These stations were chosen for their close proximity to the coast, a relatively long period of record, even latitudinal spacing, and an absence of upwind terrain, which can lead to modulation by rain shadowing.
A key question regards the duration of precipitation events causing flooding over the region. Several recent storms in the Pacific Northwest, such as those in November 2006, December 2007, and January 2009, have produced flooding on Washington and Oregon rivers in association with precipitation events of one to three days (see the NCDC Storm Events database at http://www4.ncdc.noaa.gov/cgi-win/wwcgi.dll?wwEvent~Storms). The “time of concentration” is the time required for a large rise in river level after rain begins at a significant rate [usually around 0.50–0.75 cm (0.2–0.3 in.) of rain per hour] and is generally less than 24 h for the Northwest U.S. rivers examined in this study (L. Schick, U.S. Army Corps of Engineers, Northwest region, lead forecaster, 2011, personal communication; D. Lettenmeier, University of Washington, Department of Civil Engineering, 2011, personal communication).1
To further evaluate the period of precipitation correlating best with regional river levels, U.S. Geological Survey (USGS) average daily river discharges for naturally flowing rivers along the West Coast (Table 1) were compared to daily precipitation from the GHCN dataset. Daily values were used since very few observing sites offer continuous long-term hourly precipitation or hourly river levels. It is important to note that one-day GHCN values are potentially problematic since extreme precipitation events could span calendar days. At each of the six coastal precipitation sites, one- through four-day precipitation totals were correlated with the mean daily river gauge discharge on free-flowing rivers for lags of zero, one, two, and three days.2 For all precipitation sites, the highest correlations were with a nearby river.
Table 2 shows the average correlations (for all six precipitation sites) of all combinations of lag and precipitation period. For all lags, the 24-h period had the lowest correlations, undoubtedly due to the event splitting (calendar day) issue noted above. The correlations are highest for the 72-h period, with the 48-h period close behind. There were high correlations for zero- and one-day lags and a fairly rapid decline for longer lags. These results suggest that the relevant precipitation period for hydrologically meaningful events is one or two days. First, the relevant period is clearly greater than a few hours, since the 24-h precipitation period has a lower correlation than longer intervals. Secondly, the fact that the correlations are relatively constant for 48 h and greater suggests that 48 h contains the most relevant precipitation amounts. Furthermore, the substantial dropoff for a greater than one-day lag suggests that antecedent precipitation is not relevant. Thus, based on the physical nature (response time) of Northwest drainages, our analysis of the correlations between precipitation and river flow, and the availability of daily precipitation data, two-day total precipitation was used to define extreme events in this study. Furthermore, only the cool, wet season (October–March) was considered since flooding is extremely rare during the warm season in this region.
Daily GHCN precipitation totals were used to find the top 50 two-day events at each of the six stations (Fig. 1).3 Dates were eliminated if they were within five days of another event with greater precipitation, ensuring that each event represented a distinct synoptic situation. Additionally, the data were quality controlled to eliminate days with large daily totals following days with missing data, since the resulting large totals are usually not associated with a major event but are rather indicative of equipment malfunction or observer absence. As a check on the validity of the two-day event assumption, the top 50 one- and three-day events were also determined from the same quality-controlled dataset. A nearly identical list of events was found.
As noted earlier, coastal stations were used to reduce the potential of mesoscale modulation of precipitation intensity that would be unrepresentative of the regional precipitation. To further explore the applicability of coastal stations, the National Oceanic and Atmospheric Administration (NOAA) Climate Prediction Center (CPC) quarter-degree Daily U.S. Unified Precipitation Dataset (Higgins et al. 1996) for 1950–2009 was used to examine precipitation spatial correlations associated with the 50 extreme two-day coastal precipitation events [CPC U.S. Unified Precipitation data are provided by the NOAA Office of Oceanic and Atmospheric Research Earth System Research Laboratory Physical Sciences Division (NOAA/OAR/ESRL PSD), Boulder, Colorado, from their website at http://www.esrl.noaa.gov/psd/]. Specifically, the top 50 extreme two-day coastal precipitation events were correlated with temporally coincident two-day gridded precipitation across the region (Fig. 2). High correlations surround each station on a southwest–northeast axis at every location except Newport, whose swath was oriented more west–east, and Forks, which lies at the northern boundary of the analysis domain. Most stations show strong anticorrelations over adjacent areas, particularly to the south, where there is generally suppression of precipitation by high pressure located to the southeast of ARs. As described below, the high pressure helps provide the large height gradients and thus strong onshore flow required for major events. Terrain enhancement can be seen in the correlation plots, such as over the Cascades and the Olympic Mountains in the Forks and Astoria maps (Figs. 2a,b). Some areas of low correlation, such as over Puget Sound for heavy precipitation events at Forks, are associated with large variability of precipitation during extreme events due to mesoscale variations in rain shadowing. Other areas, such as eastern Washington, are nearly always dry and thus show relatively low correlations. The swaths of high correlation are relatively narrow, ~200 km wide, consistent with the narrow nature of ARs described in the literature (e.g., Ralph et al. 2004). The structure of the spatial correlations between coastal time series of precipitation with gridded analyses for extreme precipitation events (Fig. 2) is noticeably different than the correlations for all precipitation intensities (Fig. 3). Considering all precipitation events, there are large areas of high positive correlation surrounding each station with very little structure, and no adjacent zones of negative correlation. This suggests that more typical precipitation events have a broader structure than extreme events (such as those associated with ARs) or are associated with a range of possible evolutions.
The amount of two-day coastal precipitation varies substantially among the top 50 events (Table 3). The two-day precipitation for the number 1 ranked storm ranged from 31.0 cm (12.21 in.) at Forks, WA, to 16.6 cm (6.55 in.) at Astoria, OR. The number 50 ranked storm ranged from 15.8 cm (6.23 in.) at Forks, WA, to 7.72 cm (3.04 in.) at Eureka, CA. The precipitation gauges at Forks, North Bend, Brookings, and Eureka have relatively high terrain to the east, resulting in increased precipitation from topographic enhancement for the most extreme events. In contrast, Astoria and Newport have lower terrain nearby and thus generally smaller two-day totals for the top ranked storms. For the number 50 ranked storms, the two-day totals were highest at the stations upwind of the highest terrain (Forks and Brookings).
Searching for the top 50 events in two-day precipitation at each coastal station produced a total of 207 events (several appear in the top 50 for multiple stations). The majority of these events occurred in November, December, and January (Fig. 4a), similar to the results of Lackmann and Gyakum (1999) and Neiman et al. (2008b, 2011). Examining decadal variability, the 1960s, 1970s, and 2000s experienced relatively few extreme precipitation events, while the 1950s, 1980s, and 1990s received more (Fig. 4b). Similar decadal variability was found using extreme one- and three-day precipitation totals (not shown). This decadal modulation is consistent with previous studies that found strong Pacific and Atlantic storm tracks in the 1950s, 1980s, and 1990s, and relatively weaker storm tracks during the 1960s and 1970s (Chang et al. 2002; Chang and Fu 2002).
While daily precipitation is useful, it lacks sufficient temporal resolution to evaluate the detailed evolution of extreme precipitation events. At two of the above stations, Astoria and Brookings, hourly precipitation data were available from NCDC for most of 1950–2009 and were used to examine the evolution and durations of the top 50 storms defined by the two-day totals. In the Pacific Northwest during winter months, it is common for some locations to receive persistent rain for days or weeks without flooding. It is therefore necessary to separate the extreme events from the background precipitation in order to examine storm duration and evolution. To ensure that we included the entire storm for each of the top 50 events, two days were added before and after the two-day periods noted above to create a six-day examination period for each event.
Table 4 shows the average periods associated with various percentages of the six-day storm totals at Astoria and Brookings. The results at both stations are very similar, with 50%–95% of the storm total precipitation occurring over periods of roughly 20 to 90 h. Arbitrarily choosing 75% of the six-day total as the definition of storm duration, the average storm length was 44 and 45 h at Astoria and Brookings, respectively. The shortest and longest “storms” were 16 (17) and 86 (82) hours at Astoria (Brookings) by this 75% definition. Plotting a histogram of storm lengths at these two locations using the 75% criterion (Fig. 5) indicates considerable variability in the storm periods, with little evidence of a dominant mode. Thus, although the time of concentration for most Northwest drainages is less than 24 h, a period similar to the shortest storms noted in Table 4, many storms had significant precipitation over one to three days. Examining individual storms, there is considerable variability in their temporal evolutions (Fig. 6). Some storms experience heavy rain in one spike, some in several spikes. Others receive moderate rain for many hours, without large spikes.
3. Synoptic structure and evolution
Although some studies have examined composites of the synoptic flow associated with heavy precipitation affecting the West Coast (Lackmann and Gyakum 1999; Neiman et al. 2008b, 2011), few have examined how the synoptic evolution associated with such events varies along the coast. Some insight into this issue is found in Neiman et al. (2008b), who looked at the synoptic conditions associated with major AR events for two large spatial areas: the entire Pacific Northwest (including southern BC) and California. The study presented below examines how synoptic evolution associated with heavy precipitation events varies progressively along the coast using six stations from northern Washington to northern California, without the assumption that all events must be associated with ARs. Furthermore, we have examined each event individually to determine the range of synoptic evolutions associated with major precipitation events.
The NCEP–NCAR 6-h reanalysis data (Kalnay et al. 1996) for precipitable water, 500- and 700-hPa geopotential heights, sea level pressure, and 850-hPa temperature were used to composite the synoptic evolution of the top 50 two-day precipitation events (NCEP reanalysis–derived data are provided by the NOAA/OAR/ESRL PSD, Boulder, Colorado, from their website at http://www.esrl.noaa.gov/psd/). Each event is compared to the 60-yr climatology for the corresponding days and the anomalies from climatology for each event were averaged to create an anomaly composite. Assuming the anomalies have an approximately normal distribution, the Student’s t statistic is used to identify areas that are statistically different from climatology at the 95% level. Since the extreme precipitation dates were based on a two-day cumulative total, it was necessary to identify a “0” time (t0) for the composites. If one of the two one-day totals was more than double the other, then the grid time closest to the middle of that day was used. If not, the grid time between the two precipitation days was used. Composites of synoptic variables were made every 6 h for three days before and after t0 for each of the six stations. Because of similarities among adjacent stations, composites from only three of the six stations, Forks, Newport, and Eureka, are shown.
At the time of maximum precipitation the 500-hPa (not shown) and 700-hPa composite geopotential height anomalies (Fig. 7) are negative to the northwest (troughing) and positive to the southeast (ridging) of each station, with the anomalies shifting progressively with station position. The negative anomalies are a factor of 2 greater than the positive anomalies at all stations except North Bend and Eureka (Fig. 7c), where the factor increases to 3 to 4. At Forks, the negative anomaly extends farther west than those of the other stations. The sea level pressure anomalies (not shown) possess similar features to these upper-level patterns. At all stations, the flow at all levels is generally southwesterly, although the flow at Forks contains more of a southerly component than the other stations.
The 850-hPa temperature anomalies along the coast and the near-shore waters are large, positive, and spatially extensive for Forks (Fig. 8a), reaching approximately +6°C at the time of heaviest precipitation. For stations to the south, the positive anomalies substantially weaken (to 2°–3°C), shift southward, and are displaced mainly equatorward of the locations in question (Figs. 8b,c). At all stations, significant negative anomalies are seen to the northwest of the observing site, with the magnitude of these anomalies increasing for the southern locations.
At the time of heaviest coastal precipitation there are much higher than normal precipitable water values extending southwest off the coast and inland to the northeast from each station (Fig. 9). The largest positive anomalies (10–15 mm) are located southwest of each station and suggest an AR signature (Lackmann and Gyakum 1999; Ralph et al. 2004, 2006; Junker et al. 2008; Neiman et al. 2008b). At all stations, there are significant negative anomalies to the north and south of the water vapor plume. There is a progressive southward shift and widening of the positive anomalies for stations southward down the coast.
Individually, the precipitable water and height/wind fields show large anomalies with coherent structures: this is reflected in the combination of the two, namely the integrated water vapor flux (IWVF). Here, IWVF is defined as , vertically integrated to 300 mb (Neiman et al. 2008b), where is mean specific humidity between adjacent layers, is the mean magnitude of the horizontal wind between adjacent pressure levels, dp is the pressure difference, and g is acceleration due to gravity. At each grid point, these values were summed from 1000 to 300 hPa to yield IWVF in kg m−1 s−1. At all of the stations, anomalously high IWVF is seen extending southwest of the station into the subtropical central Pacific (Fig. 10). The positive precipitable water anomalies turn westward and pass immediately north of Hawaii. The 850-hPa wind vector anomalies indicate enhanced southwesterly flow in the middle of the flux maxima, as well as enhanced northerlies over the Gulf of Alaska. There is confluence of the 850-hPa winds along the entire length of the IWVF maxima.
To examine the evolution of the synoptic fields associated with heavy precipitation events, anomaly composites for several variables were made for each station every 6 h for three days before and after t0. In the interest of space, only composite evolutions at Forks and Eureka, the stations farthest north and south, are shown (Figs. 11–14 ). At 500 hPa, there are notable differences in the evolution of the geopotential height anomalies between the two stations, although both start with a region of negative anomalies stretching from the central Pacific to the northwest United States, British Columbia, and southeast Alaska (Fig. 11). Forks begins with a larger negative height anomaly over the central Pacific Ocean and a smaller one near the coast. Two days prior to the heaviest precipitation, Forks has a far deeper anomaly over the central Pacific, while Eureka has an extensive weak positive anomaly over the south portion of the domain. At one day prior, the negative anomaly for Forks is deepest to the south of the Gulf of Alaska and is displaced considerably westward of the negative anomaly at Eureka; at the same time Forks has developed a modest positive anomaly over the West Coast. At t0, the negative anomalies are deepest, the Forks anomaly is located in the Gulf of Alaska (Fig. 11d), and the Eureka anomaly is found offshore near Vancouver Island (Fig. 11h). The positive anomaly is larger at Forks and covers the entire western United States, whereas the positive anomaly for Eureka is located over the southwest United States. In summary, the negative height anomalies at 500 hPa for Eureka were nearly stationary and increased in amplitude in time as positive anomalies developed over the southwest. In contrast, the synoptic evolution at Forks was far more dynamic, with dual negative maxima; the one over the central Pacific moved eastward and amplified into the dominant feature as positive anomalies developed in place over the Southwest.
At the surface, the sea level pressure evolution is very similar to the upper-level development (Fig. 12). However, at the surface the central Pacific negative anomaly was even more dominant at Forks, and the positive anomaly was relatively weaker. Thus, at Forks there is a strong suggestion of the importance of a westward-moving feature in heavy precipitation events, while at Eureka the amplification of a preexisting trough appears more significant.
The 850-hPa temperature evolutions also show significant differences among the stations (Fig. 13). At the two northern stations, Forks and Astoria, there are weak (1°–2°C) positive 850-hPa temperature anomalies over the central Pacific at −72 h that amplify and move to the West Coast during the next three days. In contrast, at Eureka there is little evidence of the eastward movement of positive anomalies over the Pacific; rather, a modest positive anomaly (far weaker than the one for Forks and Astoria) develops during the last day of the event. All stations had an offshore negative temperature anomaly during the period of heaviest precipitation; for Forks there was some suggestion of the eastward movement of the negative anomaly, while for Eureka the negative anomaly developed in place. It is interesting that the positive temperature anomalies near the time of the event are far more widespread than the precipitable water or integrated water vapor flux anomalies (Figs. 9 and 10).
The evolutions of the precipitable water composites reflect the varying developments between the northern and southern coastal locations (Fig. 14). At Forks, an amorphous and weak area of positive anomalies strengthens and moves westward during the three days. During that period, the positive anomaly becomes narrower and more focused, with negative anomalies developing and moving eastward to the north and south. For Eureka, the pattern appears to develop in place as a core of positive anomalies and adjacent negative anomalies increase in time. Composite integrated water vapor flux anomalies generally mirror the precipitable water anomaly fields and thus are not shown.
4. Alternate synoptic evolutions
a. Non-AR extreme precipitation
The composites of the top 50 storms for each of the six stations and the inspection of each individually suggest that most are associated with extensive water vapor transport from the subtropics and tropics that typify ARs. However, there is a small subset of heavy coastal precipitation events that rank among the top 50 that are not associated with ARs. These events occur only at three of the southern stations: Newport, North Bend, and Eureka. The infrequent occurrence of non-AR storms means that they are not represented in the synoptic composites; thus, it is necessary to take a closer look at these storms individually.
One such event occurred on 13–15 March 1967 when 10.97 cm (4.32 in.) of rain fell at Newport (the 27th greatest 48-h wintertime precipitation total at that location). From 0600 UTC 13 March to 0600 UTC 15 March, the synoptic pattern included a stationary high-amplitude, shortwave trough offshore at 700 hPa (Fig. 15a) and 500 hPa (not shown), and an offshore surface low (Fig. 15b) that deepened substantially in the first 24 h and weakened during the subsequent day (not shown). In addition, there were negative 850-hPa temperature anomalies over the entire coast (Fig. 15c) and negative precipitable water anomalies along the Oregon coast (Fig. 15d).
This case and the four other non-AR heavy precipitation events shared similar features; the precipitable water anomalies and 850-hPa temperature anomalies were negative along the entire coast and all possessed low-pressure areas northwest of the station that were nearly stationary during the 48 h of heavy precipitation. For each of these events, heavy precipitation began after an initial frontal passage; was associated with convection in cooler, unstable, postfrontal air; and included the interaction of this flow with coastal terrain (Kreitzberg and Brown 1970; Hobbs et al. 1975; Pike 1987). The mean static stability dθ/dz (Figs. 16a,b) over the Pacific Northwest for the three non-AR events that occurred at North Bend, OR (the other two occurred at Newport and Eureka), was considerably less than for the remaining 47 AR-associated events. For the five non-AR extreme precipitation events, much of the higher-elevation precipitation fell as snow, since temperatures were generally about 5°–8°C cooler at 850 hPa for these events compared to those associated with ARs (Figs. 16c,d). According to U.S. Department of Commerce Storm Data, none of the above non-AR events was associated with flooding.4 In fact, one of these storms, 8 January 2005, a top-50 event at Eureka, CA, triggered a heavy snow advisory for the mountainous terrain above Eureka.5 Open-cellular convection can clearly be seen in the infrared satellite image for this event (Fig. 17), a sign of postfrontal convection, instability, and cooler temperatures.
b. Atmospheric rivers without extreme precipitation
Most wintertime extreme precipitation events along the Pacific Northwest coast are associated with AR conditions, but not all ARs result in extreme precipitation near the coast. Since the majority of atmospheric water vapor during AR events is found in the lowest ~2.5–3.0 km of the atmosphere (Browning and Harrold 1970; Browning and Pardoe 1973; Browning et al. 1974; Peixoto and Oort 1992, 278–285; Ralph et al. 2005), it is important to understand the climatological annual cycle of low-level flow and precipitable water on the synoptic scale. Figure 18 shows 60-yr seasonal means of precipitable water and 850-hPa geopotential heights from the NCEP reanalysis. During the winter months [January–March (JFM)], the precipitable water values in the Pacific basin are lowest and increase toward the south, and the westerly 850-hPa flow is at its peak (Fig. 18a). During the spring [April–June (AMJ)], lower-tropospheric high pressure develops over the south-central Pacific and the counterclockwise flow near the surface advects water vapor northward and eastward out of the tropical western Pacific (Fig. 18b). In the summer [July–September (JAS)], the low-level air temperature is warm, the average precipitable water values in the Pacific basin are highest, the Pacific high extends farthest north, low-level westerly flow approaching the northwest is weak, and the mean circulation moves high precipitable water values north and east (Fig. 18c). In the autumn and early winter [October–December (OND)], the northeast Pacific high pressure area recedes, the mean maximum low-level flow increases and moves southward, and the mean precipitable water values decrease (Fig. 18d). Interestingly, the highest values of precipitable water (>35 mm) along the West Coast are found during the warm season when West Coast precipitation is smallest [also noted in Neiman et al. (2008b)]. In contrast, the majority of extreme precipitation events along the West Coast occur in November, December, and January, when the mean precipitable water values offshore are much lower. Consistent with the above climatology, most summertime atmospheric rivers approach the West Coast from the west, in contrast to the southwesterly origin during the cooler months. Neiman et al. (2008b) discussed similar seasonal differences as they pertained to atmospheric rivers using satellite-observed integrated water vapor.
To find days of high precipitable water values but modest precipitation, 6-h NCEP reanalysis grids during the cool season (October–March) of 1950–2009 were examined for precipitable water values >35 mm at each of three meridians (125°, 130°, and 135°W between 37.5° and 52.5°N) located off of the West Coast of the United States (Fig. 19). The CPC unified precipitation analysis was used to find daily precipitation totals at every marked grid point (crosses) along the coast in which there was less than 5 cm of daily precipitation.6 Only six cases were found during 1950–2009 with such high precipitable water values and relatively modest precipitation. Of these, all occurred in October (2), November (3), or December (1). During the same 60-yr period, 34 days with precipitable water values greater than 35 mm and less than 5 cm of rain along the coast occurred during the warm season (April–September). All but two occurred in June (7), July (8), August (5), and September (12). Clearly, high precipitable water values alone are not sufficient for producing extreme precipitation at coastal sites in the Pacific Northwest. During the summer, a number of elements work against significant precipitation. The climatological subtropical high results in a highly stable middle troposphere, often with an inversion capping a relatively shallow marine layer. Above the inversion the air is generally quite dry. Low-level flow is far weaker than during winter, resulting in less orographic uplift and precipitation. Finally, with the summertime configuration of a weakened jet stream and associated disturbances displaced to the north, there is little dynamical support for significant synoptic vertical motions, as well as a weak horizontal water vapor flux. The result is little or no precipitation even when vertically integrated water vapor amounts are high. This seasonal character of water vapor and low-level flow as it relates to atmospheric rivers and heavy precipitation along the West Coast was also discussed in Neiman et al. (2008b), where, using satellite-observed IWV, they noted large numbers of AR plumes in the summer months along the West Coast despite those months being the climatologically driest in that region.7
An example of an AR that produced little coastal precipitation occurred on 10 September 2000 (Fig. 20). The sea level pressure field is dominated by expansive subtropical high pressure that stretches from north of Hawaii to as far as 50°N and spans much of the eastern half of the North Pacific Ocean. The water vapor plume, which originates in the western tropical Pacific, is driven north by the Pacific high and upper-level flow field, extending across the Gulf of Alaska toward the Pacific Northwest. At the time of the highest precipitable water anomalies along the coast between Forks, WA, and Eureka, CA, and for three days before and after, virtually no rain fell throughout the Pacific Northwest. In this case and all other summer AR cases, upper-level ridging, weak onshore flow, or a combination of both, limited upward vertical velocities and precipitation at the coast.
The Pacific Northwest often experiences extreme precipitation that brings major flooding, landslides, property damage, and loss of life. It is clear from this study and others that atmospheric rivers are responsible for a large majority of extreme events. There is no significant trend in these events over the 60-yr period included in this study (Mass et al. 2011) but there is large interannual and interdecadal variability. This study examines the duration of the region’s extreme precipitation events, the associated synoptic structures, and how their synoptic evolutions differ at various locations along the coast. Aside from a very small subset of events, all wintertime ARs intersecting the coast cause large amounts of precipitation along the coast and nearby inland areas. Although composites of these events at various locations along the coast share similar synoptic features, there are significant differences in the synoptic evolutions leading to these events. Furthermore, there are nonconventional situations in which extreme precipitation occurs without high values of precipitable water or when high water vapor values do not produce extreme precipitation.
The top 50 extreme precipitation events in two-day precipitation were identified from 60 years of NCDC daily precipitation data at six Pacific Northwest coastal stations from northern Washington State to northern California: Forks, Astoria, Newport, North Bend, Brookings, and Eureka (Fig. 1). Of the 207 dates identified by the top 50 storms at each station, the overwhelming majority occurred during November, December, and January. The 1950s, 1980s, and 1990s experienced more events than the 1960s, 1970s, and 2000s, consistent with previous studies.
NCDC hourly precipitation data at Astoria and Brookings, the two stations with the most complete 60-yr record, were analyzed to determine storm length and the temporal variability of extreme coastal precipitation events. For each location, the associated periods of heavy precipitation ranged from 16 to 86 h, with large storm to storm variability. Many of the storms exhibited most of the precipitation in one large spike, while others showed multiple spikes or relatively steady precipitation over a prolonged period of time.
The 60-yr daily precipitation time series for winter months at each station was correlated with the daily average stream flow at nearby unregulated rivers for various time lags. Correlations for calendar day precipitation were relatively low, most likely due to large events spanning a calendar day, whereas correlations for two, three, and four days were higher. The correlations were highest at zero- and one-day lags and dropped off considerably for longer lags, suggesting that precipitation more than 48 h prior had already passed through the watersheds. These results imply that 48 h is a relevant time period for examining heavy precipitation events. The use of a two-day storm definition eliminated the calendar day problem associated with daily precipitation observations and was consistent with time of concentration estimates of less than one day for Pacific Northwest rivers.
Spatial correlations of precipitation show that, on average, precipitation associated with the top 50 two-day precipitation events at the coastal stations used in this study did not just affect the coast but also fell inland along an axis oriented southwest to northeast, with adjacent areas of suppressed precipitation (negative correlations). The width of the enhanced precipitation associated with extreme events was on the order of 200–300 km. Spatial correlations of climatological precipitation did not suggest these structures, indicating more amorphous, larger-scale features and differing precipitation evolutions for nonsevere events.
Synoptic composites were made for 500- and 700-hPa geopotential heights, sea level pressure, 850-hPa temperatures, precipitable water, and integrated water vapor flux using the top 50 events for each station at the time of maximum precipitation t0. The composites at t0 exhibit classic AR signatures and are generally consistent with previous studies (Lackmann and Gyakum 1999; Ralph et al. 2004, 2006; Junker et al. 2008; Neiman et al. 2008b). However, there are differences in the synoptic structures among stations separated by only a few hundred kilometers. The 500- and 700-hPa geopotential height fields showed a deep trough offshore and ridge axis just east of the West Coast for all stations, but at Forks, the trough stretched deep into the subtropics and the upper-level flow was more south-southwesterly over the coast. For stations south of Forks, the trough did not stretch as far to the south and the upper-level flow was southwesterly with the ridge axis farther inland than for Forks. Additionally, positive temperature anomalies were greater at stations to the north. The precipitable water anomaly composites for each station showed a large positive anomaly oriented southwest to northeast across the station (consistent with spatial correlations in precipitation). The composite precipitable water plume changed from longer, narrower, and originating closer to Hawaii for stations to the north, to shorter, wider, and originating east of Hawaii for stations farther south. Integrated water vapor flux anomaly composites closely mirror those of precipitable water and show the extension of the anomalous flux into the subtropical Pacific.
Differences in anomaly composites were not only noted for t0 but also for the entire evolution starting three days prior to t0. Negative 500- and 700-hPa geopotential height and sea level pressure anomalies extended into the central Pacific 72 h before the event only for the northernmost station, Forks, where the anomalies deepened and moved eastward in subsequent days. All other stations had negative height and sea level pressure anomalies centered closer to the coast that were relatively stationary in the days prior to the event. Positive temperature anomalies covered the entire coast for the northern coastal stations but were smaller and weaker along the coast for composites based on the southern sites. The precipitable water composite evolution showed anomalous water vapor in the central Pacific at three days prior to the event for all stations. At Forks, the water vapor anomaly consistently strengthened and propagated toward the West Coast in the days leading up to the event and remained relatively stationary in the final 24 h. At all other stations, the precipitable water anomaly initially weakened as it propagated west and then strengthened at the coast closer to t0. At two to three days prior to t0, the upper-level trough in the central Pacific was deeper and extended farther south into the subtropics and tropics for events that impacted Forks. As noted above, this evolution tends to produce a more meridional trough that supports more southerly flow that enhances water vapor advection as far north as Washington State. A more southerly flow would also enhance precipitation at Forks, which is surrounded by more south- or southwest-oriented slopes. At the stations to the south, the terrain is oriented north–south and the low-level flow during extreme precipitation events has a more westerly component.
Even though ARs are responsible for the majority of wintertime extreme precipitation events along the West Coast, this study shows there are other ways for extreme precipitation to occur. At three of the southern stations, Newport, North Bend, and Eureka, extreme precipitation sometimes occurred with anomalously low values of precipitable water. These types of storms occurred five times in the 207 events identified in this study. In each case, the precipitable water values near the station were less than 20 mm, a high-amplitude shortwave trough developed immediately offshore, and negative temperature anomalies covered the entire West Coast. These few events were not represented in the composites due to the dominance of AR-related events and were associated with postfrontal convection in cool, unstable air. The cooler temperatures meant that the majority of precipitation that fell at higher elevations was likely to be snow, minimizing possible flooding.
Anomalously high water vapor values alone are not sufficient to produce heavy precipitation. The highest values in precipitable water near the Pacific Northwest coast occur in the summer (Neiman et al. 2008b) when ocean surface temperatures are relatively warm, the atmosphere is stable, and the winds are weak. This study found only six examples in the cool season (October–March) in which precipitable water values exceeded 35 mm near the northwest U.S. coast and did not produce 24-h precipitation totals greater than five centimeters. In contrast, 34 occurrences of high precipitable water/low precipitation occurred during the summer when the jet stream is weaker or displaced much farther north and the integrated water vapor flux is small.
Many questions remain about heavy precipitation along the West Coast. What accounts for the differences in composite evolutions found in this study? To what degree do they reflect the varying topography at each location or the differing evolutions required to bring sufficient water vapor and onshore flow to locations at various distance from the subtropical/tropical sources of water vapor? How do variations in wind direction, wind speed, and stability of the low-level flow approaching the coast alter the precipitation magnitudes and spatial distributions? What mechanisms control the frequency and intensity of ARs on interannual and interdecadal time scales? Future work will be aimed at answering these questions and understanding how extreme precipitation will be affected by a changing climate.
This research was supported by the National Science Foundation (Grants AGS0709856 and AGS1041879). We wish to thank Neal Johnson for securing Astoria and Brookings NCDC hourly precipitation data, and David Carey for web development assistance. Dr. Dennis Lettenmeier and Mr. Larry Schick provided important insights regarding Pacific Northwest river hydrology, and Dr. Paul J. Neiman and two anonymous reviewers contributed important suggestions that greatly improved this manuscript.
Time of concentration was very roughly estimated by dividing the square root of the area of the drainage basin (an integer multiple of basin linear distance) by a channel flow of ~1 m s−1. (D. Lettenmeier, University of Washington, Department of Civil Engineering, 2011, personal communication).
The last day of the multiday periods was used as the value for correlations.
The reporting times for precipitation vary by station.
Two storms occurred in 1950, prior to the first publication of Storm Reports in 1959. Data come from Storm Data 1967, Vol. 9, and Storm Data 1971, Vol. 13.
Storm Data 2005, Vol. 47.
CPC analysis was used instead of observation data to cover the entire coast and grid points were chosen based on proximity to the coast and to avoid higher terrain.