ISSN: (Print) (Online) Journal homepage:

Atmosphere-Ocean ISSN: 0705-5900 (Print) 1480-9214 (Online) Journal homepage: http://www.tandfonline.com/loi/tato20 Interaction of an Intense Pacifi...
Author: Mildred Walton
4 downloads 0 Views 2MB Size
Atmosphere-Ocean

ISSN: 0705-5900 (Print) 1480-9214 (Online) Journal homepage: http://www.tandfonline.com/loi/tato20

Interaction of an Intense Pacific Low Pressure System with a Strong Arctic Outbreak over British Columbia: Forecast Challenges of the Early December 2007 Storm Quanzhen Geng , Ruping Mo , Mindy Brugman , Brad Snyder , Jim Goosen & Greg Pearce To cite this article: Quanzhen Geng , Ruping Mo , Mindy Brugman , Brad Snyder , Jim Goosen & Greg Pearce (2012) Interaction of an Intense Pacific Low Pressure System with a Strong Arctic Outbreak over British Columbia: Forecast Challenges of the Early December 2007 Storm, Atmosphere-Ocean, 50:1, 95-108, DOI: 10.1080/07055900.2012.656261 To link to this article: http://dx.doi.org/10.1080/07055900.2012.656261

Published online: 03 Feb 2012.

Submit your article to this journal

Article views: 141

View related articles

Citing articles: 3 View citing articles

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=tato20 Download by: [37.44.207.107]

Date: 14 January 2017, At: 21:21

Interaction of an Intense Pacific Low Pressure System with a Strong Arctic Outbreak over British Columbia: Forecast Challenges of the Early December 2007 Storm Quanzhen Geng1,*, Ruping Mo2, Mindy Brugman2, Brad Snyder1, Jim Goosen1 and Greg Pearce1 1 2

Pacific Storm Prediction Centre, Environment Canada, Vancouver, British Columbia, Canada National Laboratory for Coastal and Mountain Meteorology, Environment Canada, Vancouver, British Columbia, Canada [Original manuscript received 15 February 2011; accepted 01 December 2011]

The interaction of a warm moist air mass from an intense Pacific low pressure system with a cold air mass from a strong Arctic outbreak during 1–5 December 2007 produced a record number of high-impact weather events across British Columbia, including heavy snow, freezing rain, heavy rain, strong winds and extreme wind chill. The unusual concurrence of these two strong weather systems caused many forecast challenges for both the Canadian numerical weather prediction (NWP) models and meteorologists at the Pacific Storm Prediction Centre (PSPC) of Environment Canada. In this study, the evolution of the weather systems and the observed severe weather events during the 1–5 December 2007 storm are analyzed. Weather forecasts by the NWP models and PSPC meteorologists are compared with the observed high-impact weather events. It is shown that the Canadian NWP models forecast the storm reasonably well. Meteorologists at PSPC further improved the model forecasts by considering various local effects of the complex terrain over British Columbia that the model has difficulty resolving and model biases caused by inadequacies in its boundary layer parameterizations.

ABSTRACT

[Traduit par la rédaction] L’interaction d’une masse d’air chaud et humide issue d’une forte dépression dans le Pacifique avec une masse d’air froid générée par une forte coulée polaire entre le 1er et le 5 décembre 2007 a engendré un nombre record d’événements météorologiques à fort impact en Colombie-Britannique : neige abondante, pluie verglaçante, pluie forte, vents forts et indice extrêmement élevé de refroidissement éolien. Pour les modèles canadiens de prévision numérique du temps (NWP) et les météorologues du Centre de prévision des tempêtes du Pacifique (PSPC) d’Environnement Canada, la coïncidence inhabituelle de ces deux systèmes météorologiques forts a posé de nombreux défis. Dans la présente étude, nous analysons l’évolution des systèmes météorologiques et les intenses événements météorologiques observés au cours de la tempête survenue entre le 1er et le 5 décembre 2007. Nous comparons les prévisions de la météo établies à partir des résultats des modèles NWP et par les météorologues du PSPC avec les événements météorologiques à fort impact qui ont été observés. Nous démontrons que les modèles canadiens de NWP, ont prédit raisonnablement bien l’évolution de la tempête. Les météorologues au PSPC ont rajusté à leur tour les prévisions établies par ces modèles en intégrant les divers effets locaux des terrains accidentés en Colombie-Britannique, qui sont difficiles à résoudre par ces modèles, et leurs biais attribuables au manque de précision des paramétrages de la couche limite. RÉSUMÉ

KEYWORDS Pineapple Express system; Arctic outbreak; high impact weather; local effects of complex terrain; forecast challenges

1 Introduction In early December 2007, a strong Pacific low pressure system invaded British Columbia (BC) Canada. The warm moist air mass from this system, in combination with the cold continental air mass associated with a strong Arctic outbreak, produced a record number of almost all types of severe weather. The storm resulted in extensive flooding, mudslides and avalanches in BC. Power outages were reported across most of

*

the province; some highways and schools were closed, and several communities were isolated because of heavy snow, freezing rain and flooding. A traffic accident in heavy snow near Prince George claimed five lives (A. McCarthy, J. Steele, D. Jones and D. Lundquist, personal communication, 2007). The snowpack, weakened by the sharp rise in temperatures and heavy rain from the storm, caused avalanches and

Corresponding author’s email: [email protected] ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

96 / Quanzhen Geng et al. increased the potential for avalanches. The long-lasting effect on the snowpack over some BC mountain ranges is believed to be the major cause of 9 of the 11 avalanche fatalities from 8 December 2007 to 16 January 2008 (Klassen, 2008). The storm also brought severe damage to the coasts of Oregon and Washington, United States, with some regions receiving the strongest gales since 1962 (Read, 2007; Reiter, 2008). These two very strong weather systems and their interaction with the complex terrain of BC created huge forecast challenges for both the numerical weather prediction (NWP) models and operational meteorologists. Figure 1 shows the complex terrain and public weather forecast regions of BC. Located east of the Pacific Ocean, the bulk of the province is mountainous. The most significant mountain ranges are the Coast Mountains to the west, the Columbia Mountains in the centre and the Rocky Mountains to the east. The Interior Plateau is an area of extensive high-elevation flat terrain characterized by ragged hills and deep river gorges. Often during winter, an area of high pressure forms in very cold air over Alaska and northern Canada. If this cold air builds west of the Continental Divide, it has an easy path southwards into the northern and central interior of BC. At

least once or twice each year, the advance of the Arctic air is so strong that it spreads into the southern interior and flows through the mountain valleys to the coastal regions (Jackson and Steyn, 1994; Johnson and Mullock, 1996). This is usually called an Arctic outbreak. In this situation, strong northerly Arctic outflow and extreme wind chill are expected to develop over parts of the BC interior and coastal valleys and inlets. On the other hand, numerous marine storms are generated in the North Pacific each winter. Some of them follow the Pacific storm track towards BC, producing mild, wet and windy conditions along the coast (Lange, 1999; Geng and Sugi, 2003). Occasionally, each winter, the atmospheric circulation acts to maintain a long fetch of very warm and moist air stretching from the subtropics near the Hawaiian Islands to the west coast of North America (e.g., Fig. 2); this system is frequently called the “Pineapple Express” (hereafter PE; Heidorn, 2004; Mass, 2008). Associated with the system, heavy precipitation and flooding will often occur when the warm moist air is forced to rise over the coastal mountain ranges (Lyons, 1901; Loukas and Quick, 1996; Lackmann and Gyakum, 1999; Colle and Mass, 2000; Galewsky and Sobel, 2005; Knippertz and Martin, 2007; Ahrens et al.,

Fig. 1 Physiographic and public weather forecast regions of BC. The smaller forecast regions bounded by blue lines are grouped by red circles into seven large regions. The red dots with letter identifications are weather stations mentioned in the paper. Major mountain ranges are grouped with white circles; the Coast Mountains are to the left, the Rocky Mountains to the right, and the Columbia Mountains in the middle right bottom. The names of the forecast regions are (1) South Coast: 1.1 Metro Vancouver, 1.2 Greater Victoria, 1.3 Fraser Valley, 1.4 Howe Sound, 1.5 Whistler, 1.6 Sunshine Coast, 1.7 Southern Gulf Islands, 1.8 East Vancouver Island, 1.9 West Vancouver Island, 1.10 Inland Vancouver Island; (2) North Coast: 2.1 North Vancouver Island, 2.2 Central Coast – Coastal, 2.3 Central Coast – Inland, 2.4 North Coast – Coastal, 2.5 North Coast – Inland, 2.6 Queen Charlottes; (3) Southwest Interior: 3.1 Okanagan Valley, 3.2 Similkameen, 3.3 Fraser Canyon, 3.4 Nicola, 3.5 South Thompson, 3.6 Shuswap; (4) Kootenay District: 4.1 Boundary, 4.2 Arrow/Slocan Lakes, 4.3 West Kootenay, 4.4 Kootenay Lake, 4.5 East Kootenay, 4.6 Elk Valley; (5) Columbia District: 5.1 West Columbia, 5.2 East Columbia, 5.3 North Columbia, 5.4 Kinbasket, 5.5 Yoho/Kootenay Park, 5.6 North Thompson; (6) Central Interior: 6.1 100 Mile, 6.2 Chilcotin, 6.3 Cariboo, 6.4 Prince George, 6.5 Yellowhead, 6.6 McGregor, 6.7 Williston, 6.8 Bulkley Valley & the Lakes; and (7) BC Peace River: 7.1 BC Peace River. ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 97

Fig. 2 The 1200 UTC 3 December 2007 mean sea level pressure (contour interval 4 hPa, solid lines), precipitable water (colour shading), and 700 hPa wind barbs (knots), from the North American Regional Reanalysis. The frontal system over the North Pacific is drawn to illustrate the “Pineapple Express” phenomenon. Note that Hawaii (where the “pineapple” reference originates) is located near the southwest end of the cold front.

2012). This has been seen in many locales along the coast from California to BC. Although intense PE storms and strong Arctic outbreaks usually occur once or twice each winter over BC (Johnson and Mullock, 1996), the probability of experiencing both scenarios at the same time is extremely low, probably only once or twice in a decade (see Section 3d). If this does occur (i.e., a strong PE system meeting an Arctic outbreak over the complex terrain of BC) the warm moist Pacific air mass will override the cold, dry Arctic air mass trapped within the mountain valleys and low-lying land areas, producing various kinds of severe winter weather, such as heavy snow, freezing rain, heavy rain and strong winds. Usually the precipitation will fall first as heavy snow when the PE meets the cold air mass from the Arctic outbreak. The snow will then change to freezing rain or a mixure of snow and rain, as the warm moist air from the PE system continues to build in the middle and upper levels, while the cold Arctic air mass still lingers in the mountain valleys and low-lying land areas. The cold air mass will eventually be scoured out by the approaching warm air and the precipitation will change to heavy rain. At the same time, the strong northerly Arctic outflow winds will quickly switch to strong southerly winds from the Pacific low pressure system. This is exactly what transpired in early December 2007: a rare event that happened only once during the last decade. In this paper the evolution of the weather systems and the observed severe weather events during the 1–5 December

2007 storm are analyzed. Public weather forecasts by the Global Environmental Multiscale (GEM) model and meteorologists at the Pacific Storm Prediction Centre (PSPC) are compared with the observed high-impact weather events. From an operational meteorologist’s point of view, we would like to highlight the strengths and limitations of the GEM model in handling such high-impact weather events in BC. We will demonstrate that meteorologists at PSPC significantly improved the model forecasts by considering local effects of the complex terrain that the GEM model cannot fully resolve and model biases caused by inadequacies in its boundary layer parameterizations. A description of the data used in this study is given in Section 2. The evolution of the weather systems and the associated high-impact weather events are described in Section 3. The performance of the GEM model and the PSPC meteorologists in forecasting the severe weather events are examined in Sections 4 and 5. A summary and conclusions are provided in Section 6.

2 Data Operational analysis and forecast products of the GEM model from the Canadian Meteorological Centre were used to analyze the December 2007 storm. The GEM Global (GEMG) model has a horizontal resolution of approximately 33 km at 49°N. The version of the GEM Regional (GEM-R) model

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

98 / Quanzhen Geng et al. running operationally at the time of this event was also a global model but with a higher horizontal resolution (about 15 km) over most of Canada (Côté et al., 1998; Mailhot et al., 2006). In this study, short-term forecast products (0 to 48 hours) are from the GEM-R model, and the extended forecast products beyond 48 hours are from the GEM-G model. Some of the model forecasts used in this study, such as temperatures, winds, clouds and probability of precipitation, are outputs from the Updateable Model Output Statistics (UMOS; see Wilson and Vallée, 2002). There are no UMOS products for precipitation amounts and precipitation types. The forecast precipitation amounts are direct model outputs. Precipitation types are direct outputs from a diagnostic routine in the model mainly dependent on the model forecast temperature profiles (Bourgouin, 2000). To study the extent and severity of the December 2007 storm and objectively verify the severe weather warnings, a dataset of weather warning events with quantitative severity was created for each forecast region for each weather element (i.e., wind chills, winds, snowfall, freezing rain and rainfall). Weather observations were obtained from various sources in BC. All observation stations used in this study are the PSPC warning verification sites, which are considered to be representative of the general conditions of each forecast region in BC. In most cases there are several representative stations in one forecast region. To quantify the severity of the weather for each forecast region, the most extreme observed value for each weather element was chosen from all observations in that forecast region during the storm period. Here, the most extreme values for the weather elements are defined as the coldest wind chills during the storm, the strongest winds during the storm, the maximum snowfall amounts in a 24-hour period, the maximum rainfall amounts in a 24-hour period, and the highest total freezing rain amounts and longest durations during the storm for each forecast region. A weather warning event was defined as one where the most extreme value in a given region exceeded the public weather warning criteria (Goosen, 2006; Tatar, 2006). To assess the performance of the forecasts objectively, equivalent datasets of weather warning events for each weather element were created for the forecasts of both the GEM model and the meteorologists at PSPC.

3 Evolution of the early December 2007 storm a Evolution of the Weather Systems A high pressure area associated with air of Arctic origin started to build over Yukon at the end of November 2007 and gradually advanced southward, pushing cold Arctic air into the central and southern interior of BC (Fig. 3). By 1200 UTC , 1 December, the strong ridge of high pressure covered most parts of the BC interior. Cold northeasterly flow prevailed over the northern and central parts of the province and northerly outflow winds began to develop along the coastal valleys and inlets as the cold air spilled out across the Coast

Mountains. Meanwhile, a Pacific low pressure system began to develop rapidly over the offshore waters and track eastward toward Vancouver Island, with a central pressure of 1003 hPa at 1200 UTC, 1 December (Fig. 3a) and of 994 hPa at 1200 UTC, 2 December (Fig. 3b). This low played an important role in producing periods of snow across the BC South Coast on 1 December and the following day by pushing moist marine air into the area. Locally heavy snow was reported along the east coast of Vancouver Island, where the northerly outflow through the mainland inlets served as an onshore flow (Jackson and Steyn, 1994). At 1200 UTC, 2 December the low over Vancouver Island started to dissipate while another low pressure system began to develop rapidly in the central Pacific (Fig. 3b). This low quickly deepened and tracked northeastward toward the BC and Washington coasts. As shown in Fig. 3c, southeast of the low was a long low-level flow across the eastern Pacific, maintaining a persistent stream of warm moist air from subtropical regions toward the BC and Washington coasts, which is a typical PE pattern (see the description of a PE in Section 1). At 1200 UTC, 3 December, the low centre was about 1500 km southwest of Vancouver Island (Fig. 3c). However, a warm front accompanying the PE had already spread moist air over southern BC (Fig. 2). The PE arrival forced the Arctic air mass to retreat from the South Coast and the southern interior of BC, where heavy rain and strong winds began to develop on 3 December. The low itself eventually arrived and moved onshore on 4 December (Fig. 3d). The shaded 1000–500 hPa thickness lines in Figs 3b to 3d clearly show that warm air pushed gradually from the Pacific into the southern and then central interior of BC.

b Temporal Variations of the Observed Weather At the beginning of the storm event, the very cold air mass from the Arctic high pressure over Yukon and northern BC generated strong northeasterly outflow winds through the coastal valleys and inlets. Figure 4 shows the hourly observations of surface winds at Pam Rocks (CWAS, located in Howe Sound) during 1–5 December. There was a long period of strong, cold northerly outflow winds before the arrival of the PE system. The northerly winds peaked during 0000–1200 UTC, 2 December. However, at about 1700 UTC, 3 December the strong northerly outflow winds suddenly changed to strong southerly inflow winds, indicating the arrival of the PE system and the retreat of the Arctic air mass. The winds at Vancouver International Airport (CYVR) showed the same pattern, but the speeds were much weaker. To illustrate the progression of weather over the province as the PE system advanced, Fig. 5 gives the hourly observations of surface temperatures and weather at five stations from west to east (or from coast to interior) of BC (i.e., Vancouver (CYVR), Abbotsford (CYXX), Hope (CYHE), Kelowna (CYLW) and Revelstoke (CYRV)). The locations of these stations are shown in Fig. 1. Under the control of the Arctic

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 99

Fig. 3 The 1200 UTC mean sea level pressure (contour interval 4 hPa, solid lines) and 1000–500 hPa thickness (contour interval 6 dam, dashed lines) for the period 1–4 December 2007 from the GEM-R model operational analysis.

Fig. 4 Hourly observations of surface winds for the period 1–5 December 2007 at Pam Rocks (CWAS) in Howe Sound (region number 1.4 in Fig.1). Negative (blue) and positive (red) values represent cold northerly outflow and warm southerly inflow winds, respectively.

air mass, precipitation fell as snow prior to 0000 UTC, 3 December. As the PE approached, warm air spread onshore at higher elevations. However, the cold air remained entrenched in the boundary layer and mountain valleys, as seen in the observed soundings over KUIL (Quillayute, WA) and CWLW (Kelowna, BC) (not shown). Surface temperatures remained just below freezing. This inversion pattern caused widespread freezing rain over the province starting around 0000 UTC, 3

Fig. 5 Hourly observations of surface temperatures and precipitation type at five stations from west to east across BC (i.e., Vancouver (CYVR), Abbotsford (CYXX), Hope (CYHE), Kelowna (CYLW) and Revelstoke (CYRV)) for 1–5 December 2007. The locations of these stations are shown in Fig.1.

December over the South Coast and gradually spread eastward across the interior the following day. As warm air from the Pacific system continued to pump into the South Coast, surface temperatures rose sharply after 3 December (Fig. 5).

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

100 / Quanzhen Geng et al. Precipitation changed to rain in Vancouver which gradually spread eastward to the Fraser Valley and then into the interior regions. As the system weakened during 0000–1200 UTC, 5 December, rain eased over most regions, marking the end of this notorious storm.

c Spatial Distributions of the Observed Severe Weather Events Figure 6 shows the observed severe weather events in BC that exceeded the public weather warning criteria during 1–5 December 2007. Extreme wind chill, strong winds, heavy snow, extended periods of freezing rain and heavy rain affected most parts of the province. 1 EXTREME WIND CHILL OVER THE NORTH COAST AND STRONG WINDS OVER THE SOUTH COAST

As shown in Fig. 6a, strong Arctic outflow combined with cold temperatures produced very cold wind chill values

(e.g., lower than −20°C) for a long period over the North and Central Coasts. More extreme wind chill values (lower than −40°C) were observed in the Peace River District in northeastern BC. Meanwhile, strong southerly winds spread across the BC South Coast with the arrival of the PE system around 1200 UTC, 3 December. Southeasterly winds above or near the warning criteria (65 km h−1) were observed over most of the exposed areas of the South Coast.

2 HEAVY SNOW

The interaction between the warm, moist air ahead of the PE system with the cold air mass from the Arctic outbreak produced widespread heavy snowfall over the southern and central parts of the province (Fig. 6b). East Vancouver Island, Howe Sound and Whistler received between 35 and 50 cm of snow in 24 hours. Campbell River on East Vancouver Island reported 52 cm of snow in 24 hours. Other parts of the South Coast also received 25 to 35 cm of snow in 24 hours.

Fig. 6 The observed severe weather events that exceeded public weather warning criteria for 1–5 December 2007 over British Columbia. (a) Coldest wind chill and strongest winds, (b) Maximum snowfall in 24 hours, (c) Storm total freezing rain, and (d) Maximum rainfall in 24 hours. ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 101 Record or near-record snowfall amounts were also observed in some of the interior regions such as Fraser Canyon, West Columbia and East Columbia. Lytton in the Fraser Canyon region reported 48 cm of snow in 24 hours, the fourth highest amount since 1917. Golden in the East Columbia region had 44 cm of snow in 24 hours, the most since 1998. 3 FREEZING RAIN

The approaching Pacific low pressure system spread warm air over the entrenched Arctic air in the boundary layer, producing widespread and prolonged freezing rain in the southern and central parts of BC (Fig. 6c). The warning threshold for freezing rain in BC is only 1 mm of ice accumulation or 2 hours of light freezing rain. However, Howe Sound, Whistler and Fraser Canyon reported 25 to 50 mm of freezing rain lasting 10 to 20 hours. Such events had not been seen since PSPC started their warning verification program in 1998 (Goosen, 2006; Tatar, 2006). In the last 10 years, Whistler experienced three freezing rain events, including this one. The first two events lasted only 3 to 6 hours with little accumulation. In comparison, the 12 hours and 27 mm of freezing rain during this event were exceptional. Many other regions over the southern part of the province also reported 5 to 10 mm of freezing rain lasting 5 to 10 hours.

express the rapidness of the arrival of the PE system, because only hourly reports are available from CWAS. The criterion of a wind speed change larger than 60 km h−1 used to select the possible cases of strong interaction between a PE and an Arctic outbreak is arbitrary. Choosing other criteria such as 50 km h−1 or 70 km h−1 will identify more or less possible cases but this will not affect our final conclusion of which event has the strongest interaction between a PE system and an Arctic outbreak. Using the criteria mentioned above, 26 events were identified in the 10-year period from 2000 to 2009 (Fig. 7). Among them, only the 19 November 2001 case is comparable to the 3 December 2007 case in terms of strength and persistence. However, further examination (not shown) indicates that the cold air mass over BC originated from an Arctic ridge of high pressure over northern Canada on 3 December 2007, which was much colder than the cold air mass from a continental high pressure over Alberta on 19 November 2001. The event on 1 January 2006 is also a fairly strong case, but compared to the 19 November 2011 and the 3 December 2007 events, it lacks the strength of cold air in terms of the strength of cold northerly flow. The northerly winds during the 1 January 2006 event never exceeded 50 km h−1 (Fig. 7). This is much weaker than the other two events, during which there was at least one hour with northerly winds stronger than 60 km h−1, reaching the wind warning criterion.

4 HEAVY RAIN

As the warm, moist air of the PE system pushed further inland, the Arctic air mass retreated to northern BC. Freezing rain or snow changed to heavy rain and then spread over most of southern BC. More than 100 mm of rain had fallen in a 24hour period over most of Vancouver Island and the Lower Mainland (Fig. 6d). West Vancouver Island and Howe Sound received 218 mm and 170 mm of rain, respectively, over 24 hours.

d A Very Rare Storm The interaction of a warm, moist marine air mass with a cold, dry Arctic air mass caused widespread severe winter weather over BC in early December 2007. Such a dramatic event, in which an intense PE system and a strong Arctic outbreak occurred at the same time and collided over BC, is very rare. Figure 7 shows events over the past 10 years characterized by very strong surface pressure gradients, those that might have been associated with the interaction of a Pacific low pressure system with an Arctic outbreak. The wind speed and direction changes at Pam Rocks (CWAS, see Fig. 1) are used to select the events since, because of its geographical location, this station easily feels the effects of Arctic outbreaks and approaching Pacific low pressure systems. To express the rapid collision of a Pacific low pressure system with an Arctic outbreak, we required that the one-hour wind speed change from a northerly to a southerly direction at CWAS be larger than 60 km h−1. Furthermore, if the wind changed direction from northerly to southerly it had to remain southerly for at least the following five hours. Here we choose one-hour wind speed change to

4 Forecast by the GEM model The GEM model forecast the general circulation pattern reasonably well during 1–5 December. Figure 8 shows the 72-hour forecast of sea level pressure by the GEM-G model (issued at 0000 UTC, 29 November, valid at 0000 UTC, 2 December) and the corresponding analysis. Both the strength and position of the low just off Vancouver Island on 1 December were well forecast. Snow over many areas of the South Coast on 1 December was forecast 72 hours ahead by the GEM-G model issued at 0000 UTC , 29 November; the model run also predicted 10 to 20 mm of precipitation on 2 December and over 100 mm on 3 December for most areas of the South Coast (figures not shown). This provided an early sign of the potential for heavy rainfall over the South Coast. Figure 9 shows the GEM-R model 24-hour and 48-hour forecasts for mean sea level pressure and 1000–500 hPa thickness, valid at 1200 UTC , 2 and 3 December. Compared with the analysis shown in Fig. 3, positions of the low over Vancouver Island at 1200 UTC, 2 December and the Pacific low at 1200 UTC, 3 December were well forecast, although the depth of the lows were over-forecast. The modelled low over Vancouver Island at 1200 UTC, 2 December was too deep and led to misleading forecast guidance of strong southeasterly winds and warm advection across the BC South Coast. For example, the GEM guidance for Metro Vancouver, based on the model run initialized at 1200 UTC, 1 December (Saturday), was given as: “… Tonight: Snow changing to

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

102 / Quanzhen Geng et al.

Fig. 7 Cases of possible rapid interaction between an Arctic outbreak with a Pacific low pressure system approaching the BC south coast for 2001 to 2009 winters. These cases are selected by using the following conditions. The 1-hour wind speed change from a northerly to a southerly direction at CWAS (Pam Rocks, see Fig.1) must be higher than 60 km h−1, and if the wind changed direction from northerly to southerly it must stay southerly for at least the next 5 hours. Numbers in the figure show the wind speeds 10 hours before, at, and 10 hours after the hour of maximum wind speed change from a northerly to a southerly direction. Green and blue indicate northerly flow higher than 40 and 60 km h−1, respectively. Orange and red indicate southerly flow higher than 40 and 60 km h−1, respectively. The y-axis shows the dates of the selected cases.

Fig. 8 (a) The GEM-G model 72-hour forecast of mean sea level pressure (contour interval 4 hPa, solid lines) and 1000–500 hPa thickness (contour interval 6 dam, dashed lines), valid at 1200 UTC , 2 December 2007, and (b) the corresponding analysis data. ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 103

Fig. 9 The GEM-R model 48-hour and 24-hour forecast of mean sea level pressure (contour interval 4 hPa, solid lines) and 1000–500 hPa thickness (contour interval 6 dam, dashed lines), valid at 1200 UTC , 2 and 3 December 2007, respectively. The corresponding analyses are shown in Figs 2b and 2c, respectively.

rain near midnight. Snowfall amount 5 cm. Becoming windy. Low zero with temperature rising to 6 by morning. Sunday: Rain. Temperature steady near 5” (EC, 2007). As it turned out, the warm advection brought in by the low was much weaker. Temperatures observed at CYVR were steady near −2°C through Saturday night and the maximum temperature on the following day was only 2.4°C. Snow in Metro Vancouver did not change to rain until Sunday evening when the arrival of the PE system brought warmer air. Figure 10 shows the GEM-R model performance with respect to the observed warning events for 1–5 December 2007. It can be seen in Fig. 10a that the GEM model missed the Arctic outflow warnings over the coastal sections of North and Central Coasts, forecasting wind chill values much warmer than −20°C. The poor performance is likely due to the GEM model’s low resolution that cannot resolve the complex terrain in BC, so that the depth of the cold air is not properly represented. Figure 11 shows BC weather stations on a terrain elevation map and compares the real and model elevations of these stations on a scatterplot. All these weather stations are carefully chosen by the PSPC to represent the general conditions of a forecast region for weather warning verifications. This figure shows that almost all the stations are located in the valleys or other relatively low areas. From the scatterplot in the figure we can see that the

model topography is much smoother and the valleys are much shallower than they are in reality. This could lead to the weaker outflow gap winds through the coastal valleys and warmer surface temperatures at the valley bottom (shallower valleys hold less cold air), resulting in much warmer wind chill values. This can be seen from the GEM-R model forecasts for Cathedral Point (CWME) located in a deep valley of the Central Coast–Coastal (region 2.2 in Fig. 1), which indicate that the model forecast much weaker northeasterly outflow winds and warmer surface temperatures than the observations (figures not shown). Figure 10a also shows the difference between the model forecast wind chills and observations. Therefore, it is no surprise to see that the model produced wind chill forecasts over northern BC that were too warm, even in the regions where the model correctly forecast the wind chill warning events. Figure 10a shows that the GEM-R model correctly forecast all the southeasterly wind warning events over the exposed areas of the South Coast, where the local effects on surface winds are minimal. However, the model forecast stronger southeasterly winds than observed, probably because the model forecast a deeper low than the analyses showed (see Figs 9c and 9d and Fig. 3c). Figure 10b shows that the GEM-R model correctly forecast more than half (63%) of the snowfall warning events for

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

104 / Quanzhen Geng et al.

Fig. 10 The GEM-R model forecast of the severe weather events for 1–5 December 2007. The colours in the forecast regions indicate the model’s performance in forecasting the severe weather events during the storms; darker colours indicate hits; lighter colours indicate misses; grey indicates unverifiables; and yellow indicates false alarms. The squares indicate the model’s bias (the difference between the model’s forecasts and observations) for (a) the coldest wind chill and strongest winds; (b) the maximum snowfall in 24 hours; (c) the storm total freezing rain; and (d) the maximum rainfall in 24 hours.

1–5 December 2007. While the model did well for the coastal regions and the windward side of the Columbia Mountains in the interior, it missed almost all the warning-level events over the leeward (eastern) sides of the Coast Mountains and the Columbia Mountains. The differences between the model forecast and the observed snowfall in Fig. 10b suggest that the model under-forecast the snowfall amounts for these regions. One possible explanation is that the “spillover” effect was not handled correctly in the model. A spillover on mountain ridges refers to mid-level hydrometeors being carried by strong winds over the crest to the leeward slope resulting in modification of precipitation distribution such that more precipitation falls downstream of the ridge than would be usual (Browning et al., 1974, 1975). Both the GEM-R and GEM-G models use the Sundqvist microphysical scheme, in which hydrometeors are not explicitly predicted (Sundqvist, 1988). Instead, excess moisture is removed and allowed to fall immediately to the ground as precipitation in one time step. Therefore, whenever the flow is very strong, the lack of hydrometeor drift in the model could lead to precipitation being over-forecast on the windward slopes and

under-forecast on the leeward slopes. As mentioned earlier, the model topography is smoother than the real topography. Would this allow the model winds to flow over the mountains more easily, carrying the moisture with them and so counteract the spillover effect? The answer is probably no, because the moisture being carried over would not experience further orographic uplift; therefore, would not immediately fall as precipitation on the leeward side of the mountains. Note that our recent internal PSPC warning verification statistics (J. Goosen, personal communication, 2011) have found that the higher resolution model (i.e., the 2.5 km GEM-LAM model) achieved significant improvement in precipitation distribution prediction over the mountainous terrain of BC. The improvement could be partly a result of the explicit prediction of hydrometeors in the mesoscale model. The GEM models failed to forecast almost all of the 14 freezing rain events during the 1–5 December (Fig. 10c). In most cases the model simply forecast snow changing to rain (Fig. 10c). One of the major reasons for this is again likely caused by the shallower mountain valleys in the model. The shallow valley results in an insufficient depth of cold air,

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 105

Fig. 11 Locations of the representative observation stations over British Columbia. The scatterplot compares the actual and GEM-R model elevations for these stations.

and then the cold air is removed too quickly compared to reality. Another possible reason could be the inadequacies of the model’s boundary layer parameterizations. The rate of low-level warming is determined, in part, by the mixing of warm air aloft, and this is a difficult process for NWP models to get quantitatively correct. The NWP model guidance, in general, tends to mix down the warm air prematurely in situations such as the storm in this study, even in locations of relatively flat terrain. A study of the UMOS temperature guidance for Vancouver in 2007 also found a noticeable warm bias, especially with the scenario of warm air advection over a cold air mass (Mo, personal communication, 2008). As shown in Fig. 12, the surface temperatures forecast for Abbotsford (CYXX) were several degrees warmer than the observations and rose sharply a few hours earlier than the observations when the PE approached. And most importantly, the model forecast surface temperatures did not have the requisite period of near-freezing conditions. Figure 10d shows that the GEM model successfully forecast 9 of the 15 (or 60%) rainfall warning events. The model correctly predicted all the rainfall events over the windward side of the Coast Mountains but over-predicted the amounts. Meanwhile, the model under-forecast the rainfall amounts and missed all the rainfall warning events over the leeward sides of Vancouver Island and the Coast Mountains. These biases could be attributed to the same factors as those mentioned earlier for the snow forecasts. However, Fig. 10d shows that

Fig. 12 Observed and model forecast surface temperatures for 1–5 December 2007 at Abbotsford (CYXX). The red and blue lines represent the modelled and observed surface temperatures, respectively. This station is located in region 1.3 in Fig. 1.

the GEM-R model forecast too much rain for the Lower Mainland area over the South Coast. The reason for this is still unknown and further investigation is needed in a future study.

5 Forecast by the Pacific Storm Prediction Centre PSPC meteorologists significantly improved the forecasts of severe weather events compared to the model. Figure 13 shows the performance of the PSPC forecasts for the observed weather warning events of 1–5 December 2007. The same

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

106 / Quanzhen Geng et al.

Fig. 13 As in Fig. 9 but for the PSPC forecasts.

methodology used to assess the model performance was applied to the PSPC forecasts. Figure 13a shows that PSPC successfully forecast all the wind chill events over the Peace River District and the North and Central Coasts. Figure 13b shows that PSPC correctly forecast over 77% of the observed snowfall warning events compared to 63% for the model (Fig. 10b). As shown in Fig. 13c, PSPC greatly improved the freezing rain forecast. Out of 14 events, PSPC successfully forecast seven while only one event was correctly forecast by the model. In these scenarios, cold air usually lingers in the mountain valleys even when mild air from the Pacific system has already pushed in aloft, posing a risk of freezing rain in these areas. Applying this local knowledge resulted in an improvement of the model-generated forecasts. For the rainfall events (Fig. 13d), PSPC forecasters achieved significant improvement over the model forecasts, with 80% correct compared with 60% correct for the model. In summary, PSPC issued a total of 48 severe weather warnings for the coastal regions during the 1–5 December 2007 storm, which is a record number for one storm cycle since warning verification began in 1998. For the interior, 51 warnings were issued, which is the fourth highest since 1998. In

general, PSPC performed well in warning the public of the high-impact weather events. The critical success index (Schaefer, 1990) for all warning events for this storm was 0.68, which can be considered a very successful skill score. A comparison of PSPC forecasts and GEM model forecasts with the observed values of weather elements is shown in Fig. 14. In most instances, the PSPC outperformed the model with respect to high-impact weather events. Specifically, by comparing Fig. 13 with Fig. 10, PSPC forecasters made significant improvement to the model forecast of freezing rain and wind chill events during the storm. These two types of high-impact weather are most affected by the complex terrain in BC, which the NWP model has difficulty resolving. Based on their experience and knowledge of various local effects for BC, PSPC forecasters know that the model usually forecasts higher surface temperatures in the valleys in a situation when warm moist Pacific air approaches the province, because of its smoother terrain and shallower valleys which cause the cold air to be scoured out too quickly and because of the inadequacy of boundary-layer parameterizations causing warm air aloft to be mixed down prematurely. Under such circumstances, the forecasters will

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie

Interaction of an Intense Pacific Low Pressure System / 107 While the GEM model performed reasonably well, its smoothed topography compared to reality could have several consequences: (1) Weaker northerly outflow winds or stronger southerly inflow winds could be caused by the inland sea level pressures falling too quickly in the model because of an insufficient depth of cold air being trapped in the model valleys (M. Gélinas, personal communication, 2008). (2) Warmer surface temperatures are forecast because the cold air cannot be retained long enough in the shallower valleys. (3) Stronger southeasterly winds are forecast because the smoother topography did not adequately block the onshore winds. Fig. 14 Comparison of the PSPC and GEM-R model forecasts with observed values of weather elements. The closer the forecast value is to the observed one the better. The red, green and blue bars indicate, respectively, the percentage of PSPC forecasts that are better than, the same as or worse than the GEM model forecast.

adjust the timing of the cold air retreat in places where the mountain valleys are relatively deep, thus improving the forecast of freezing rain and wind chill.

6 Summary and discussion During 1–5 December 2007 a strong Arctic high pressure formed over Yukon and the BC interior spreading very cold air into coastal regions through inland valleys. At the same time, an intense Pacific low pressure system of subtropical origin, often referred to as a “Pineapple Express”, approached the BC coast, spreading abundant moisture and warm air over the province. The interaction between these two air masses produced a record number of high-impact weather events and created huge forecast challenges for both the NWP models and the operational meteorologists. In this study, the evolution of the weather systems and the observed severe weather events during 1–5 December 2007 are analyzed. It is shown that the extent and severity of the high-impact weather events from this storm broke BC’s single storm records since PSPC started its public weather warning verification program in 1998. The widespread and prolonged freezing rain reported across much of southern and central BC was exceptional; historical data show no evidence of a similar event in the province during the last decade. The widespread severe weather associated with this storm was caused by a very warm, moist air mass from an intense PE system overrunning an unusually strong and prolonged Arctic outbreak. Weather forecasts by the GEM model and PSPC meteorologists were compared with observations. The GEM model did reasonably well in forecasting the severe weather events during the storm; however meteorologists at PSPC further improved the model forecasts for all severe weather events. Knowledge of the local effects of the complex terrain allowed operational forecasters to correct the model biases caused by the inadequacy of its boundary-layer parameterization, leading to significant improvement over the model forecasts.

The combinations of these effects resulted in some noteworthy biases in the GEM models in forecasting severe weather during this storm: (1) The model forecast for wind chill over the North Coast was too warm and the forecast for southeasterly winds over the South Coast was too strong. (2) The model under-forecast snowfall amounts over most of the province except over upslope regions of high mountain ranges. (3) The model forecast the transition from snow to rain to happen too quickly during a freezing rain event for most regions. It was noticed that the GEM-R model under-forecast precipitation on the leeward sides and over-forecast on the windward sides of the mountains. Our conjecture (which we have not investigated) is that the problem is likely caused by the model treatment of the spillover of precipitation to the leeward sides of the mountains. The Sundqvist cloud parameterization scheme (Sundqvist, 1988) used in the GEM-R model does not explicitly predict hydrometeors. Rather excess moisture is allowed to fall immediately to the ground as precipitation. Therefore, the spillover effect is not predicted by the model. Even a model with the hydrometeor advection mechanism could still underestimate the spillover effect if its topography is smoother than the actual topography. According to Browning et al. (1974, 1975), spillover precipitation depends mainly on the contribution of mid-level hydrometeors, because the forced ascent on the windward slopes of terrain may act to cause the hydrometeors aloft to extend to higher levels and, therefore, be carried further downstream to the leeward slopes. The smoother topography in the model would result in weaker forced ascent on the windward slopes and, therefore, less leeside spillover precipitation. Weather forecasting for BC is challenging because of the various local effects of its complex terrain and inadequacies in the model’s physical parameterizations. Decision-making during the forecast process involves considering model guidance, accounting for various local effects and recognizing weather patterns based on the meteorologist’s own experience. Successful forecast of the early December 2007 storm

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 Canadian Meteorological and Oceanographic Society

108 / Quanzhen Geng et al. provided a good example of meteorologists at the PSPC using these decision-making skills. Acknowledgements We would like to thank all PSPC meteorologists who were involved in forecasting this severe winter storm for the

period 1–5 December 2007. Special thanks go to Ford Doherty and Don Tatar for their assistance with data collection and weather warning verification. Michel Gélinas, Barry Brisebois and Ian Okabe provided valuable discussions. Suggestions and critical comments from two anonymous reviewers have resulted in significant improvement of the paper.

References and C.E.J. JACKSON. 2012. Meteorology Today: An Introduction to Weather, Climate, and the Environment (1st Canadian Ed.). Toronto, Ontario: Nelson Education Ltd. BOURGOUIN, P. 2000. A method to determine precipitation types. Weather Forecast. 15: 583–592. BROWNING, K.A.; F.F. HILL and C.W. PARDOE. 1974. Structure and mechanism of precipitation and the effect of orography in a wintertime warm sector. Q. J. R. Meteorol. Soc. 100: 309–330. BROWNING, K.A.; C.W. PARDOE and F.F. HILL. 1975. The nature of orographic rain at wintertime cold fronts. Q. J. R. Meteorol. Soc. 101: 333–352. COLLE, B.A. and C.F. MASS. 2000. The 5–9 February 1996 flooding event over the Pacific Northwest: Sensitivity studies and evaluation of the MM5 precipitation forecasts. Mon. Weather Rev. 128: 593–617. CÔTÉ, J.; S. GRAVEL, A. MÉTHOT, A. PATOINE, M. ROCH and A. STANIFORTH. 1998. The operational CMC-MRB Global Environmental Multiscale (GEM) model: Part I. Design considerations and formulation. Mon. Weather Rev. 126: 1373–1395. EC (Environment Canada). 2007. FPCN11 CWVR 011900 Forecasts for the south coast of British Columbia issued by Environment Canada at 11.00 am PST Saturday 1 December 2007 for today and Sunday. Archived on the Pacific Storm Prediction Centre internal website. Available upon request. GALEWSKY, J. and A. SOBEL. 2005. Moist dynamics and orographic precipitation in northern and central California during the New Year’s Flood of 1997. Mon. Weather Rev. 133: 1594–1612. GENG, Q. and M. SUGI. 2003. Possible change of extratropical cyclone activity due to enhanced greenhouse gases and sulfate aerosols – Study with a highresolution AGCM. J. Clim. 16: 2262–2274. GOOSEN , J . 2006. A review of weather warning performance for the BC interior and Yukon: 1998–2005. Technical Report of the Pacific Storm Prediction Centre, Vancouver, British Columbia: Environment Canada. HEIDORN, K.C. 2004. The BC Weather Book: From the Sunshine Coast to Storm Mountain. Markham, Ontario: Fifth House Ltd. JACKSON, P.L. and D.G. STEYN. 1994. Gap winds in a fjord. Part I: Observations and numerical simulation. Mon. Weather Rev. 122: 2645–2665. JOHNSON, K. and J. MULLOCK. 1996. Aviation Weather Hazards of British Columbia and the Yukon. Kelowna, British Columbia: Environment Canada. KLASSEN, K. 2008. Persistent weak layers and the winter of 2007/08. Avalanche Rev. 26(4): 24–27, Retrieved from http://www.american avalancheassociation.org/tar/TAR26-4_LoRes.pdf AHRENS, C.D.; P.L. JACKSON

and J.E. MARTIN. 2007. A Pacific moisture conveyor belt and its relationship to a significant precipitation event in the semiarid southwestern United States. Weather Forecast. 22: 125–144, doi: 10.1175/ WAF963.1 LACKMANN, G.M. and J.R. GYAKUM. 1999. Heavy cold-season precipitation in the northwestern United States: Synoptic climatology and an analysis of the flood of 17–18 January 1986. Weather Forecast. 14: 687–700. LANGE, O.S. 1999. The Wind Came All Ways: A Quest to Understand the Winds, Waves and Weather in the Georgia Basin. North Vancouver, British Columbia: Environment Canada. LOUKAS, A. and M.C. QUICK. 1996. Spatial and temporal distribution of storm precipitation in southwestern British Columbia. J. Hydrol. 174: 37–56. LYONS, C.J. 1901. The storms of the Hawaiian Islands. Mon. Weather. Rev. 29: 68–68. KNIPPERTZ, P.

MAILHOT, J.; S. BÉLAIR, L. LEFAIVRE, B. BILODEAU, M. DESGAGNÉ, C. GIRARD, A. GLAZER, A.-M. LEDUC, A. MÉTHOT, A. PATOINE, A. PLANTE, A. RAHILL, T. ROBINSON, D. TALBOT, A. TREMBLAY,

and A. QADDOURI. 2006. The 15-km version of the Canadian regional forecast system. Atmosphere–Ocean, 44: 133–149. MASS, C. 2008. The Weather of the Pacific Northwest. Seattle, Washington: University of Washington Press. READ , W . 2007. The major coastal gale of December 1–3, 2007. Retrieved from http://www.climate.washington.edu/stormking/ REITER , M . 2008. December 1–4, 2007 storm events summary. Weyerhaeuser Company Report. Retrieved from http://www.weyerhaeuser.com/pdfs/ company/media/December2007StormEventSummary.pdf SCHAEFER, J.T. 1990. The critical success index as an indicator of warning skill. Weather Forecast. 5: 570–575. SUNDQVIST, H. 1988. Parameterization of condensation and associated clouds in models for weather prediction and general circulation simulation. In M.E. Schlesinger (Ed.), Physically-based modelling and simulation of climate and climatic change (pp. 433–461). Dordrecht, The Netherlands: Kluwer Academic Publishers. TATAR , D . 2006. A review of weather warning performance for the BC coast: 1997–2004. Technical Report of the Pacific Storm Prediction Centre. Vancouver, British Columbia: Environment Canada. WILSON, L.J. and M. VALLÉE. 2002. The Canadian Updateable Model Output Statistics (UMOS) system: Design and development tests. Weather Forecast. 17: 206–222. P. VAILLANCOURT, A. ZADRA

ATMOSPHERE-OCEAN 50 (1) 2012, 95–108 http://dx.doi.org/10.1080/07055900.2012.656261 La Société canadienne de météorologie et d’océanographie