1. Introduction

Heavy precipitation forecasting and the quantitative precipitation forecasting (QPF) problems remain among the most challenging of tasks to an operational forecaster (Fritsch et al. 1998; Olson et al. 1995). Roebber and Bosart (1998) have pointed out that synoptic-scale forecasts of heights have improved during the past several decades, while much slower progress has occurred in predicting precipitation amounts. Although numerical model resolutions and computational capabilities have increased dramatically during recent years, notable problems with specific precipitation forecasts have occurred recently. These problems are tied to a combination of initial condition uncertainty and/or model physics. Zhang et al. (2002) discuss the issues relating to poor QPFs in association with the 24–25 January 2000 snowstorm in North Carolina (e.g., Zhang et al. 2002).

The case discussed by Roebber and Bosart (1998) was characterized by moist convection, and therefore its predictability was likely to be limited compared with cold-season, higher-latitude cases. However, there has been no published work that systematically documents statistically significant synoptic-scale precursors associated with precipitation events in the northeastern United States. Research on precursors that are likely responsible for triggering major cold-season precipitation events has been performed for the Mackenzie River basin in the Northwest Territories of Canada (Lackmann and Gyakum 1996, 1999) and for southeastern Quebec, Canada (Fischer 1997). The notion of pattern recognition is still used by operational forecasters (Funk 1991) in conjunction with the presence of more sophisticated numerical weather prediction systems. Durran and Snellman (1987) point out that good forecasts are based upon the understanding of factors that produce the weather. Indeed, much of our understanding of such factors is focused on synoptic-scale circulations. Studies by Lackmann and Gyakum (1996) for precipitating events in the Mackenzie River basin and by Fischer (1997) for the Montreal, Quebec, region have focused on the identification of the associated large-scale environments. A unique aspect of our work is the focus on large-scale circulation precursors to varying-intensity precipitation events in the Burlington, Vermont (KBTV), region that have occurred over a relatively long period of record (33 yr).

The objectives of this study are as follows:

1. To identify large-scale precursors to several intensity classifications of cold-season precipitation events at KBTV. KBTV (indicated with a star in Fig. 1) is located in the Champlain Valley of northern Vermont and New York. Burlington is chosen because it is the most populated area in the region of responsibility for the forecasting office. The Champlain Valley is a narrow north–south valley between the Adirondack Mountains to the west and the Green Mountains to the east (Fig. 1d). Although the precipitation details are impacted by the surrounding orography (e.g., Roebber and Gyakum 2003), the synoptic scale has been shown to be prominently associated with such important cold-season events as the 1998 ice storm (Gyakum and Roebber 2001).

2. To provide forecasters with objective guidance to determine that a significant meteorological event will occur that may result in the issuance of a weather watch, warning, or advisory.

The next section describes the data and methodology. Section 3 focuses on the precipitation climatology for the Burlington, Vermont, station. The composites of sea level pressure (SLP) and 1000–500-hPa thickness for three classifications of precipitation amounts are examined in section 4. Surface cyclone tracks are studied in section 5. Water vapor transports and precipitable water composites are discussed in section 6. A concluding discussion is given in section 7.

2. Data and methodology

For the purpose of this study, we define the cold season as the period between November and March, inclusive, which corresponds to the snowiest 5 months of the year climatologically at KBTV. This period also defines the specific range of time during which winter weather advisories, watches, and warnings are issued, and also corresponds to the period with the least convection.

We use hourly precipitation data from KBTV for the period between 1963 and 1995 available on CD-ROM from the National Climatic Data Center (NCDC). The data were included on the Solar and Meteorological Surface Observation Network (SAMSON), CD-ROM (NCDC 1993) and on the Hourly United States Weather Observations (HUSWO) CD-ROM (NCDC 1997). The precipitation observations were all taken prior to the introduction of the Automated Surface Observing System (ASOS). Any freezing, or frozen, precipitation was measured as the precipitation fell into a weighing rain gauge containing antifreeze and melted.

Twenty-four-hour running totals were extracted from the data using spreadsheet software. An event was defined as the greatest 24-h running total of precipitation when compared to previous and subsequent 24-h totals. It should be noted that actual precipitation events span more and less than 24 h. Once this 24-h event has been identified, we specify that at least 48 h must elapse before another event's initiation may begin. This time-spacing requirement is designed to provide for independence of each case.

We identify 1387 precipitation events in which at least 0.2 mm (0.01 in.) of precipitation was observed. The precipitation was categorized based on QPF categories with which operational forecasters in the National Oceanic and Atmospheric Administration's (NOAA) National Weather Service (NWS) are familiar. These categories are also similar to those forecast by the model output statistics (MOS; Klein and Lewis 1970; Dallavalle and Erickson 2000). The categories chosen were 0.2–3 mm (0.01–0.09 in.), 4–6 mm (0.10–0.24 in.), 7– 13 mm (0.25–0.49 in.), 14–24 mm (0.50–0.99 in.), and 25–50 mm (1.00–2.00 in.). A heavy precipitation event is defined as 25–50 mm (1.00–2.00 in.) in 24 h, a moderate event as 14–24 mm (0.50–0.99 in.), a light event as 7–13 mm (0.25–0.49 in.), and a very light event as 4–6 mm (0.10–0.24 in.).

Forty-eight heavy precipitation events occurred in the category of 25–50 mm. There was only one event found to be over 50 mm (2.00 in.), and this extreme event was not used in our study. To maximize the distinction between the categories of cases, we select 50 events in each of the moderate and light categories. These 50 events were chosen by the criterion that their amounts are closest to the median of each category, with 25 events below and 25 above the median amount. The range of precipitation amounts occurring about the median of each were 17–21 mm (0.65–0.84 in.) for moderate events, 9–10 mm (0.34–0.40 in.) for light events, and 4–5 mm (0.16–0.20 in.) for very light events.

Six-hourly fields of SLP, 1000- and 500-hPa geopotential heights, wind, and water vapor mixing ratio were used from the National Centers for Environmental Prediction (NCEP) global reanalysis grids with a horizontal grid spacing of 2.5° of latitude–longitude and 28 vertical levels (Kalnay et al. 1996). The climatological anomalies of these fields were computed by taking the difference between the total field and the monthly climatological means. These climatological means are weighted according to the number of cases occurring in each month. The monthly mean values are computed from a 33-yr (1963–95) period of record. A similar procedure for compositing anomalies has been used successfully by Lackmann and Gyakum (1999) in their study of Mackenzie River basin precipitation events. Using the same compositing methodology, we have produced fields of precipitable water, the vertical integral of the horizontal moisture transport, and its convergence.

The start of heaviest 24-h period of precipitation was applied to the nearest 6-h time in the NCEP reanalysis (0000, 0600, 1200, and 1800 UTC) defined as time zero or t+00. The time 12 h before the beginning of the heaviest 24-h period was designated t−12, and the time in the middle of the heaviest period t+12, and so on. Twelve-hourly composites were obtained from the time as early as 108 h (t−108) to 60 h (t+60) after the event.

The General Meteorological Package (GEMPAK; Koch et al. 1983) was used to analyze and display the resulting grids. Extratropical cyclone tracks for the heavy and moderate events present over the United States and southern Canada at the time of t+00 were identified from the Global Tropical/Extratropical Cyclone Climatic Atlas (hereafter GTECCA; NCDC 1996). Previous and subsequent positions of the cyclone were obtained from the GTECCA and by combining the data analysis capability of the Grid Analysis and Display System (GrADS; Doty et al. 1997).

3. Climatology

The mean annual precipitation at KBTV (using the 1961–90 climatology) is 877.95 mm (34.47 in.). The elevation of the station is 104 m (340 ft) above mean sea level. The monthly amounts for November, December, January, February, and March are, respectively, 79.50 mm (3.13 in.), 61.47 mm (2.42 in.), 46.23 mm (1.82 in.), 41.40 mm (1.63 in.), and 56.64 mm (2.23 in.). The flanking regions of higher elevations, including the Green and Adirondack Mountains, typically receive larger amounts of precipitation. The minimum value of our heaviest 24-h category (25 mm) approaches 55% of January's monthly climatology and 32% of November's climatology. The more recent KBTV precipitation climatology (1971–2000) shows an annual mean of 915.67 mm (36.05 in.) with little change in the seasonality.

The synoptic-scale climatology (Fig. 1) through this cold-season period shows several features of relevance. A climatological thickness trough lies meridionally in eastern North America throughout the cold season. The SLP features include a climatological-mean anticyclone in the subtropical Atlantic and over the continental United States. The relatively strong surface trough that extends from Atlantic Canada into the western Atlantic is associated with the time-mean Icelandic low. There is also a mean SLP trough in the Great Lakes, most prominent in November (Fig. 1a). This is likely related to the sensible-heating-driven pressure reduction (Petterssen and Calabrese 1959) that peaks during the late autumn season in the Great Lakes. The monthly mean precipitable water values range from 11 mm in November to 7.5 mm in January for Burlington, Vermont. The particularly large value of precipitable water in November, relative to the subsequent cold season, is likely to be one prominent factor in the relatively large number of significant precipitation events during the month of November.

The 24-h precipitation histogram (Fig. 2) shows that 251, or 18%, of the 1387 events in the 33-yr period were light, 157 (11%) were moderate, and 49 (3.5%) were heavy. The vast majority (930 or 67%) of the 24-h precipitation events at KBTV in the cold season were extremely light precipitation events of less than 7 mm (0.25 in.). While 24-h precipitation amounts in excess of 25 mm are certainly extreme events for KBTV, the moderate category (14–24 mm) would likely trigger a winter weather watch or warning for freezing and/or frozen precipitation, and light category events (7–13 mm), in general, would not. Forecasters therefore need to distinguish between the moderate and light categories when making winter weather watch or warning decisions.

Figure 3a shows the monthly histogram, normalized by 30-day periods, of all cold-season precipitation events. The relatively large number of events is nearly equally distributed among November, December, and January, with a substantially reduced frequency for February and March. It is likely that the seasonal migration of the jet stream and storm tracks southward away from the KBTV region is responsible for the reduction in the number of events by February. Figure 3b shows the monthly distribution of each composited precipitation intensity category's frequency. The total number of heavy cases, 48, is represented in the composite. Only the sample of 50 cases in each of the light and moderate composited categories is represented in the figure. November has largest number of events in the heavy category, with 40% of the cases represented. December and March each contribute nine cases, or 19% of the total number of heavy cases. These results are consistent with the fact that November is the largest monthly contributor to the climatological cold-season precipitation amounts, as discussed earlier in this section.

4. Composites of sea level pressure, 1000–500-hPa thickness, and their anomalies

The purpose of this section is to identify key synoptic-scale circulation features, and their climatological anomalies, associated with each of the intensities of precipitation. These features will be displayed at 24-h intervals from t−108 to t+12 h. Additionally, we display differences in anomalies of specified categories, and the statistical significance of these differences, for t−36 through t+12 h. The purpose of this display is to identify significant synoptic-scale precursors that are unique to a specific categorization of precipitation. A secondary objective of this display is to assess our capability to distinguish between a major and more benign precipitation event.

5. Storm tracks

Figure 10 shows the 6-hourly positions of the surface cyclones that are located within 1100 km of KBTV at t+12 h for the 48 heavy precipitation events. Most, but not all, events were associated with surface cyclones. Some events were associated with concurrent cyclone centers. These tracks reveal details that are otherwise masked or smeared by the compositing procedure. For example, though we have referred to individual low centers in our prior discussions of the SLP composites, these features are the result of the compositing procedure and do not reveal the variability of low centers' locations with respect to the averaged low center. Figure 10 shows that many of heavy cases feature a surface cyclone that tracks from the Middle Atlantic states into New England. Other surface cyclones take a more inland track over the Great Lakes. The mean position of the surface cyclone during the 24-h period of most intense precipitation (t+00 to t+24) is optimally located to the southwest of KBTV.

Figure 11 shows the cyclone tracks for the moderate cases. When compared to tracks of the heavy cases, we find that the surface cyclone tracks again converge on New England. However, the mean track of these cyclones is slightly more zonal and to the north (relative to the mean cyclone track in the heavy cases) across the United States through the Ohio Valley, prior to its passage toward KBTV. The composite central pressures decrease by 7 hPa (Fig. 11b) during the 24-h period of most intense KBTV precipitation, compared with 9 hPa in the heavy precipitation composite (Fig. 10b). Both samples are associated with surface cyclones that are deepening, and generally reach their maximum intensities much later, at t+84 h, than the occurrence of the KBTV precipitation event. Our results are consistent with our previous observations that there is weaker cyclogenesis in the moderate versus the heavy cases, and that this weaker cyclogenesis is associated with weaker onshore geostrophic flow from the south.

6. Water vapor and its transport

We relate the synoptic-scale circulation structures found in earlier sections to the composite water vapor features of the heavy, moderate, and light events.

The precipitable water depth, w, is defined by

View Expanded

where g is the gravity, ρ_l is the density of liquid water (assumed to be 1000 kg m⁻³), q is the specific humidity, and p is the pressure. We define the vertical integral of the water vapor transport vector, Q, as

View Expanded

where v is the horizontal wind vector.

The expression for precipitation amount, P, may written in the context of a moisture budget for synoptic-scale flows (Kuo and Anthes 1984) as

View Expanded

where the first term on the right-hand side of (3) is the horizontal convergence of Q; the second term is the storage term, or the time rate of change of precipitable water; and the third term, E, is the evaporation.

We show the composites of SLP; the precipitable water, w; the vertical integral of the water vapor transport, Q; and the composite divergence of Q for each class of precipitation amounts in this section.

Figure 12, showing the heavy precipitation composites, reveals very large precipitable water values of 20 mm and greater throughout New England during these events (Fig. 12d). These values are nearly 6 mm larger than the mean found in the moderate composite (Fig. 12f). Prior to the event, substantial northward moisture transports are occurring in the southerly geostrophic flow to the east of the surface cyclone discussed in section 4a. These transports are directed from the Gulf of Mexico as early as t−60 h and increase as the event time approaches. Furthermore, an organized region of convergence of moisture transport of approximately 4 mm (12 h)⁻¹, centered in Oklahoma (Fig. 12a), migrates eastward and amplifies to 12 mm (12 h)⁻¹ by t+12 h. Perhaps the most revealing and important distinction between the heavy and moderate composites is shown by the difference flux vectors of anticyclonic flow that are directed more northward in the region of the precipitation, and preferentially from the southeast and the Atlantic Ocean in the heavy versus moderate composites (Figs. 12e,f). This suggests that the heavy precipitation events are being physically enhanced by preferential transports of water vapor from the Atlantic Ocean.

The moderate cases (Fig. 13) show a similar large-scale structure to that of the heavy composites. However, the precipitable water amounts are smaller in the region of precipitation by almost 6 mm, the northward transports are weaker into the precipitation region, and there is a less obvious conduit of water vapor flux from the Gulf of Mexico and the Atlantic Ocean (Figs. 13b–d). Although there are larger values of precipitable water in the moderate cases versus those of the light cases, the differences are only 1–2 mm. The generally drier air to the southwest and moister air to the east, with the accompanying larger respective northerly and southerly transports, are associated with the stronger synoptic-scale signals discussed earlier for the moderate cases, as compared with the light precipitation cases. Much of the enhanced southerly moisture flux in the moderate versus light cases is located eastward of the precipitation region in the Atlantic Ocean.

Figure 14, showing the light cases, reveals the following features: The precipitable water increases with time to approximately 14 mm over KBTV (Fig. 14d). The values of precipitable water are almost 8 mm less than those associated with the heavy cases. The water vapor fluxes are directed weakly from the southwest into the precipitation region prior to and during the event (Figs. 14a–d). The difference (heavy minus light) fields of flux vectors reveals a very strong relative northerly transport in the southwestern domain and a southerly flux of water vapor directly from the subtropical regions of the Atlantic Ocean in the heavy cases. This southerly flux and the link to the Tropics was also demonstrated by Gyakum and Roebber (2001) for the great ice storm of 1998.

7. Conclusions

Our work has produced several results that have implications for forecasting heavy precipitation. Among the 5 months designated as the cold season, and used for this study, November stands out as having the most heavy and moderate precipitation events (Fig. 3b). This fact may be due to a combination of baroclinic storm tracks migrating southward through the KBTV region, along with the sea surface temperatures of the western Atlantic being relatively warm in November compared with the other cold-season months.

As might have been expected from this study a priori, a composite surface cyclone is associated with each intensity category of precipitation (Figs. 4f, 6f, and 8f). The strength of the composite cyclone is stronger as the precipitation amounts increase (Figs. 4h, 6h, and 8h). An especially interesting result of this study is that the downstream surface anticyclone is especially strong for the heavy cases (Fig. 4f). The implication for this result is that a stronger southerly component of the geostrophic flow exists in the heavy-case composites versus the other composites (Figs. 4g,h and 6g,h). This southerly component of flow tends to be more directly from the relatively warm ocean waters of the subtropical regions of the Atlantic Ocean. This fact relates well to the importance of the November maxima in the heavy precipitation cases discussed above.

The upper-level structures generally show increasingly favorable surface cyclone development as the precipitation intensity increases (Figs. 5, 7, and 9). The amplitude of the warm 1000–500-hPa thickness anomaly in the heavy cases averages to +21 dam (+10.5°C). The combination of this warm anomaly occurring in the relatively warm cold-season month of November helps to explain why the heavy events are typically rainstorms.

The upper-level trough associated with the heavy events can be readily traced back to a position off of the west coast of British Columbia (Fig. 5). These statistically significant negative circulation anomalies can be found at t−108 h off the Washington–Oregon coast and actually as early as t−144 h (6 days) prior to the onset of a heavy event at Burlington, Vermont. This negative anomaly moves through the Rocky Mountains, the Gulf of Mexico coastal region, then into the New England region.

The overall results of this study are very similar to that of Fischer's (1997) study of synoptic-scale circulation anomalies associated with heavy precipiation in southeastern Quebec. This similarity provides us with an a posteriori assessment that the choice of KBTV is a representative choice for the Champlain Valley–Saint Lawrence Valley region.

Finally, the atmosphere's water vapor content is especially large in the heavy composites (Fig. 12), as compared with climatology (Fig. 1) and with the moderate (Figs. 12e,f and 13) and light (Fig. 14) composites. The relatively strong water vapor transport from the marine regions in the heavy cases, as compared with the other composites (Figs. 12, 13, and 14), exemplifies the fact that extreme cold-season precipitation events benefit from the transports of relatively large amounts of water vapor into the region from the lower latitudes of the warm oceans. This particular crucial characteristic was observed in the great 1998 ice storm that affected the same region (Gyakum and Roebber 2001).