SEARCH

SEARCH BY CITATION

Keywords:

  • hurricanes;
  • extreme precipitation;
  • North America

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

Despite the catastrophic societal costs of hurricanes and considerable recent attention to possible trends, little is known about the general contribution of hurricane-related activity to extreme precipitation over North America and the underlying dynamics. Here, a 25-year observational analysis of daily data shows important contributions to extreme events over large regions of North America, including a pronounced signal over northern and inland areas, with an average span of influence extending to several hundred kilometers. Over large areas of the Northeast, there are stations where more than two-thirds of all extreme events are linked to hurricane-related activity. Large-scale vertical velocity, maximum wind speed, and tropical/extratropical character are shown to be important factors in the strength and range of influence. Analysis of dynamical factors, including buoyancy, moisture availability, and lifting, show the largest changes in lifting, with rising motion typical of the deep tropics occurring over inland and northern latitudes.

1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

Individual hurricanes and their remnants can produce exceptionally intense rainfall [Rodgers and Adler, 1981; DiMego and Bosart, 1982; Hellin et al., 1999] and the associated flooding, even independent of storm surge, is one of the leading causes of hurricane-related death in the U.S. [Rappaport, 2000]. The societal impacts of hurricanes are catastrophic [Diaz and Pulwarty, 1997] and concerns about possible recent trends in strength and number have received considerable attention [Emanuel, 2005; Webster et al., 2005; Hoyos et al., 2006; Santer et al., 2006; Shepherd and Knutson, 2007; Mann et al., 2009; Bender et al., 2010]. While the maximum contribution of hurricane-related activity to seasonal precipitation totals is generally less than 25% even for the most-affected coastal regions [Rodgers et al., 2000; Englehart and Douglas, 2001; Rodgers et al., 2001; Larson et al., 2005; Corbosiero et al., 2009; Jiang and Zipser, 2010], the contribution to the number of extreme events in the southern coastal regions can be significant [Englehart and Douglas, 2001; Knight and Davis, 2009; Konrad, 2001; Shepherd et al., 2007; Konrad and Perry, 2010]. However, a general assessment of the extent of the continental influence, especially over northern and non-coastal regions, has yet to be done, and the underlying dynamical mechanisms are not well understood.

Here, observational analysis of 25 years, 1975–1999, of daily data is used to examine both the average radius of influence for hurricane-related activity (hurricanes, tropical storms, and midlatitude remnants) and the pattern of influence over North America. Influence is considered in terms of the number of extreme events that occur within some radial distance of the activity and is tested for significance with a Monte Carlo approach. An absolute (4 inches [10.16 cm] or greater) and a local measure (10 wettest days at each station) are used to define extreme daily precipitation. The dynamics of the influence are also investigated through analysis of three fundamental factors in the occurrence of precipitation: moisture availability, buoyancy, and large-scale lifting (vertical motion).

2. Data

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

Track positions, speed of the system, maximum wind speed, and tropical/extratropical character of the hurricane-related activity are taken from the 2006 version of the US National Hurricane Center's HURDAT dataset [Jarvinen et al., 1984; Davis et al., 1984], including all track categories (tropical, subtropical, extratropical, wave, remnant low) in the eastern Pacific and Atlantic for 1975–1999. Daily precipitation is taken from 9,464 stations in the Global Daily Climatology Network [National Climatic Data Center, 2002] for North America that report at least 50% of the time in that 25-year period. Most US stations report at 99% or higher, but the 50% level was chosen to retain some coverage in Central America. Buoyancy data, in terms of Convective Available Potential Energy (CAPE), is taken from the North American Regional Reanalysis [Mesinger et al., 2006] as daily means at 1/3° × 1/3°, to provide high resolution over the continent, while atmospheric water content (precipitable water and specific humidity at 925 hPa) and vertical velocity are taken from the NCEP/NCAR reanalysis [Kalnay et al., 1996] as daily means at 2.5° × 2.5°, to include adjacent tropical ocean areas for comparison.

3. Radius of Influence and Contributing Factors

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

The average influence on extreme events as a function of radial distance, including all tracks and all stations, is shown as the black line in Figure 1a (“AVG”) in terms of likelihood. The local definition of extreme, the 10 wettest days at each station, is used here to provide a more comparable definition of extreme across the greatly varying mean precipitation amounts across North America. The maximum likelihood in the average case is 8.5%, which represents more than a 75-fold increase over the random chance of 0.11% (10 days over 25 years). Statistical significance is assessed through a monte-carlo approach where the dates of the hurricane-related activity are randomly reshuffled in time and then all calculations are repeated 1000 times based on the random data (year and day-of-month are randomized, but calendar month is retained, to avoid introducing a seasonal bias); the 99% significance level is shown as a green line.

image

Figure 1. Radius of hurricane-related influence. (a) Probability of an extreme daily precipitation event as a function of distance from hurricane-related activity for the 1975–1999 period. The average probability (“AVG”) is given as a thick black line, along with several subsets of events: extratropical (“ET”), maximum wind speed greater than 50 kt (“VMAX > 50”), and pressure vertical velocity less than 0.2 Pa/s (“VVEL < −0.2”). An extreme event is defined as one of the 10 wettest days (largest 24-hour precipitation accumulations) at each station (station distribution shown in Figure 2). The 99% significance level for the average case (all tracks) is shown as a green line. (b) Number of days in the 25-year period that are within 500 km of the track center of a hurricane, tropical storm or remnant, as recorded in NOAA's HURDAT database, shown on a 0.5° × 0.5° degree grid for land only. Contouring is at intervals of 25, with an additional contour at 1. The maximum value is 400 days, in southwestern Mexico.

Download figure to PowerPoint

To investigate the aspects of the hurricane-related activity that affect the strength of influence, the activity is composited by intensity, speed, tropical/extratropical character, and large-scale vertical velocity (2.5° × 2.5° grid). The average considering only activity with a maximum wind speed of at least 50 knots (25.7 m/s, the strongest 17.6% of systems) is shown as the “VMAX” curve in Figure 1a, and has nearly double the likelihood of an extreme event near the track center compared to the overall average (“AVG”). Storms moving slower than 2.5 m/s (the slowest 17% of systems, not shown), while providing a longer time to accumulate precipitation at any given station, actually have slightly less influence on extreme events than the average, at least for the 24-hour accumulations considered here. An average of only extratropically-transitioned systems, the “ET” curve in Figure 1a, shows less influence than the average of all systems out to around 200 km, but considerably increased influence at greater distances, consistent with the larger scales of extratropical systems [Hart and Evans, 2001]. Finally, the large-scale vertical velocity (lifting), while not a traditional metric of hurricane strength, is also analyzed here as vertical motion is a primary factor in the occurrence and strength of precipitation. The average including only activity with a large-scale pressure vertical velocity less than −0.2 hPa/s at 500 hPa—the strongest 17.6% of systems in terms of rising motion—is shown as the “VVEL” curve. Hurricane-related activity with strong rising motion is shown to have the highest probability of an extreme event of all factors considered out to nearly 600 km.

4. Pattern of Influence

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

The geographic pattern of influence is investigated through examination of the frequency with which extreme events occur within the range of influence of the hurricane-related activity. Based on Figure 1a, 500 km is a conservative estimate for an average radius of influence and is used as the threshold here; the results are not dependent on the exact value. The number of days in the 25-year period that are within 500 km of hurricane-related activity are shown in Figure 1b. This pattern is similar to plots of track density (National Hurricane Center, Historical track distributions at http://www.nhc.noaa.gov/pastprofile.shtml#ori; U.S. Geologic Survey, Hurricane return period map at http://www.usgs.gov/hazards/hurricanes/) and return period [Hart and Evans, 2001].

Figure 2 shows the influence on extremes using the local measure (10 wettest days). Based on the Monte Carlo analysis, only stations with changes locally significant at the 95% level are colored; the field significance exceeds 99.9%. While extreme events are associated with the hurricane activity over most of the possible range, the number of influenced days (or track density plots) is not a good indicator of the strength of influence by this measure (cf. Figures 2 and 1b). Note that all the events are not necessarily caused by hurricane-related activity—some may be primarily trigged by other mechanisms and only modified by the activity; full attribution for a specific extreme event would require additional investigation.

image

Figure 2. Hurricane-related extremes, local definition. The number of the 10 wettest days at each station over the 1975–1999 period associated with hurricane-related activity (estimated as above). Red indicates 7 or more of the 10 wettest days, orange indicates 4–6 of the 10 wettest days, yellow indicates 1–3 of 10, and gray indicates none of the 10 wettest days are associated with hurricane-related activity (by this estimate). Areas with no stations shown, such as parts of Central America, indicate a lack of station data, not necessarily a weak signal. Only stations locally significant at higher than 95% are colored; the field significance exceeds 99.9%.

Download figure to PowerPoint

Areas of greatest influence on locally-defined extreme events occur in coastal North Carolina in the US, the tip of the Baja Peninsula and the southwestern coast in Mexico, and the island of Puerto Rico, all areas with frequent proximity to hurricane-related activity (Figure 1b) and documented links to extremes [e.g., Englehart and Douglas, 2001; Knight and Davis, 2009; Konrad, 2001; Konrad and Perry, 2010; Shepherd et al., 2007]. However, the Gulf Coast and Florida in the US and Northwest Mexico are also regions with similar numbers of affected days but a much lower occurrence of extreme events. The Northeast US, in contrast, has much fewer affected days, but still includes several stations with more than a third of their extreme events coming in conjunction with hurricane-related activity, even away from the coast.

To provide a more complete assessment, an absolute definition of extreme event—4 inches (10.16 cm) or more in a day—is also considered. The number of days at each station meeting this threshold is shown in Figure 3a. The percentage of 4-inch days that occurred within 500 km of hurricane-related activity is shown in Figure 3b, showing only stations that reported at least one 4-inch day. As before, only changes locally significant at the 95% level are colored, and the field significance exceeds 99.9%. Here, an even greater emphasis on the Northeast US is seen. While the number of hurricane-related extreme events is actually larger in the Gulf and southern coastal areas [e.g., Hart and Evans, 2001], the overall contribution of hurricane-related activity (percent of overall extremes) is much smaller in those regions than in the Northeast. In the Gulf Coast and Florida, strong convection and heavy rain are a much more regular occurrence [Changnon, 2001], whereas the mechanisms for extremely heavy precipitation in the Northeast are much more limited, allowing hurricane-related activity to play a more dominant role. Indeed, 4-inch events for 147 of the stations occurred only in the presence of hurricane-related activity during this period.

image

Figure 3. Hurricane-related extremes, absolute definition. (a) The total number of extreme precipitation days in the 25-year period, as measured by 24-hour accumulations of 4 inches (10.16 cm) or more. Shading interval is 5 days, shown in legend in lower-left of the plot. (b) The percentage of extreme precipitation days associated with hurricane-related activity, analyzed as above. Red indicates 67% or more, orange indicates 33–67%, yellow indicates 33% or less, and grey indicates none. Stations not recording any 4-in plus days are omitted in Figure 3b. Only stations locally significant at higher than 95% are colored; the field significance exceeds 99.9%.

Download figure to PowerPoint

5. Dynamical Factors

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

Finally, the underlying physical mechanisms of the linkage are investigated through examination of the hurricane-related changes to three fundamental factors in the occurrence and strength of precipitation: moisture availability, buoyancy (convective available potential energy), and large-scale lifting (vertical velocity). Changes in moisture and buoyancy (not shown), are generally less than 25% over almost the entire domain with the exceptions of the Southwest and Northeast US, where local moisture increases can exceed 40%. Vertical velocity, however, exhibits a striking increase over almost the entire region of North America affected by hurricane activity. Figure 4a shows the climatological mid-level vertical velocity during hurricane season (all days when a track was present anywhere in the domain) and Figure 4b shows the vertical velocity during hurricane-related activity for each grid box (averaged individually for each grid box only for those days when the activity was within 500 km of the center of that grid box, as with the station data; grid boxes that are never within 500 km are shaded gray in both plots). When under the influence of hurricane-related activity, vertical velocity attains values typical of tropical convection over almost all of the affected continental region, including the high latitudes. This is especially notable in the Northeast and Southwest US, which switch from weak descent during average conditions to deep tropical values of ascent when influenced by hurricane activity, providing a very strong dynamical forcing for extreme rainfall and consistent with the results of section 4.

image

Figure 4. Dynamics: changes to vertical velocity. (a) Climatological 500 hPa pressure vertical velocity during the hurricane season, with gray shading masking the areas influenced by hurricanes only 5 or fewer days during the period. Red indicates ascent and blue indicates descent. (b) 500 hPa pressure vertical velocity when under the local influence of a hurricane, with the same contour intervals and shading convention. That is, the vertical velocity averaged for each grid box only when that grid box is within 500 km of hurricane-related activity.

Download figure to PowerPoint

6. Summary and Implications

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

This investigation of the contribution of hurricane-related activity to daily extreme precipitation events over North America reveals a large influence outside the coastal regions usually considered for hurricane impacts: well over 50% of events for several regions and exceeding 25% for large swaths of the continent, with a pattern that does not simply relate to track density. The role of two typically-tracked storm parameters, maximum wind speed and extratropical character, are quantitatively evaluated and shown to play a key role. Moreover, large-scale vertical velocity (even at resolutions typical of operational weather data without additional sampling of individual storms) is here also shown to be a critical factor, and, indeed, has the largest influence out to 600 km.

In terms of the dynamics underlying the forcing of extreme precipitation, changes to large-scale lifting (vertical velocity) are much larger than changes to moisture availability or buoyancy. Surprisingly, the values of vertical velocity are largest over the Northeast US, which suggests that further investigation of the interaction of the hurricane-related activity with the large-scale circulation, perhaps especially with respect to the jet stream, will be key to understanding the influence on extreme precipitation. Future work is also needed to separate the contributions to the radial influence from different geographical regions and with respect to the quadrant relative to the track [Matyas, 2007], as well as to consider multi-day precipitation extremes, and to assess the degree to which current forecast models are able to capture the observed relationships.

Acknowledgments

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References

Comments by Suzana Camargo, Frank Colby, and two anonymous reviewers helped significantly improve this work. Support was provided through NSF 0621237.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Data
  5. 3. Radius of Influence and Contributing Factors
  6. 4. Pattern of Influence
  7. 5. Dynamical Factors
  8. 6. Summary and Implications
  9. Acknowledgments
  10. References