3. PRE Climatology 3.1 Statistical Climatology 3.1.1 Overview As stated in section 1.1, 47 PREs were identified downstream of 21 Atlantic basin TCs during 1998–2006. The average PRE-producing TC (PPTC) thus spawned 2.2 PREs, implying that the production of multiple PREs downstream of a single TC is common. Seventeen of the 47 (36%) Atlantic basin TCs that made landfall in the U.S. during 1998–2006 produced at least one PRE. Three PPTCs that did not make landfall in the U.S. [Alex (2004), Irene (2005), and Ophelia (2005)] recurved off the East Coast, and one [Emily (2005)] made landfall in northern Mexico. While not all PPTCs made landfall in the U.S., they all were located northwest of 20°N, 65°W when they produced PREs; thus, all TCs crossing this point were deemed capable of producing PREs. Twenty one of the 83 (25%) TCs in the study period crossing this point produced at least one PRE. Figure 3.1 shows a histogram of the occurrence of Atlantic basin PPTCs and a line graph of the occurrence of all Atlantic basin TCs during 10- and 11-day periods encompassing the dates for which Atlantic basin TCs were officially recognized by NHC during 1998–2006. Atlantic TC occurrence peaked during 1– 10 September, but PRE occurrence peaked during 21–31 August and 11–20 September. The PRE minimum during 1−10 September corresponds to a relative minimum in U.S. landfalling TCs (Fig. 3.2), but both of these minima are likely statistical artifacts resulting from the limited nine-year TC sample. PREs also can 51 occur early or late in the tropical season, but Fig. 3.2 indicates that they were more likely to form in June and July than in October and November during 1998– 2006. 3.1.2 TC Tracks Favorable for PRE Development The procedure and motivation for separating TCs into different categories based on similarity of TC track is discussed in section 2.2.2. The tracks of all SR TCs listed in Table II are shown in Fig. 3.3a so the tracks taken by TCs in this category can be visualized. Seven of the 11 (64%) TCs represented in Fig. 3.3a produced at least one PRE, which is the highest percentage of PPTCs in any track category containing at least 10 TCs. The tracks of the seven SR PPTCs are displayed in Fig. 3.3b on a map containing dots that represent PRE formation locations. Each dot is colored to match the track color of the TC that helped produce it. Sixteen PREs formed in association with the seven SR PPTCs, which is an average of 2.3 PREs per PPTC. Nearly all of the PREs occurring downstream of SR PPTCs formed in a southwest–northeast band from the Gulf Coast to southern New England. Comparison of Fig. 3.3b with Fig. 3.4 shows that PRE formation locations were generally along or south and east of the Appalachian ridge line and in proximity to the Gulf of Mexico and/or Atlantic Ocean. Six of the 15 (40%) AR TCs whose tracks are shown in Fig. 3.5a produced at least one PRE, which is the second highest percentage of PPTCs in any track category containing at least 10 TCs. The tracks of the six AR PPTCs and the 52 formation locations of the PREs that occurred downstream of them are illustrated in Fig. 3.5b. The three nonlandfalling PPTCs that recurved off the East Coast (see section 3.1.1) are represented by the green, red, and blue tracks, respectively. Twelve PREs formed in association with the six AR PPTCs, which is an average of 2.0 PREs per PPTC. Figure 3.5b shows that the PREs occurring downstream of AR PPTCs formed along the immediate mid-Atlantic or southern New England coastline or near elevated terrain of the Northeast (refer to Fig. 3.4). Three of the 10 (30%) CG TCs whose tracks are shown in Fig. 3.6a produced at least one PRE, which is the third highest percentage of PPTCs in any track category containing at least 10 TCs. The three CG PPTCs depicted in Fig. 3.6b tracked west of the Appalachian ridge line (refer to Fig. 3.4). Thirteen PREs formed in association with the three CG PPTCs, which is an average of 4.3 PREs per PPTC—about twice as many as formed in association with SR or AR PPTCs. Figure 3.6b shows that PRE formation locations downstream of CG PPTCs spanned the distance from the Upper Midwest to the East Coast during 1998–2006. Only six of the 47 (13%) PREs occurring during 1998–2006 formed downstream of TCs following tracks other than the three already described. Two TCs [Bret (1999) and Emily (2005)] out of 11 (18%) that made landfall southwest of the Texas/Louisiana border produced at least one PRE. Two TCs [Harvey (1999) and Wilma (2005)] out of six making landfall in Florida, but passing southeast of SR TCs (see section 2.2.2 for exact definition) produced at least 53 one PRE. Finally, Isabel (2003) and Fran (1996) both made landfall in the midAtlantic and produced PREs, but were not classified as AR cases because they continued moving northwestward into the U.S. after landfall. No null cases following tracks similar to Isabel and Fran have been identified since 1996. 3.1.3 PRE Locations Relative to TC Track Anticipating the likely location of PRE formation relative to the eventual TC track can help forecasters assess the potential impacts of PRE and TC rainfall. In particular, areas affected by AT PREs may experience extreme flooding because of the occurrence of heavy TC rainfall after the predecessor rain. Areas affected by LOT and ROT PREs do not receive heavy TC rainfall, but the PRE rainfall in those areas still may be significant enough to cause flooding. Figure 3.7 illustrates that 26 of the 47 (55%) PREs in the study period occurred LOT. In comparison, 12 of the 47 (26%) PREs were AT and 9 of the 47 (19%) PREs were ROT. The AT percentage implies that, given a PRE will form, there is approximately a one-in-four chance that the rainfall directly associated with a TC will subsequently fall over the same area affected by a PRE. 3.1.4 Other Statistical Properties of PREs Box plots were constructed for the distributions of PRE SDs, time lags, durations, 24-h rainfall rates, and speeds. Four box plots will be shown for each set of data: one showing the distribution for all PREs that occurred during 1998– 2006, and one each for the distributions of LOT, AT, and ROT PREs. According 54 to Wilks (2005), box plots are a concise way of displaying the five-number summary of a data distribution, which includes the smallest and largest data points, the lower and upper quartiles, and the median. In the box plots described below, the box defines the bounds of the interquartile range (IQR), which contains the middle 50% of the data and provides a measure of statistical variability. The whiskers extend to values within a distance of 1.5 times the IQR from the upper and lower quartiles. The median is represented by a black dot within the IQR, and outliers are plotted individually outside the IQR. Positive (negative) skewness can be inferred from the box plots in the figures that follow if the top (bottom) whisker is longer than the bottom (top) whisker and if the median is less (greater) than the mean. The median SD between all PPTCs and PREs is 935 km. Positive skewness can be inferred in the all-PRE SD distribution shown in Fig. 3.8 because the mean SD is approximately 150 km greater than the median. The LOT PRE SD distribution is similar to the all-PRE SD distribution, in part because LOT PREs comprise more than half the PRE database. The narrow IQR of the AT PRE SD distribution indicates that SDs vary less with AT PREs than with LOT or ROT PREs. The ROT PRE SD distribution demonstrates a notably higher mean and median than the other two PRE categories and exhibits a slight negative skewness. The median time lag for all PREs is 36 h, but time lags could not be calculated for one LOT and two ROT PREs because the centers of their parent TCs never reached the latitudes of the PRE centroids. Figure 3.9 shows that the 55 all-PRE time lag distribution is positively skewed because the mean of 45 h is 9 h greater than the median. The extreme outlier in the all-PRE time lag distribution occurred when the remnants of Hurricane Dennis (2005) stalled for several days over the lower Mississippi Valley before reaching the latitude of the PRE. The LOT PRE time lag distribution in Fig. 3.9 shows a greater mean and median and greater variability than the all-PRE time lag distribution. The AT and ROT PRE time lag distributions exhibit much less variability than the LOT PRE time lag distribution. The mean and median time lags for ROT PREs are less than for LOT or AT PREs. Given the expectation that small time lags would be associated with small SDs, it is counterintuitive that ROT PREs have a larger median SD than LOT or AT PREs (refer to Fig. 3.8). However, calculation of TC speeds during PREs provides no statistical evidence that the large SDs and short time lags associated with ROT PREs are attributable to faster TC movement in ROT PRE cases than LOT or AT PRE cases. The four box plots summarizing the data distributions of PRE durations (Fig. 3.10) all have means and medians between 12 and 15 h, suggesting that the location of a PRE relative to the track of its parent TC has no effect on how long it lasts. However, each duration distribution, except for the one associated with AT PREs, contains an outlier of 36 h, suggesting that long-lasting PREs can occur. The AT and ROT PRE duration distributions exhibit the least variability of the four distributions because they have the smallest range and IQR, respectively. 56 Figure 3.11 reflects the median rainfall rate of all PREs during 2001–06 of 203 mm (24 h)−1 and the lower bound of 100 mm (24 h)−1 required for classification as a PRE. Comparison of the AT and ROT PRE rainfall distributions with the LOT PRE rainfall distribution in Fig. 3.12 reveals that the AT and ROT PRE rainfall distributions have larger IQRs and greater mean and median rainfall rates than LOT PREs. Specifically, the mean 24-h rainfall rates of AT and ROT PREs are approximately 60 mm greater the mean rainfall rate of LOT PREs. Despite the smaller samples of AT and ROT PREs compared to LOT PREs, the highest 24-h rainfall rates of all PREs occurred with AT and ROT PREs. The box plots shown in Fig. 3.12 reveal that PRE speeds can range from nearly zero to greater than 20 m s−1. The speed distribution of all PREs is positively skewed with mean and median values of 9.5 and 7.4 m s −1, respectively. The LOT and AT PRE speed distributions are similar. Despite one extreme outlier, the mean ROT PRE speed is 5.8 m s −1, which is ~4–6 m s−1 slower than the mean LOT and AT PRE speeds. The IQR of the ROT PRE speed distribution is considerably smaller than that of the all-PRE speed distribution, indicating that ROT PRE speeds have much less variability. Comparison of Fig. 3.11 with Fig. 3.12 shows that high ROT PRE rainfall rates correspond to slow ROT PRE speeds, but that high AT PRE rainfall rates are associated with fast AT PRE speeds. 3.2 Composite Climatology 57 Synoptic-scale geography-relative composite plots spanning the 24-h period centered on PRE initiation time were constructed using 11 of the 16 PREs forming downstream of SR PPTCs. The dates and times included in the composites are listed in Table III alongside the TCs associated with the PREs. Two types of composite plots will be shown every 12 h to emphasize important synoptic-scale signatures: 1) 700-hPa geopotential height overlaid on upward vertical motion to diagnose the geostrophic flow pattern in which PREs form downstream of SR PPTCs and 2) 925-hPa geopotential height and θe overlaid on 200-hPa wind speed to diagnose the low-level temperature and moisture fields and upper-level jet structure. Figure 3.13a shows that the composite TC (represented by the green star) is located just west of the Florida peninsula 12 h prior to PRE initiation. Upward vertical motion is maximized northeast of the TC center along the Georgia coast, within a confluence zone created by a 700-hPa trough northwest of the TC and a 700-hPa ridge east of the TC. A 200-hPa jet streak with peak winds greater than 40 m s−1 is oriented from southwest to northeast over the northeastern U.S. and is collocated with a 925-hPa θe gradient downstream of a trough axis in southeastern Canada (Fig. 3.13b). Southwesterly geostrophic flow at 925 hPa is oriented parallel to a θe-ridge axis extending northeastward from the TC (Fig. 3.13b), implying poleward transport of warmer and moister air. The synoptic-scale pattern changes little during the next 12 h. Figure 3.14a shows that the composite PRE forms along the North Carolina/Virginia border downstream of the nearly stationary 700-hPa trough and north of the 58 maximum upward vertical motion just prior to TC landfall along the western Florida coast. The PRE also forms near the equatorward entrance region of the 200-hPa jet streak, just to the west of the 925-hPa θe-ridge axis (Fig. 3.14b). Twelve hours after it forms, the composite PRE is located in northeastern Virginia and is accompanied by a second composite PRE that forms in close proximity to the first PRE (Fig. 3.15a). While it may seem contradictory that two PREs are located in virtually the same place at the same time, the composite position of the second PRE at this time indicates that the subsequent PRE occurring downstream of a SR PPTC forms 12 h after the first PRE forms and within a synoptic-scale environment similar to that of the first PRE. Figure 3.15a shows that the composite TC and the composite PREs are located downstream of the 700-hPa trough. A significant change from 12 h earlier, however, is that there now are two maxima in the upward vertical motion field—one near the TC, and one near the two PREs northeast of the TC. Figure 3.15b indicates that the PREs are located on the anticyclonic shear side of the 200-hPa jet streak and near an implied axis of dilatation at 925 hPa between ridges to their west and east. Although composites of PREs in other TC track categories have not been constructed, the synoptic-scale environments in which the PREs formed have been examined subjectively. Nine of the 12 PREs occurring downstream of AR PPTCs and three of the 13 PREs occurring downstream of CG PPTCs formed in association with the four synoptic-scale signatures evident in the SR composites—namely, ahead of a midlevel trough, near the equatorward entrance 59 region of an upper-level jet, within a low-level θe ridge or gradient, and in an area receiving warm, moist air from the TC. Including the 11 PREs in the SR composites, 23 of the 47 (49%) PREs identified during 1998–2006 formed in environments characterized by all the synoptic-scale signatures identified by the SR composites. The synoptic-scale environments of the remaining PREs varied more widely and featured only some of the signatures seen in the SR composites. Possible explanations of this observation are that: 1) some of the four synoptic-scale signatures seen in the SR composites may play greater roles than others in producing PREs and 2) mesoscale and physiographic processes below the 2.5° × 2.5° resolution of the composites may have a greater influence in producing PREs than synoptic-scale processes. 60