NCEP Short-Range Ensemble forecasts of an historic rainfall event: The 23-27 June 2006 East Coast Floods By Richard H. Grumm National Weather Service Office, State College PA 16803 And Norman W. Junker National Centers for Environmental Predictions 1. INTRODUCTION A quasi-stationary north-south oriented frontal boundary set-up along the East Coast on the 24 th of June 2006. This frontal boundary slowly, indicative of a synoptic type flood event (Maddox et al. 1979) drifted westward over time focusing the potential for heavy rain over the Mid-Atlantic region and northeastern United States for about 4 days. Heavy rains were observed along and mainly on the warmer side of this boundary from Virginia to southern New England. The heavy rains produced both flash flooding and river flooding. The rain began on the 22 nd ahead of a cold front that stalled along the coast on the 23 rd early on the 24 th. Initially, heavy rains on Saturday, 24 June were confined to coastal regions. Heavy rains, with some 125 mm (5 inch) swaths extended from Massachusetts s southwestward across Long Island, New Jersey, Delaware, Maryland and Virginia. The rain shifted westward on the 25 th with axis of the heaviest rains following Appalachian Mountains from North Carolina to south-central New York. The front slid a bit farther east on the 25 th producing 2 days of heavy rains from Maryland to central New York. The largest aerial coverage of heavy rain Figure 1 Estimated rainfall for the 24-hour period ending at 1200 UTC 28 June 2006. Values are as shown on color key to right. Image courtesy of the National Weather Service precipitation analysis project.
occurred on the 27 th. Widespread flooding and flash flooding was observed. In central Pennsylvania, 20 of the 33 counties in the State College County Warning Area (CWA) experienced flash flooding. Major flooding was observed along the Susquehanna River and its tributaries at Bloomsburg, Middletown, and Harpers Tavern. An additional 7 locations experienced moderate flooding. Flooding was also observed along the Delaware River and the Mohawk River in New York. Large sections of the New York State Thruway (Interstate 90) were closed due to flooding along the Mohawk River. Flooding caused serious problems from Maryland to New York State during the course of this long duration event. It will be shown that forecasts from both the National Centers for Environmental Predictions (NCEP) medium- and shortrange ensemble forecast systems (MREF and SREF respectively) correctly predicted the pattern associated with this event. There were clear signals for a prolonged synoptic type (Maddox et al 1979) heavy rainfall event in the Mid- Atlantic region to southern New England. The SREF forecasts from 23 and 26 June 2006 showed the potential for a strong north-south frontal boundary stalling along the East Coast. The SREF showed a prolonged period of above normal precipitable water (PWAT) along and ahead of the front. Significant anomalies were forecast and forecast to persist. Conditions favoring a Maddox Synoptic type event were evident in the forecasts. Shorter-range forecasts showed an anomalous low-level jet into the region of excessively high PWAT. Overall, the signal for a prolonged heavy precipitation event was well forecast. However, the span of the event exceed the overall forecast length of the SREF system. In addition to the overall pattern, the anomalies of key parameters often associated with heavy rain events were significant (Hart and Grumm, 2001). The strong low-level southerly jet and high above normal PWAT have been shown to be key indicators of heavy rain events (Grumm and Hart 2001 and Hart and Grumm 2001a). The precipitable water anomalies in this event were similar to those in the devastating New England Flood of 7-8 October 2006 (Grumm 2005). An examination of model quantitative precipitation forecasts revealed that the details were not as well forecast. The probability for heavy rains, though high, was not forecast consistently from cycleto-cycle and the area identified by the ensembles though useful, was not as successful as the pattern predicted by the ensembles. Model jumpiness and QPF s played a role in the forecast problems associated with this event.
2. METHODS All SREF data were archived in realtime along with all the NAM and GFS model data. All images were recreated after the event using GrADS (Doty and Kinter 1992). Standard displays of spaghetti and spread are shown. The spread is normally shown with gray shading. The spaghetti plots show distinct contours for each parameter (Sivillo et al 1997). Lower panels for some fields show the ensemble mean with color shading showing the ensemble mean in standard deviations from normal, referred to as standardized anomalies (Grumm and Hart 2001;Hart and Grumm 2001). Figure 2. SREF forecasts of total accumulated rainfall from forecasts initialized at 0900 UTC 24 June 2006. Data display concepts are similar to those shown in Figure 4. This paper will document the event of 24-27 June 2006. The emphasis is on the forecasts produced by the NCEP shortrange ensemble prediction systems (EPS). A companion document will evaluate the MREF performance during this event. The goal is to demonstrate the overall value of ensembles in predicting the patterns associated with significant and historic precipitation events. Some weaknesses, as they presently exist will also be presented. Details of flooding and river flooding are not covered in this document. Precipitation displays show the critical thresholds over a time period, normally 24 hours and 0.50 inches of quantitative precipitation forecast (QPF). Upper panels normally show the probabilities and the consensus of the specified thresholds. Lower panels show the position of the specified threshold for each EPS member and shading shows the ensemble mean QPF. The spaghetti plots follow the standard ensemble concepts outlined by Sivillio et al (1997). This paper is a good reference point for those learning to use ensembles. Forecast data shown here will primarily from the NCEP SREF. All dates and times will be in the format of day and hours. Such that 0900 UTC 25 June 2006 would be displayed at 25/0900. Precipitation probabilities are shown for long periods, such as 60-hours to illustrate the long duration event. This event had several surges in the overall
length of the system. Thus illustrating the need for longer range prediction systems being used in conjunction with more detailed shorter range tools. SREF forecasts. 3. RESULTS The SREF forecasts from 23 June captured the same synoptic set-up indicated in the MREF (not shown). Furthermore, the SREF forecasts did not extend far enough into the future to capture the entire event. Thus, in long duration events, the SREF is useful in refining forecasts to define high threat areas but may not be able to capture the rainfall associated with a multi-day extreme rainfall event. The SREF kept the heaviest rainfall focused near the frontal boundary but never indicated the heavy rainfall that affected coastal regions on the 24 th into the 25 th. i. 24 June SREF Figure 3. As in Figure 1 except precipitation analysis valid for the 24-hour periods ending a) 1200 UTC 25 June 2006 and b) 1200 UTC 26 June 2006.. precipitation that contributed to the heavy rainfall amounts over the period of several days. The SREF played a critical role in identifying areas of heavy rainfall at shorter ranges, but was of limited value at longer ranges as the event lasted longer than the forecast Figure 2 shows the 84-hour accumulated rainfall from the 15- member SREF from forecasts initialized at 24/0900 UTC. These forecasts capture the heavy rain focused along the frontal system. But they do not indicate the heavy rains which were observed (Figure 3) on the 24 th along the coastal region. Rainfall estimates were over 2-5 inches on Long Island and a large portion of New Jersey. Over the length of these forecasts, The SREF gave no explicit indication of heavy rains over eastern areas where heavy rain was observed early in the
a. 24/1800 b. 25/1800 Figure 4. SREF forecasts of PWAT initialized at 0900 UTC 24 June showing the ensemble mean and the standardized anomalies valid at a) 1800 UTC 24 June and 1800 25 June 2006. event. On a positive note, the SREF captured the high probability of heav y rains along the Appalachian Mountains from North Carolina to southern New York. Maximum amounts (not shown) were on the order of 5 inches at several locations indicated by plume diagrams for select points. T he SREF properly forecast the frontal boundary and high PWAT field to focus along the East Coast on 24 June (Figure 4) and it forecast this boundary to drift westward with time. By 25/1800 UTC the SREF was forecasting an anomalously strong southerly jet (Fig. 5) in the above normal moisture axis in the Figure 5. SREF 850 hpa winds initialized at 0900 UTC 24 June valid at 1800 UTC 25 June 2006. Data shown include a) 850 hpa winds and U-wind anomalies and b) 850 hpa winds and V-wind anomalies. Mid-Atlantic region. This feature was forecast to persist and slowly move northward over the next 36 hours. T he location of the moisture and strong jet appeared to play a significant role in determining where SREF focused the rain. This forecast cycle showed higher probabilities of heavy rain after 25/1800 UTC than previous cycles (not shown). The 2100 UTC 26 June ensemble mean SREF system ensemble mean forecasts of 850-hPa moisture flux anomalies (not shown) were above 3 SDs for the entire period from 26/0000 UTC through 28/0000 UTC and were over 4 SDs over southern PA at 26/1200 UTC and over PA and southern NY at 28/0000 UTC.
Figure 7. As in Figure 2 except from SREF initialized at 2100 UTC 24 June 2006. Overall, the PWAT, 850 hpa winds and moisture flux fields showed the same synoptic flood pattern indicated by the MREF. SREF forecasts initialized at 24/2100 UTC showed a stronger low-level jet than previous forecasts and pushed the frontal boundary farther to the west. The V-wind anomalies exceeded +3 SD s above normal from 26/0000 UTC through 27/0000 UTC and the forecasts valid at 26/1200 UTC are shown in Figure 6. This resulted in a westward shift in the precipitation shield (Figure 7) and oddly, a reduced probability of heavy rains in excess of 2 inches. Figure 6. As in Figure 3 except from SREF initialized at 2100 UTC 24 June 2006. Maryland and Pennsylvania were the only areas forecast to received an inch or more QPF during the 24 hour period ending at 26/1200 UTC (Fig. 7). Comparing these forecasts to the verification (Fig. 8) it appeared that the general pattern was well forecast by the SREF though the maximum rainfall was grossly under forecast. Individual members did forecast as much as 5 inches at various grid points (not shown) but overall the forecasts were low. The pattern associated with the strong lowlevel jet and PWAT anomalies were clues to higher rainfall amounts than forecast.
Figure 8. As in 5 except 24-hour accumulated rainfall valid at 1200 UTC 26 June 2006. ii. 25 June SREF The 25/0900 UTC SREF showed a strong southerly 850 hpa wind anomaly on the order of 3 to 4SD s above normal over Pennsylvania from around 26/0600 UTC through 28/0600 UTC. This strong jet extended into New York State and was associated with +1 to +2 SD above normal precipitiable water anomalies. The 850 hpa jet valid at 27/0600 UTC is shown in Figure 9. This jet and the moisture produced the band of heavy rain over the period of the entire SREF forecasts as shown in the 84-hour accumulated precipitation (Fig. 10b) and the shorter-range 24 hour Figure 9. As in Figure 6 except SREF initialized at 0900 UTC showing forecasts valid at 0600 UTC 27 June 2006. precipitation ending at 26/1200 (Fig. 10a). The 24 hour data is comparable to the values shown in Figure 3b. The 24- hour QPF pattern looked good though the SREF under forecast the probability of 2 inches of QPF. The forecasts from 25/2100 UTC showed similar results and for brevity are not shown. iii. 26 June SREF The SREF forecasts initialized at 26/0900 UTC showed a sharp increase in the magnitude of the low-level jet. Southerly wind anomalies exceeded +4SDs above normal by 2100 UTC on
Figure 10a. As in Figure 2 except SREF initialized at 0900 UTC 25 June valid for the 24 hour period ending at 1200 UTC 26 June 2006. Figure 10b. As in Figure 2 except SREF initialized at 0900 UTC 25 June valid for the 84 hour period ending at 2100 UTC 28 June 2006. the 27 th (Fig. 11) and approached +5SD s above normal at 28/0000 UTC (Fig. 12) falling below +4SDs above normal after 28/0900 UTC. The focus of the jet was over Pennsylvania and New York during this period. PWAT values remained above 40 mm and were around +2 SDs above normal during most of this period in the affected region (not shown). Though not alluded to earlier, the 250 hpa jet showed a strong entranc e region over the eastern United States. The strong southerly jet is shown in Figure 13. Unidirectional shear is a k ey element in many flood events and this deep strong jet played a critical role in this system. The 84-hour accumulated rainfall for the period ending at 29/2100 UTC is shown in Figure 12. This SREF cycle forecast a narrow region of 2 or more inches of rain region along the ax is of the warm moist low-level. This forecast covered the two critical 24 -hour periods ending at 1200 UTC 27 and 28 June 2006 (Fig. 14) when the heaviest rains fell in Pennsylvania and New Yor k leading to significant flooding on several large rivers. The SREF had two problems including too little QPF and showing the precipitation shield too far west. To its credit, it forecast the synoptic flood pattern and did s o with anomalous southerly winds at both 850 and 250 hpa. It also had the enhanced moisture plume (PWAT). The signals were there but the specificity of the QP F was still missing.
Figure 11. SREF forecasts initialized at 0900 UTC 26 June 2006 valid at 2100 UTC 27 June 2006 showing 850 hpa winds and a) U-wind anomalies and b) V-wind anomalies. Forecasts from 26/2100 UTC (not shown) indicated a similar pattern to those initialized at 26/0900 UTC. Key features included the anomalous jet and PWAT fields. Forecasts continued to miss the maximum rainfall and tended to pull the heavy rainfall too far west of the observed locations. Figure 14 shows the 60-hour QPF which covers the two periods of heavy rain in Figure 15. These data show that the SREF placed the threat too far west and did not produce a threat of heavy rainfall over the Pennsylvania and New Jersey border, where the Delaware river would go above flood stage due to heavy rains. This poor QPF had significant impacts on forecasts of flash flooding and river flooding. Figure 12. As in Figure 9 except valid at 0300 UTC 28 June 2006. 4. CONCLUSIONS These data suggested a Maddox Synoptic (Maddox et al 1979) type storm with all the ingredients for heavy rains (Doswell et al 1996). The well forecast surge of tropical moisture, documented in the PWAT fields, suggests the models properly handled the interaction of the tropical system and the frontal system. These data demonstrate how using ensemble mean fields and climatic anomalies along with ensemble QPFs can provide confidence in a significantly above normal rainfall event. The strong and anomalous signals in the forecasts produced high confidence in a major
Figure 13 As in Figure 9 except for 250 hpa winds. event. The QPF and QPF probability forecasts suggest the details of where the heavy rain, based on the models composing the EPS still lack skill in identifying the area of highest impact. This case demonstrates the utility of ensemble fields of precipitable water to recognize heavy rain patterns and parameters. These fields were only recently made available from NCEP SREF forecasts in the manageable small SREF forecast files. These data, when used in combination with the QPF forecasts they can help in the forecast process. These PWAT fields, combined with the wind fields, in this event, the anomalous southerly low-level jet, lend Figure 14. SREF 60-hour accumulated rainfall from forecasts initialized at 2100 UTC 26 June 2006. Probabilities of exceeding 2 inches is in the upper panel and the ensemble mean and each members 2 inch contour is shown in the lower panel. themselves to future AI forecast applications to predict heavy precipitation events. In the interim they will serve to help identify historic and near historic precipitation events. As the event unfolded the patterns identified in the ensembles and high resolution deterministic models contributed to human interpretations of the forecast. In a forecast discussion the anomalies were employed to heighten awareness before the final surge of extreme heavy rainfall over east-central Pennsylvania EXTREME FLASH FLOOD THREAT FOR TODAY WITH APPROACH OF ANOMALOUS 850 MB JET TAPPING 2.00+" PWATS IN
A RICH TROPICAL MOISTURE PLUME SOURCED FROM THE EASTERN CARIBBEAN. SREFS DEPICTING 850 MB WIND ANOMALIES OF +3 TO +4 STANDARD DEVIATIONS COMBINED WITH PWAT ANOMALIES OF +2 TO +3 S.D. ANOMALOUS 850 MB JET WILL ENHANCE CONVERGENCE ALONG THE QUASI- STATIONARY BOUNDARY DRAPED ACROSS CENTRAL PA...PRODUCING TRAINING CONVECTIVE CLUSTERS WITH LOCALLY HEAVY RAINFALL AND LIKELY WIDESPREAD FLASH FLOODING ACROSS THE SOUTH CENTRAL AND CENTRAL MOUNTAIN REGIONS TODAY INTO TONIGHT. THIS IS PERHAPS THE MOST POTENT SYNOPTIC SETUP FOR FLOODING I'VE EVER SEEN OVER CENTRAL PA THAT HAS NOT BEEN ASSOCIATED WITH A (DECAYING) TROPICAL CYCLONE...WE WILL HAVE OUR HANDS FULL. The event suggested several SREF strengths, though as shown in Figures 16 and 17, it likely under forecast the strength of the observed low-level jet. But overall, the SREF correctly predicted: i. the pattern to include the frontal system retrograding to the west with time; ii. The amplified pattern with the sharp trough and ridge pattern (Fig. 17). Though for brevity SREF forecasts of these features were not presented. iii. The high PWAT s, anomalous lowlevel jet and associated anomalously high moisture flux; iv. and a Maddox Synoptic type flood pattern in association with anomalous values of winds and moisture These products and the overall well anticipated event mark the success of the NCEP SREF system. Furthermore, they imply that the combined use of ensemble data with climatic data allows the forecaster, and eventually, artificial intelligence programs, to identify and predict record to historic weather events. However, one weakness in this event was properly defining the threat area, based on EPS QPF for heavy rain and rainfall amounts. Model and therefore EPS QPF is an area where considerable progress still needs to be accomplished. 4. ACKNOWLEDGEMENTS Ron Holmes, National Weather Service in State College for archival of SREF data and rainfall map production. Jun Du of NCEP for SREF data for 25 June 2006 which was lost after the event. 5. REFERENCES Doswell, C. A. III, H. E. Brooks, and R. A. Maddox, 1996: Flash flood forecasting: An ingredients-based methodology. Wea. Forecasting, 11, 560-581. Doty, B. and J.L. Kinter III, 1992: The Grid Analysis and Display System (GrADS): A practical tool for research science visualization. International Conference on Interactive Information and Procession Systems, Atlanta, Georgia, 5-10 January, 1992. Grumm, R.H., and RE Hart. 2001: Standardized Anomalies Applied to
Significant Cold Season Weather Events: Preliminary Findings. Weather and Forecasting,16, 736 754. Grumm, R. H., and Hart R., 2001: Anticipating heavy rainfall events: Forecast aspects. Preprints, Symp. on Precipitation Extremes: Prediction, Impacts, and Responses, Albuquerque, NM, Amer. Meteor. Soc., 66 70. Hart, R.E., and R H. Grumm. 2001: Using Normalized Climatological Anomalies to Rank Synoptic-Scale Events Objectively. Monthly Weather Review,129,2426 2442. Hart, R., and Grumm R. H., 2001a: Anticipating heavy rainfall events: Climatological aspects. Preprints, Symp. on Precipitation Extremes: Prediction, Impacts, and Responses, Albuquerque, NM, Amer. Meteor. Soc., 271 274. Maddox, R. A., C.F. Chappell and L. R. Hoxit, 1979: Synoptic and Meso-alpha Scale Aspects of Flash Flood Events. Bull. of Amer. Meteor. Soc., 60, pp 115-123. Sivillo J.K, JE. Ahlquist and Z Toth. 1997: An Ensemble Forecasting Primer. Weather and Forecasting,12, 80 9 818. Zhang, F, C.Snyder, and R. Rotunno, 2003: Effects of moist convection on mesoscale predictability. JAS,60,1173-1184.
Figure 16 NAM 00-hour forecasts initialized at 0600 UTC 27 June 2006 showing (left) 850 hpa winds with U and V wind anomalies and (right) precipitable water and 1000 hpa winds and V-wind anomalies. Figure 17. As in Figure 16 except showing 500 hpa heights and 500 hpa temperatures and anomalies.