Southern Heavy rain and floods of 8-10 March 2016 by Richard H. Grumm National Weather Service State College, PA 16803 1. Introduction Heavy rains (Fig. 1) produced record flooding in northeastern Texas and northwestern Louisiana 9-11 March 2016. The period of heaviest rain was the 24 hour period from about 1800 UTC 8-9 March 2016. Some areas of northwestern Louisiana had over 300 mm (11 inches). Point maximum data indicated during the longer period of time rainfall may have exceed 20 inches in some locations. The heavy rains produced record flooding along the Sabine River which runs along the Texas and Louisiana border. Record crests along the Sabine were measured at several locations including Burkeville, Deweyville, and Bon Wier (TWC, USA Today, AP 2016). Several of the sites broke records which were had stood for over 100 years. This paper will document the pattern in which the heavy rainfall occurred and present methods which may aide anticipating this and similar events. The focus is on using climate data to identify the potential for heavy rainfall including the pattern in which the event occurred and the value of Average recurrence intervals (ARI: NOAA14) to characterize the observed and forecast Figure 1. Total QPE form the Stage-IV 6-hourly data files for the periods of a) 1800 UTC 8-9 March and the 48 hour period from b) 1800 UTC 8 to 10 March. Values in mm as indicated by the color bar with contours every 50 mm. Return to text.
precipitation. The estimate precipitation (QPE) and forecast precipitation (QPE) are shown along with ratios relative to the ARI values. Forecasts from several NCEP forecasts systems are also shown relative to the ARI values as ratios. 2. Methods and data The climate forecast system re-analysis (CFSR) data was used to reconstruct the pattern and the standardized anomalies associated with the event. The CFSR is used to show the pattern, which forecasters often use to gain confidence in a potential significant weather event. These same patterns, when forecast may produce high end QPF which may reinforce confidence in the forecast. The Stage-IV rainfall data (Seo 1998) was used to estimate the rainfall over 6-hour intervals and produce accumulations in 12, 24, and 48 hour intervals. The NOAA14 data were used to produce the recurrence intervals. The focus here was on 100 year ARI values of 6 and 24 hour durations. The 6 and 24 hour durations at 100 recurrence intervals have been seamed together with NOAA40 data to cover the entire USA. Other intervals do not cover Texas and most of the western United States and thus were used limitedly herein. The NCEP GEFS, GFS, and 3km HRRR was used to forecasts of the extreme QPF. 3. Results a. The pattern The 500 hpa heights and anomalies (Fig. 2) and the 250 hpa winds (Fig. 3) showed a textbook heavy rainfall pattern. The 500 hpa trough moving across the southwestern United States deepened to -6σ below normal before entering Texas. To the east a larger 500 hpa ridge with +2 to +3σ heights was present. Between the deep trough to the west and strong ridge to the east, deep southerly flow developed. Ahead of the developing trough on the 8 th and 9 th there was a surge of high precipitable water (PW: Fig 4) into east Texas and Louisiana along with a strong low-level 850 hpa jet (Fig. 5). Based on decades of research on heavy rainfall (Junker and Schneider 1997; Junker et al. 1999; Grumm and Hart 2000; Hart and Grumm 2000) this was an ideal pattern for heavy rain, often classified as a Maddox-Synoptic event (Maddox et al 1979) with deep southerly flow and a plume of relatively deep moisture. The standardized anomalies often help characterize the potential for heavy rainfall (Grumm 2011). The pattern produced the heavy rainfall on 8-10 March (Fig. 1) was dominated by periods of heavy rainfall on the 8 th and 9 th (Fig. 5) which persisted into the 10 th (Fig. 6). The 2 day rainfall total (Fig. 1) showed areas in Louisiana where rainfall exceeded 500 mm (20 in). The multi-day event produced the heaviest rain in Texas and Louisiana on the 8-9 th and heavier rain in Arkansas on the 9-10 th.
b. GEFS forecasts Ensembles allow the forecaster to assign probabilities to an outcome and to visualize uncertainty based on the variation in the forecasts amongst forecast members. When the forecasts systems get the large scale pattern correct they often contain useful signals. In this event the NCEP GEFS did particularly well with the pattern and thus predicted an enduring heavy rainfall event in the general region where the heavy rainfall was observed. The GEFS 24 and 48hour probabilities of 100 mm (4 inches) or more QPF are showing in Figures 7 & 8. The 24 hour forecasts ending at 1800 UTC 9 March indicated heavy rain would fall over eastern Texas and potentially into western Louisiana. The axis of the heavy rainfall was too far west relative to observations. The internal GEFS QPF climate (M-Climate: Fig. 9) from the 1200 UTC 6 March 2016 GEFS forecast cycle suggested that 4 and 4-5 inches or more QPF in 24 and 48 hours respectively was an extreme event in the GEFS QPF climatology. Other GEFS forecasts were equally impressive with one time shown here for illustrative purposes. c. Deterministic forecasts-ncep GFS M-Climate data for the NCEP GFS does not exist, therefore the ARI data was used to gage the NCEP GFS QPF to determine the potential for an extreme rainfall event. Six GFS forecasts valid at 1800 UTC 9 March (Fig. 10) and 0000 UTC (Fig. 11) show the GFS QPF (inches) and the ratio of the QPF/ARI as a percentage. Similar to the GEFS, the GFS did not get the location of the heavy rainfall correct but it got the correct region. The QPF varied between GFS forecasts cycles and longer range forecasters generally forecast higher QPF amounts. This could be a spin-up issue in the model for the shorter range forecasts. The salient point here is that the GFS forecast 4 to 8 inches of QPF which at times represented 75 to 100% of the 100 year recurrence interval data for the 24 hour duration. The GFS was showing a potentially heavy rainfall event with nearly all forecasts indicating at least 50% of the 24h/100yr ARI values. The forecast valid at 0000 UTC show that some of the issues with QPF/ARI ratios were related to timing differences between GFS cycles. The matching QPE/ARI ratios for the period ending at 1800 UTC 9 March 2016 (Fig. 12) show that the Stage-IV data had over 12 inches of rainfall, higher than GFS forecasts, and the QPE/ARI ratios were 100 to 125% indicating the verifying QPE exceed the 24h/100yr ARI values. d. NCEP 3km HRRR Short range forecasts from the 3km NCEP HRRR during the event combined with the QPF/ARI ratios suggested that the HRRR has some potential in stronger forced events to show areas where the rainfall is approaching or exceeding specified ARI thresholds. Figure 13 shows 6 HRRR forecasts initialized from 0000 to 0600 UTC 9 March. These HRRR forecasts showed 4 to 6 inches of QPF over northwestern Louisiana. At times the 6h/100yr ARI ratios were 100 to
125% implying that the HRRR was forecasts 6hour rainfall amounts in greater than the 6h/100yr ARI. 4. Conclusions A high impact rainfall event impacted portions of Texas, Louisiana, Arkansas, and Mississippi on 8-10 March 2016. The heaviest rain was observed in northwest Louisiana (Fig. 1 & 12). The heavy rain produced flooding with record flooding along the Sabine River along the Texas and Louisiana border. The flooding damaged thousands of houses (AP 2016a) and damaged many roads and bridges. This was a relatively well predicted high impact rain and flood event by the NCEP GEFS and GFS. The GEFS forecast a record event. The 24 and 48 hour QPF values showed a high probability of over 100 mm of QPF in portions of eastern Texas and Louisiana (Figs. 8-9). These forecasts when examined the GEFS M-Climate show that the GEFS forecasts of 100 mm or more QPF were an extreme event verse the GEFS internal QPF climatology. Whenever a forecast system forecasts a record event relative to its internal climatology the potential for a rare event is something forecasters should be mindful of. Similar to the NCEP GEFS, the GFS showed heavy rain in the correct geographic area. The GFS is of higher resolution than GEFS and thus forecast higher QPF amounts. M-Climate data does not exist for the GFS so QPF/ARI ratios were used. These data showed that the GFS forecast between about 50 and 100% of the 24h/100yr ARI values which was lower than the observed 24h/100yr ARI values which the Stage-IV data suggested were in the 100 to 125% range over portions of northwest Louisiana. The HRRR forecasts were shown relative to the 6h/100yr ARI data. The HRRR showed that the QPF at times was forecast to be 100 to 125% of the 6h/100year ARI values. The HRRR QPF forecasts examined were generally closer to the observed rainfall locations relative to those produced by the GEFS and GFS. The more accurate location forecasts were likely the result of the models resolution and the relatively short forecast lengths of the HRRR relative to the GEFS and GFS. 5. Acknowledgements 6. References Associated Press 2016a: Louisiana, Mississippi: Thousands of homes damaged by floods. AP 2016 and similar stories. Associated Press 2016b-: Unusually widespread flooding across Louisiana, Mississippi. AP 2016 and similar stories
Maddox,R.A., C.F Chappell, and L.R. Hoxit. 1979: Synoptic and meso-alpha aspects of flash flood events. Bull. Amer. Meteor. Soc.,60,115-123. Grumm, R. H. 2011: New England Record Maker Rain Event of 29-30 March 2010. National Weather Association, Electronic Journal of Operational Meteorology, 2011-EJ4 Grumm, R.H. and R. Hart, 2000: Anticipating heavy rainfall events: Climatological aspects. Preprints, Symposium on Precipitation Extremes: Prediction, Impacts, and Responses, Albuquerque, New Mexico, Amer. Meteor. Soc. 66-70. Hart R., and R.H. Grumm, 2000: Anticipating heavy rainfall events: Forecast aspects. Preprints, Symposium on Precipitation Extremes: Prediction, Impacts, and Responses, Albuquerque, New Mexico, Amer. Meteor. Soc. 271-275. Junker, N.W. and R.S Schneider, 1997: Two case studies of quasi-stationary convection during the 1993 great Midwest flood. National Weather Association Digest., 21,5-13., R.S. Schneider, and S.L. Fauver, 1999: A Study of heavy rainfall events during the great Midwest flood of 1993. Wea. Forecasting., 14, 701-712. Seo, D.J., 1998: Real-time estimation of rainfall fields using rain gauge data under fractional coverage conditions. J. of Hydrol., 208, 25-36.
Figure 1. CFS 500 hpa heights (m) and 500 hpa height anomalies in 12 hour increments from a) 0000 UTC 8 March through f) 0000 UTC 11 March 2016. Heights very 60 m. Return to text.
Figure 2. As in Figure 2 except for 250 hpa winds (ms-1). Return to text.
Figure 3. As in Figure 2 except for precipitable water (mm) in 6 hour increments from a) 1800 UTC 8 March through f) 0000 UTC 10 March 2016. Contours every 5 mm. Return to text.
Figure 4. As in Figure 4 except for 850 hpa winds (ms-1) and 850 hpa wind anomalies. Return to text.
Figure 5. Rainfall in 6-hour increments from a) 1200 UTC 8 March through e) 1200 UTC 9 March 2016 and f) the total QPE for the period in mm. Return to text.
Figure 6. As in Figure 6 except for the period covering 1200 UTC 9 to 1200 UTC 10 March 2016. Return to text.
Figure 7. NCEP GEFS forecasts showing the probability of QPE greater than or equal to 100 mm in the 24 hour period ending at 1800 UTC 09 March from GEFS forecasting initialized at a) 0000 UTC 4 March, b) 0000 UTC 5 March, c) 0000 UTC 6 March, d) 0000 UTC 7 March, e) 0000 UTC 8 March, and f) 1200 UTC 8 March 2016. Return to text.
Figure 8. As in Figure 7 except for the 48 hour period ending at 1800 UTC 10 March 2016. Return to text.
Figure 9. NCEP GEFS forecast of QPF showing the ensemble mean QPF and the QPF relative to the GEFS M-Climate as a percent of the climatological value. Upper panel is the 24 hour forecast valid 1800 UTC 9 March and lower panel is the 48 hour forecast valid at 1800 UTC 10 March 2016. Return to text.
Figure 10. NCEP GFS forecasts of QPF (contours in inches) for the 24 hour period ending at 1800 UTC 9 March 2016 and the ratio of the QPF to the 24h/100yr ARI as a percentage as per the color bar. Forecasts include those initialized at a) 0000 UTC 7 March, b) 1200 UTC 7 March, c) 1800 UTC 7 March, d) 0000 UTC 8 March, e) 0600 UTC 8 March and f) 1200 UTC 8 March 2016. Return to text.
Figure 11. As in Figure 10 except for the 24 period ending at 0000 UTC 10 March 2016. Return to text.
Figure 12. Stage-IV QPE (inches) and the ratio of the 24 hour QPE to the 24h/100yr ARI values. Data valid for the 24 hour period ending at 1800 UTC 9 March 2016. Return to text.
Figure 13. As in Figure 9 except for NCEP 3km HRRR forecasts of 6 QPF (inches) and the ratio of the QPF to the 6h/100yr. Return to text.