Figure 1. Official track of Hurricane Irma during its approach to the US coast. Red H icons are hurricane status, yellow is tropical storm status, and green is tropical depression or weaker. Full path from eastern Atlantic is in the appendix.

Hurricane Irma Forecasts and Heavy rainfall

by Richard H. Grumm

National Weather Service office State College, PA 16803. and

Corey Guastini National Centers for Environmental Predicts/Environmental Modeling Center

1. Overview Hurricane Irma began as an African Wave in the eastern Atlantic and developed into a tropical storm near the Cape Verde Islands on 30 August 2017. It then rapidly developed into a hurricane. Irma peaked as a Category 5 storm on the Saffir-Simpson scale and the peak winds, around 185 mph were reached on 6 September 20171. The storm came across the Florida Keys on 9 September as category 4 storm. The storm came ashore in in Cuba as category 5 storm on 8 September. Earlier in its life as a Category 5 storm it severely damaged the islands of Barbuda, Saint Martin, and the Virgin Islands.

The impacts of the storm included heavy rain, coastal flooding with the storm surge and wave action, and severe wind damage. In Florida, Georgia, and South Carolina the storm produced heavy rainfall (Fig. 2). The area heavy rainfall in the Florida Keys matched up with the eyewall. The heaviest rainfall on the Florida Peninsula was along and west of the track (Fig. 1) of the storm. There

1 Table of track and intensity data during peak time and used in Figure 1 is in a table in the Appendix.

was an extensive area of over 300 mm of rainfall in northeastern Florida. The data in Figure 2 imply that a significant percentage of the State of Florida had over 200 mm (8 inches) of rainfall. The rainfall amounts diminished dramatically as the surface cyclone weakened over Georgia and Mississippi. Across the lee islands, Cuba, and southern Florida included Irma packed a triple punch of damage with a mix of storm surge, extreme winds, and flooding due to rain. Many of the islands were devastated by winds over 100kts. The US Virgin Islands had extensive damage to homes and infrastructure due to the extreme winds. It was estimated that it could take months to restore power to most of the Islands inhabitants (WP 2017; CNBC 2017). The island of Barbuda was devastated by Irma (USA Today) damaging “an estimated 95% of the Barbuda’s infrastructure”. The island had to be evacuated due to the massive destruction to the infrastructure. Reports indicated that this was the first time in over 300 year that the island was not inhabited. Forecasts of Irma by the NCEP GEFS and EC EFS all indicated a powerful, long-lived storm. These forecasts facilitated advanced warnings about the storm and its potential impacts. There was considerable uncertainty with the storm relative to its intensity and track. But overall and similar to Harvey the forecast guidance provided useful guiadance related to the storm and its potential impacts with an extraordinary amount of lead-time.

Figure 2. Total estimated rainfall from 1200 UTC 9 September to 1200 UTC 12 September 2017. Values in mm as in the color bar. Rainfall had not ended by the time of this image. Return to text.

This paper will document the track and intensity of hurricane Irma. Forecasts from the NCEP, EC, and UKMO are presented to show the track forecasts and thus potential impact areas. The general westward track in more recent forecasts is self-evident. 2. Methods and data The Climate Forecast Center reanalysis data was used to show the pattern and evolution of the system. These data are far to course to draw conclusions related to tropical storms and hurricanes but they do serve as a dataset of record. These data were displayed using GrADS. The Stage-IV rainfall data were used to show 1-hourly, 6-hourly, 24-hourly, and storm total estimated rainfall amounts. These data were displayed in GrADS and Python. GEFS forecast data were from the NCEP website and were plotted using GrADS. The predictability horizon diagrams were produced using the same data but were processed with Python. Forecast tracks were provided using the NCEP/EMC data feed and software. These tracks show the uncertainty and potential impact of the storm.

Figure 3

3. Pattern and anomalies The large scale pattern over south Atlantic from 31 August to 5 September 2017 (Fig. 3) showed above normal heights which moved slowly to the northwest. A closed 5940 m anticyclone developed over the southern Atlantic. By 0000 UTC 4 September the upper level wave associated with Irma was visible with -6σ 500 hPa height anomalies. This overall pattern, with the anomalous anticyclone to the north, implied weak flow over the southern Atlantic. As the large anticyclone shifted east and weakened a second hurricane, Jose, developed and it too began its trek across the south Atlantic (Fig. 4). At the surface, Irma is shown approaching and moving over the lee Islands and Cuba (Fig. 5) from 0000 UTC 06 to 0000 UTC 11 September 2017. During this entire period the coarse CFSR data showed Irma with mean sea level pressure anomalies -6σ below normal. The CFSR also showed Jose to the east of Irma. Jose was a significantly smaller circulation relative to Irma. North of Irma, the strong subtropical ridge was present over the Atlantic. This feature weakened over time as it drifted to the east. Over the southern United States where was a deep trough which over time was replaced by a strong surface anticyclone. The changing flow to the north and west of the storm likely had an impact on the track and thus track forecasts of Irma. Note that by 0000 UTC 11 September there as a +2σ above normal pressure anomaly center over the Mid-Atlantic region. The precipitable water (PW: Fig 6) field showed that Irma, like most tropical cyclones, was collocated with a pool above normal PW. PW values would exceed 75mm at times as the storm moved westward. Jose too had a PW pool associated with its center. PW anomalies were generally +5 to +6σ above normal along Irma’s track. A frontal boundary was clearly evident in the western Atlantic off the coast of the United States. As Irma moved across the northern coast of Cuba (Fig. 7a-c) it began to make a turn to the north which took the storm over the Florida Keys (Fig. 7d) and up the West Coast of Florida. The pool of high PW air (Fig. 8) moved into southern Florida. The PW values peaked near 80 mm and +5σ above normal (Fig. 8). This likely contributed to the high rainfall amounts observed over the Florida Peninsula. Despite the deep cyclone, the PW values dropped as Irma moved northward into Florida and there was not additional moisture plume to tap. This lack of additional moisture likely contributed to the lack of heavy rainfall north and west of Georgia and South Carolina (Fig. 2). The landfalls on Florida are shown using 3-hourly 3km HRRR analyses. These data show the HRRR move toward and over the Florida Keys from 0000-1500 UTC 10 September (Fig. 9) and into Fort Meyers during the period through 1800 UTC and 0900 UTC 11 September (Fig. 10). The storm remained quite strong and quite circular as it moved over Florida. 4. Forecast Tracks Global ensemble track forecasts are used to show the evolution of the forecasts of Irma. There are too many forecast times to show so select times critical for landfall issues are addressed here. Figure 11 shows the tracks from the GEFS, EC_EFS, and UK_EFS initialized at 0000 UTC 2 September with forecast through 0000 UTC 12 September 2017. The UK forecasts do not extend out far enough in time. But both the EC_EFS and GEFS forecast an Atlantic track with the storm moving up the East Coast or along the East Coast. The lower panel show the intensity of the cyclone as observed, forecast by the high resolution deterministic model of each system and each member of the 3 ensemble forecast systems.

The trend for the next few cycles at 1200 and 0000 UTC showed a shift to the west and higher probability of Florida impact. Forecasts issued at 0000 UTC 4 September 2017 (Fig. 12) showed the GEFS had most members bringing the storm up the Florida Peninsula. A few members still favored an offshore track to the east. The EC-EFS favored a Florida landfall. All three systems showed the relatively consistent track over the Leeward Islands and a track close to the northern coast of Cuba. The trend for a more westward track continued as shown in the forecasts from 0000 UTC 6 September (Fig. 13). This trend continued and as expected as the forecast length decreased the predictability increased thus forecasts issued at 0000 UTC 9 September focused the landfall over the Florida Keys and then up the West Coast of Florida with a second landfall along the southwestern coast of Florida. 5. Quantitative Precipitation Forecasts

6. Summary Irma was a very strong long-lived hurricane that will likely set many Atlantic hurricane records. It remained a Category -5 storm for a long time; it did extreme wind and surge damage to many Caribbean Islands. It brought high winds and flooding to Haiti and Cuba. Finally, the storm made landfall on the Florida Keys and then moved inland near Fort Meyers, Florida. The impacts of Irma across the lee islands, Cuba, and Florida included a mix of storm surge, extreme winds, and flooding due to rain. Many of the islands were devastated by winds over 100kts. The extreme rainfall with Hurricane Harvey in Texas dwarfed the damage done by coastal flooding and winds. Rain was problem serious problem in Florida with Irma though the rainfall amounts were about 40% of those observed in Texas. In Florida, advanced knowledge of the approaching storm resulted in one of the largest mass evacuations in US history. The uncertainty associated with the storm likely played a critical role in the size of the evacuation efforts. But the impact of the wind, heavy rainfall, and coastal flooding was likely mitigated to some degree due do these massive evacuations. Better forecasts may have provided better knowledge on what areas to evacuated but this may be about as good as it gets. The forecasts were generally consistent with the EC showing a faster trend toward the West Coast of Florida relative the GEFS and UKEFS. But overall all three systems did extremely well slowly refining the forecasts and tracking the storm ever closer to the observed track.

7. Acknowledgments Lance Bosart for editorial comments and suggestions not all which could be reproduced here. The Albany MAP for insights and data links.

Figure 4. Return to text.

Figure 5. As in Figure 4 except for mean sea level pressure (hPa) every 24 hours from a) 0000 UTC 6 September through f) 11 September 2017. Values in hPa contoured every 4 hPa. Return to text.

Figure 6. As in Figure 5 except for precipitable water (mm) and precipitable water anomalies. Return to text.

Figure 7. As in Figure 5 except every 12 hours from a) 0000 UTC 09 to f) 1200 UTC 11 September 2017. Return to text.

Figure 8. As in Figure 7 except for precipitable water (mm). Return to text.

Figure 9. As in Figure 6 except for 3km HRRR 00-hour analysis showing mean sea level pressure and pressure anomalies in 3hr increments from a) 0000 UTC through f) 1500 UTC 10 September 2017. Return to text.

Figure 10. As in Figure 9 except for a) 1800 UTC 10 September through f) 0900 UTC 11 September 2017. Return to text.

Figure 11. Track forecasts from the NCEP GEFS, EC EFS, and UK EFS initialized at 0000 UTC 2 September 2017 showing the tacks of each ensemble member. In each panel the thin lines show each member and the thick lines show the observed track (black), ensemble mean track (as in each key) and the deterministic forecast from each forecast system. The lower right panels shows the intensity forecasts for each member of each system and the deterministic runs.. Return to text.

Figure 12. As in Figure 11 except initialized at 0000 UTC 4 September 2017. Return to text.

Figure 13. As in Figure 11 except initialized at 0000 UTC 6 September 2017. Return to text.

Figure 14. As in Figure 11 except initialized at 0000 UTC 9 September 2017. Return to text.

Appendix

Lat Lon Wind Pressure Storm Type Category 16.7° -57.7° 175 mph 929 mb Hurricane 5 16.8° -58.4° 180 mph 931 mb Hurricane 5 16.9° -59.1° 185 mph 926 mb Hurricane 5 17.1° -59.8° 185 mph 926 mb Hurricane 5 17.2° -60.5° 185 mph 916 mb Hurricane 5 17.2° -60.5° 185 mph 916 mb Hurricane 5 17.4° -61.1° 185 mph 916 mb Hurricane 5 17.7° -61.8° 185 mph 914 mb Hurricane 5 17.9° -62.6° 185 mph 914 mb Hurricane 5 18.1° -63.3° 185 mph 918 mb Hurricane 5 18.2° -64.0° 185 mph 918 mb Hurricane 5 18.3° -64.2° 185 mph 922 mb Hurricane 5 18.4° -64.5° 185 mph 920 mb Hurricane 5 18.5° -64.7° 185 mph 920 mb Hurricane 5 18.6° -64.9° 185 mph 920 mb Hurricane 5 18.7° -65.1° 185 mph 920 mb Hurricane 5 18.8° -65.4° 185 mph 914 mb Hurricane 5 18.9° -65.6° 185 mph 914 mb Hurricane 5 19.0° -65.8° 185 mph 914 mb Hurricane 5 19.1° -66.1° 185 mph 914 mb Hurricane 5 19.2° -66.3° 185 mph 916 mb Hurricane 5 19.3° -66.6° 185 mph 916 mb Hurricane 5 19.4° -66.8° 185 mph 916 mb Hurricane 5 19.5° -67.1° 185 mph 918 mb Hurricane 5 19.6° -67.4° 185 mph 918 mb Hurricane 5 19.7° -67.7° 180 mph 921 mb Hurricane 5 19.7° -67.9° 180 mph 921 mb Hurricane 5

Appendix of Hurricane Irma Track data focused on the track near time of maximum intensity. This and all previous and later data were used to plot the tracks and rate the storm in the text.