Development of a Tropical Cyclone Probabilistic Rainfall Model

Development of a Tropical Cyclone Probabilistic Rainfall Model Brian Mc. Noldy 1, Mu-Chieh Ko 2, and Frank Marks 3 1 University of Miami/RSMAS 2 University of Miami/CIMAS 3 NOAA/AOML NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs

Record-Setting Hurricane Rainfall 2017 -18 • Hurricanes Harvey, Florence, and Lane have each set state records for tropical cyclone rainfall with Harvey’s rainfall of 60+ inches setting the U. S. record Harvey (2017) - 60. 58 in Texas & U. S. Record Florence (2018) – 35. 93/26. 63 in North Carolina/South Carolina Record NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs Lane (2018) – 58. 00 in Hawaii Record 2

HFIP Goals & Metrics: Rainfall Goal 4. 5 Metric 4. 5 Baseline Target Goal 4. 6 Metric 4. 6 Baseline Target Goal 4. 7 Metric 4. 7 Baseline Target Improve accuracy and lead time of the WPC Excessive Rainfall Outlook for TCs. Brier Score of Day-3 Excessive Rainfall Outlook for landfalling Atlantic basin TCs Current Brier Score of Day-3 Excessive Rainfall Outlook, 2015 -17 CONUS-landfalling TCs. Current Brier Score of Day-2 Excessive Rainfall Outlook: Improve skill of Quantitative Precipitation Forecasts (QPF) for landfalling TCs. QPF Brier Score for TCs affecting CONUS, Puerto Rico, and USVI 2015 -17: 10% improvement over baseline Create a probabilistic TC QPF product based on HAFS/HWRF ensemble output. Disseminate probabilistic TC QPF for CONUS, Puerto Rico, & USVI TC threats. N/A Disseminate probabilistic TC QPF for CONUS, Puerto Rico, & USVI TC threats. NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 4

Parametric Modeling of TC Rainfall • R-CLIPER: Rainfall CLImatology & PERsistence • Marks & De. Maria (2003), Tuleya et al. (2007) • Accounts for intensity, size, speed, land, but not asymmetry or topography • Run experimentally at NHC 2001 -2003, operationally since 2004 • PHRa. M: Parametric Hurricane Rainfall Model • Lonfat et al. (2007) • Builds on R-CLIPER framework, but adds asymmetry and topography • Intensity and shear dependent parameterization of rainfall derived from TRMM data (Lonfat et al. 2004) • Rainfall Probability: Probabilistic PHRa. M • Utilizes NHC’s 1000 -member Monte Carlo ensemble used for wind speed probabilities • PHRa. M is run on the 1000 members to get probabilistic information • Computationally reasonable to run in real-time… how does it compare to wind speed probabilities? NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 5

R-CLIPER • 2121 TC cross-sections of rainfall collected from TRMM during 1998 -2003 • • 10 -km bins to 500 km Partitioned by TS, H 12, H 345 Partitioned by ocean basin Piecewise formulation used in R-CLIPER replaced by dualexponential EXAMPLE: H 345 vmax = 115 kt rmw = 20 km H 12 vmax = 75 kt rmw = 36 km TS vmax = 47 kt rmw = 47 km • Rain rate scales continuously with intensity, Vmax ≥ 35 kt, Rmax ≤ 100 km NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs Marks and De. Maria (2003) 6

PHRa. M • c proportionality constant, Vs 10 m wind field, & hs is elevation. Use Willoughby et al. (2006) wind model • n is exponent for power law inside Rmax (=1), X 1 is an exponential decay length in outer vortex (=250 km). • Wind field reduced to 10 -m by taking 85% of estimates • Inflow angle is not accounted for NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs Lonfat et al. (2007) 7

TC Ensemble • Start with operational Monte Carlo 1000 -member ensemble (De. Maria et al. 2009) • Includes uncertainties in track, intensity, & size randomly selected from NHC error distributions over past 5 years • Used to generate 34 kt, 50 kt, and 64 kt wind speed probabilities. • Why not reuse the realizations for rainfall probabilities? NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 8

Ensemble PHRa. M • Example for best-track position, intensity, shear, Rmax for Florence (2018) from 11 Sep 12 Z through 16 Sep 12 Z… begins three days before landfall • Rmax values from Extended Best-Track. NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 9

Ensemble PHRa. M • Example forecast values of position, intensity, shear Rmax for Florence (2018) from 11 Sep 12 Z through 16 Sep 12 Z. • Rmax values calculated using Knaff et al. (2015) empirical relationship which is function of Vmax & latitude NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 10

Florence Excessive Rainfall Outlook Day 1 Day 2 Day 3 NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 11

Rainfall Probability • Ensemble-based products include an ensemble mean, probability of exceeding a fixed amount, continued… Day 3 NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 12

Rainfall Probability • Probability of exceeding deterministic forecast by some amount, area with % chance of exceeding deterministic forecast Day 3 NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 13

Future Scenarios • Calculate linear trend of track errors over past decade at each lead time • Project that % into next decade, with optional +/- increase due to progress • e. g. : average 23% improvement in 72 h track forecasts from 2008 -2018 • Suppose progress will be 20% greater in coming decade… 28% improvement in 72 h track forecasts from 2018 -2028. 2008 -2018 72 h NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 2018 -2028 72 h 14

Future Scenarios • Calculate linear trend of intensity errors over past decade at each lead time • Project that % into next decade, with optional +/- increase due to progress • e. g. : average 29% improvement in 72 h intensity forecasts from 2008 -2018 • Suppose progress will be 20% greater in coming decade… 35% improvement in 72 h intensity forecasts from 2018 -2028. 2008 -2018 -2028 NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 15

Future Scenarios • We can also generate a Monte Carlo ensemble taking these reductions in uncertainty into account and arrive at hypothetical future rainfall probability products as well! 2018: operational NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 16

Future Scenarios • We can also generate a Monte Carlo ensemble taking these reductions in uncertainty into account and arrive at hypothetical future rainfall probability products as well! 2028: hypothetical NOAA Hurricane Forecast Improvement Project Meeting the Nation’s Needs 17