HWRF Ocean: The Princeton Ocean Model Isaac Ginis Graduate School of Oceanography University of Rhode Island HWRF Tutorial NCWCP, College Park, MD 23-25 January 2018 1 1
Why Couple a 3-D Ocean Model to a Hurricane Model? To create accurate SST during hurricane model integration Evaporation (moisture flux) from sea surface provides heat energy to drive a hurricane Available energy decreases if storm-core SST decreases Uncoupled hurricane models with static SST neglect SST cooling during integration à high intensity bias One-dimensional (vertical-only) ocean models neglect upwelling and horizontal advection, both of which can impact SST during integration 2
Early history of Princeton Ocean Model Three-dimensional, primitive equation, numerical ocean model (commonly known as POM) Originally developed by Alan Blumberg and George Mellor in the late 1970 s Initially used for coastal ocean circulation applications Open to the community during the 1990 s and 2000 s https://en.wikipedia.org/wiki/princeton_ocean_model 3
Developing POM for Tropical Cyclones Available POM code version transferred to University of Rhode Island (URI) in 1994 POM code changes made at URI specifically to address ocean response to hurricane wind forcing This POM version coupled to GFDL hurricane model at URI Coupled GFDL/POM model operational at NWS in 2001 Additional POM upgrades made at URI during 2000 s (e.g. initialization) and implemented in operational GFDL/POM Same version of POM coupled to operational HWRF in 2007 New version created in 2014: Message Passing Interface POM for tropical cyclones (MPIPOM-TC) 4
Developing MPIPOM-TC 5
MPIPOM-TC Features MPIPOM-TC uses MPI software to run efficiently on multiple processors, allowing for both higher grid resolution and a larger ocean domain than POM-TC MPIPOM-TC accepts flexible initialization options MPIPOM-TC is an adaptation of sbpom, which has community support and includes 18 years of physics updates and bug fixes MPIPOM-TC is a modernized code with NetCDF I/O MPIPOM-TC uses a single prognostic code in all worldwide HWRF ocean basins 6
MPIPOM Domains Worldwide All domains are set to the same size covering 83.2 (37.5 ) of longitude (latitude) with a horizontal grid spacing of 1/12 o. By default, HWRF runs coupled in all ocean basins except Southern Hemisphere. DTC will provide instructions on how to run coupled in all basins. 7
POM Numerical Design 8
Sigma Vertical Coordinate 40 vertical sigma levels; free surface (η) Level placement scaled based on ocean bathymetry Largest vertical spacing occurs where ocean depth is 5500 m 9
Arakawa-C Grid: External Mode Plan View Horizontal spatial differencing occurs on staggered Arakawa-C grid 2-D variables UA and VA are calculated at shifted location from η 10
Arakawa-C Grid: Internal Mode Plan View Horizontal spatial differencing occurs on staggered Arakawa-C grid 3-D variables U and V are calculated at shifted location from T and Q T here represents variables T, S, and RHO Q here represents variables Km, Kh, Q2, and Q2l 11
Vertical Grid: Internal Mode Elevation View Vertical spatial differencing also occurs on staggered grid 3-D variables W and Q are calculated at shifted depth from T and U T here represents variables T, S, and RHO Q here represents variables Km, Kh, Q2, and Q2l 12
Time Stepping POM has a split time step External (two-dimensional) mode uses short time step: 22.5 seconds during pre-coupled initialization 13.5 seconds during coupled integration Internal (three-dimensional) mode uses long time step: 15 minutes during pre-coupled initialization 9 minutes during coupled integration Horizontal time differencing is explicit Vertical time differencing is implicit 13
POM Physics 14
1-D processes Hurricanes cool the ocean surface through: (1) vertical (turbulent) mixing (2) surface heat flux Vertical mixing drives ~85% of sea surface temperature cooling Storm center Turbulent Mixing 15
Vertical Mixing Parameterization Turbulent flux terms are assumed proportional to the vertical shear of the mean variables, e.g. Momentum # w'u'(z) = K % u $ z Temperature & # ( w'θ '(z) = K θ & % ( ' $ z ' The turbulent mixing coefficient K is parameterized using either (1) Mellor Yamada level 2.5 turbulence closure model (M-Y scheme) or (2) K-Profile Parameterization (KPP scheme) : K(z) = hwg(z) h - mixing layer depth W- turbulent velocity scale G(z)- non-dimensional shape-function 16
3-D processes Strong Winds Hurricane induced upwelling and horizontal advection can enhance and/or modify surface cooling. depth (m) Horizontal Advection Upwelling km 17
Effect of Ocean Stratification on SST cooling Typical of Gulf of Mexico in Summer & Fall Typical of Caribbean in Summer & Fall 18
Effect of TC translation speed on SST cooling 2.4 m s -1 4.8 m s -1 Hurricanes have historically propagated in the Gulf of Mexico: < 5 m s -1 73% and < 2 m s -1 16% of the time And in the western tropical North Atlantic: < 5 m s -1 62% and < 2 m s -1 12% of the time So 3-D effects (e.g. upwelling) are important Yablonsky and Ginis (2009) 19 19
3-D 1-D 2.4 m s -1 3-D 1-D 4.8 m s -1 20
POM Initialization 21
Flexible Initialization Options 1. Feature-based modifications to U.S. Navy s Generalized Digital Environmental Model (GDEM) monthly temperature (T) and salinity (S) climatology with assimilated daily SST (FB) used in NATL. 2. Global Real-Time Ocean Forecast System (RTOFS) used in EPAC & CPAC. 22
Feature-based (FB) initialization The basic premise of the FB procedure is that major oceanic fronts and eddies in the western North Atlantic Ocean are poorly represented by GDEM climatology. By defining the spatial structure of these fronts and eddies using observations gathered from field experiments, crossfrontal sharpening of GDEM T & S fields is performed to increase the horizontal density gradients across the fronts. Algorithms are incorporated to initialize the Gulf Stream and Loop Current with prescribed paths and to insert eddies into the Gulf of Mexico based on guidance from near-real-time observations, such as satellite altimetry. 23
Feature-based Initialization: Gulf Stream Temperature at 400 m climatology after initialization Gulf Stream cross-section climatology after initialization Falkovich, A., I. Ginis, and S. Lord, 2005 24
Feature-based Initialization: Hurricane Irma 25
Feature-based Initialization: Gulf of Mexico 75-m Temperature (Sept. GDEM) TPC 26 C depth on 15 Sept. 2005 (1) Start with monthly climatology (3) Assimilate SST analysis 75-m Temperature AFTER assimilation (2) Adjust LC position & add warm- and cold-core eddies info derived from altimetry Future Hurricane Rita track (4) Spin-up ocean currents Yablonsky and Ginis (2008) 26
Feature-based Initialization: Examples of improved vertical temperature in Loop Current and warm-core eddy Loop Current Warm-Core Eddy Yablonsky and Ginis (2008) 27
Effect of Loop Current on Hurricane Katrina Intensity Central Pressure GFDL forecast Initialized Aug. 26, 18Z Climatological Loop Current Initialized Loop Current 28
Atmosphere-Ocean Coupling Wind speed (U a ) Temperature (T a ) Humidity (q a ) Atmospheric Model Air-Sea Interface Momentum flux (τ) Sensible heat flux (Q H ) Latent heat flux (Q E ) Surface current (U s ) SST (T s ) Momentum flux (τ) Ocean Model τ = ρ a C ( D U a U )( s U a U ) s Q H = C H ( U a U s )( T a T s ) Q E = L C V P C E ( U U )( q q ) a s a s 29
Developing Air-Sea Interface Module (ASIM) with explicit wave coupling Motivation: air-sea fluxes and turbulent mixing above/below sea surface are significantly modified by surface waves in high wind conditions. Image courtesy of Fabrice Veron 30
Wind-Wave-Current Interaction In Hurricanes Wind Sea state Surface waves!! τ air τ c τ! c τ! air Current Current τ! air Ocean currents Atmosphere Ocean Fan, Y., I. Ginis, and T. Hara, 2010 31
Sea State Dependent Drag Coefficient in WW3-MPIPOM coupled model RMS= 70 km, U 10max = 65 m/s U T =10 m/s U T =5 m/s Reichl, Hara, and Ginis, 2014 32
Sea State Dependent Langmuir turbulence in KPP 1. Sea state enhancement of K based on the Langmuir number K(z) = [ hw * G(z) ] F(La) La = u * / u St 2. Use the Lagrangian current in place of the Eulerian current in KPP Eulerian current Stokes drift Lagrangian current Reichl et al, 2016 33
Sea State Dependent Langmuir turbulence in KPP The enhancement factor needed in the KPP mixing coefficient is found by running a Large Eddy Simulation (LES) model 0 U 10 : 58.20 m s 1 LES 0 U 10 : 67.23 m s 1 KPP σ 0.5 Enhanced KPP 0.5 1 1 0 0.2 0 0.2 mixing 1 : 36.42 coefficient 1 m s mixing : 42.73 coefficient m s Reichl et al, 2016 34
Langmuir impact on the ocean response Stress (N/m 2 ) τ 0 (N/m 2 ) 0 1 2 3 4 5 6 200 An ideal hurricane is translated westward. km N 0 200 Stationary Slow Fast 200 0 200 200 0 200 U s (m s 1 ) Stokes drift (m/s) 200 0 200 Three translation speeds: 0 (stationary), 2.7 (slow), and 5.8 (fast) m/s. km N 200 0 0 0.1 0.2 0.3 0.4 0.5 Initial temperature profile: Mixed Layer Depth: 20m dt/dz=-.1 C/m 200 200 0 200 200 0 200 Langmuir number La 200 0 200 0.3 0.4 0.5 0.6 0.7 0.8 200 km N 0 200 200 0 200 200 0 200 200 0 200 Reichl et al, 2016 35
k 200 0 200 1 200 0 200 0 200 400 Current (m(m/s) s 1 ) 0 200 400 0 200 400 600 2 m/s 0 0.5 1 1.5 2 2.5 0 200 400 600 km N km N Depth Dpt (m) km from km N center 200 Langmuir impact on the ocean response Sheer-driven turbulence only (KPP-ST) Sheer-driven and Langmuir turbulence (KPP-LT) 200 0 200 0 200 0 200 400 0 200 400 600 1ms 1 Current (m s 1 ) 0 0.5 1 1.5 2 2.5 km from center U (m s 1 ) 200 200 0 0.5 1 1.5 2 2.5 0 200 0 200 0 200 400 0 200 400 600 200 50 100 0 200 0 200 SST ( C) 10 15 20 25 30 0 200 400 0 200 400 600 150 Stationary Slow Current (m s 1 ) Fast 1ms 200 200 0 0 2000.5 200 1 0 200 1.5 200 0 2 200 2.5 200 Depth Dpt (m) U (m s 1 ) 200 0 0.5 1 1.5 2 2.5 200 0 200 200 0 0 200 200 400 200 0 0 200 km 200 400 600 E km 50 100 150 200 200 0 200 200 0 200 200 0 200 km from center Reichl et al, 2016 36
Langmuir impact on the ocean response Difference in current Difference in temperature SST ( C) ( C) 0.5 0.5m/s m s 1 0.7 0.35 0 0.35 0.7 KPP-LT minus KPP-ST KPP-LT-CS - KPP-nw-CS km N 200 0 3D 200 Stationar y 200 0 200 Slow 0 200 400 Fast 0 200 400 600 KPP-LT-CS - KPP-nw-CS, 1d km N 200 0 1D 200 Stationar y 200 0 200 Slow 0 200 400 KPP-LT-CS - KPP-nw-CS, 3d Fast 0 200 400 600 Reichl et al. 2016 37
Coupled WW3-MPIPOM Modeling of Hurricane Edouard (2014) Model wind fields based on TC vitals Modeled significant wave heights Blair et al. 2017 38
Coupled WW3-MPIPOM Modeling of Hurricane Edouard (2014) 12 September 12 September Blair et al. 2017 39
Coupled WW3-MPIPOM Modeling of Hurricane Edouard (2014) Transect B Transect A Comparisons between model and observations of SST and mixed layer depth after the storm, on September 17th,1800 UTC. Blair et al. 2017 40
Wave-dependent physics in HWRF U c SST R air Q air Atmosphere Z ref α γ U ref Rib H s c p τ air KE D Ocean τ diff τ cor λ U c U λ U s Waves Ø Atmospheric model: air-sea fluxes depend on sea state Ø Wave model: forced by sea state dependent wind forcing Ø Ocean model: forced by sea state dependent wind stress modified by growing or decaying wave fields and Coriolis-Stokes effect. Turbulent mixing is modified by the Stokes drift (Langmiur turbulence). 41
References Ginis, I., 2002: Tropical cyclone-ocean interactions. Vol. 1, Advances in Fluid Mechanics Series, No. 33, WIT Press, 83-114. Falkovich, A., I. Ginis, and S. Lord, 2005: Ocean data assimilation and initialization procedure for the Coupled GFDL/URI Hurricane Prediction System. J. Atmos. Oceanic Technol., 22, 1918-1932. Moon, I.-J., I. Ginis, T. Hara, and B. Thomas, 2007: A physics-based parameterization of air-sea momentum flux at high wind speeds and its impact on hurricane intensity predictions. Mon. Wea. Rev., 135, 2869-2878. Yablonsky, R. M., and I. Ginis, 2008: Improving the ocean initialization of coupled hurricane-ocean models using feature-based data assimilation. Mon. Wea. Rev., 136, 2592-2607. Yablonsky, R. M., and I. Ginis, 2009: Limitation of one-dimensional ocean models for coupled hurricane-ocean model forecasts. Mon. Wea. Rev., 137, 4410-4419. Fan, Y., I. Ginis, and T. Hara, 2010: Momentum flux budget across air-sea interface under uniform and tropical cyclones winds. J. Phys. Oceanogr., 40, 2221-2242. Fan, Y., I. Ginis, and T. Hara, 2009: The effect of wind-wave-current interaction on air-sea momentum fluxes and ocean response in tropical cyclones. J. Phys. Oceanogr., 39, 1019-1034. 42
References (cont d) Yablonsky, R. M., I. Ginis, B. Thomas, V. Tallapragada, D. Sheinin, and L. Bernardet, 2015: Description and analysis of the ocean component of NOAA's operational Hurricane Weather Research and Forecasting (HWRF) Model. J. Atmos. Oceanic Technol., 32, 144 163. Yablonsky, R. M., I. Ginis, B. Thomas, 2015: Ocean modeling with flexible initialization for improved coupled tropical cyclone-ocean prediction, Environmental Modelling & Software, 67, 26-30. Reichl, B. G., T. Hara, and I. Ginis, 2014: Sea state dependence of the wind stress over the ocean under hurricane winds. J. Geophys. Res. Oceans, 119, 30-51. Reichl, B. G, D. Wang, T. Hara, I. Ginis,, T. Kukulka, 2016a: Langmuir turbulence parameterization in tropical cyclone conditions, J. Phys. Oceanogr, 46, 863-886. Reichl, B. G., I. Ginis, T. Hara, B. Thomas, T. Kukulka and D. Wang 2016b: Impact of sea-state-dependent Langmuir turbulence on the ocean response to a tropical cyclone. Mon. Wea. Rev., 144, 4569-4590. Blair, A., Ginis, I., Hara, T., & Ulhorn, E., 2017: Impact of Langmuir turbulence on upper ocean response to Hurricane Edouard: Model and observations. J. Geophys. Res. Oceans, 122, 9712 9724. 43