Physically-based risk assessment of hurricane storm surge in a changing climate Ning Lin Princeton University Department of Civil and Environmental Engineering Hurricane Ike 5 Year Workshop Rice University, Sep. 23-25, 2013
Hurricane Sandy (2012)
Gradient wind profile Emanuel and Rotunno (2011) Wind Field Modeling V m V m >> fr m R >> o R m R m f 4V m R 2 o Rm Ro
Gradient wind profile Emanuel and Rotunno (2011) Wind Field Modeling V m V m >> fr m R >> o R m R m f 4V m R 2 o Surface (10-m) wind Gradient to surface wind reduction (0.85; Georgiou et al. 1983) Empirical expression of inflow angle (Bretschneider;1972) Background wind component (0.55U t at 20 ; Lin and Chavas 2012) Rm Ro
Gradient wind profile Emanuel and Rotunno (2011) Wind Field Modeling V m V m >> fr m R >> o R m R m f 4V m R 2 o Surface (10-m) wind Gradient to surface wind reduction (0.85; Georgiou et al. 1983) Empirical expression of inflow angle (Bretschneider;1972) Background wind component (0.55U t at 20 ; Lin and Chavas 2012) Sensitivity/uncertainty analysis conducted by Lin and Chavas (2012) Rm Ro
Storm Surge Simulation SLOSH model (Jelesnianski et al. 1992) SLOSH mesh ~ 10 3 m ADCIRC mesh ~ 10 2 m Battery ADCIRC model (Luettich et al. 1992) ADCIRC mesh ~ 10 m (Colle et al. 2008)
Hindcast (2012) 0.7 m 2.16of m storm surge for Hurricane Sandy The Battery 2.91 m (2.74) Bergen Point 3.68 m (3.8) 3.25 m King s Point 3.2 m (2.9) Sandy Hook 3.04 m (2.71) 3.33 m Atlantic City 2.4 m (1.8) Cape May 1.21 m (1.05) Montauk 1.22 m (1.38)
Hindcast of storm surge for Hurricane Irene (2011) 1.3 m
3.2 m 3.5 m Sandy Donna Gloria 2.1 m Irene (estimated) 2011 2012 Scileppi and Donnelly (2007)
Historical NY-region Storms Annual frequency λ=0.34 55 historical storms that pass within 200 km of the Battery with V m > 20 m/s, during 1851-2011 (Hurricane Best Track Data)
Physically-based Hurricane Surge Risk Assessment Environmental conditions: Reanalysis or GCMs Wind & pressure models; Numerical grids Emanuel et al. statistical/deterministic hurricane model Hydrodynamic models: ADCIRC and SLOSH Extreme values modeling Storm Generation Surge Simulation Statistical Analysis Astronomical tide Sea level rise Extremes Return levels with confidence bonds
Reanalysis-data-driven NY-region storm simulation 5000 synthetic tracks that pass within 200 km of the Battery with V m > 20 m/s, simulated by a statistical/deterministic hurricane model (Emanuel et al. 2006 and 2008) driven by the 1981-2000 climate from NCEP/NCAR reanalysis
Some worst-case surges (tides not included) under NCEP/NCAR 1981-2000 climate Prob. = 1/14500 Prob. = 1/10000 Lin et al. 2010 Sep. - JGR Lin et al. 2012 Feb. - NCC
Density NCEP 1981-2000 climate Surge at the Battery (m)
Density NCEP 1981-2000 climate Surge at the Battery (m)
Fitting tail with Generalized Pareto Distribution (GPD) P{ H > h u h} = ς u[1 + ξ( )] σ 1 ξ Density NCEP 1981-2000 climate Exceedance probability P { H > h} Surge at the Battery (m) Surge at the Battery (m)
} { max 1 } { h H P e h H P > = > λ T h H P 1 } { max = > Surge return level Surge return level ξ σ ξ ς 1 )] ( [1 } { + = > u h h H P u For GPD: + = 1 ] ) 1 log(1 [ ξ λς ξ σ T u h u Parameter estimation: MLE Confidence interval estimation: Delta method NCEP 1981-2000 climate
Spatial variation of long-term surge risk 100-year 500-year 10000-year
GCM-driven NY-region storm simulation GFDL 1981-2000 GFDL 2081-2100 (IPCC-AR4 A1B) Model Designation Institute CNRM-CM3 CNRM Centre National de Recherches Météorologiques, Météo-France ECHAM5 ECHAM Max Planck Institute GFDL-CM2.0 GFDL NOAA Geophysical Fluid Dynamics Laboratory MIROC3.2 MIROC CCSR/NIES/FRCGC, Japan
GCM storm surge return level Green: Current climate (1981-2000) Orange: A1B future climate (2081-2100)
GCM storm surge return level Effect of the Change of Storm Size Distribution due to Climate Change Green: Current climate (1981-2000) Orange: A1B future climate (2081-2100) Purple: A1B future climate (2081-2100) with R 0 increased by 10% and R m increased by 21%
Related processes: Wave setup (210 surge events (>70-year events) for the NCEP 1981-2000 climate) Effect of wave setup is small and thus neglected in risk assessment.
Related processes: Astronomical tide Simulated surge-tide nonlinearity (L), for each of the 210 extreme events, every 3 hours during a day Curve: Simulated astronomical tide Dots: L= storm tide - surge - tide Combination of surge and tide (Each color represents a set of 210 events; 8 sets are shown) L is relatively large (9% of the surge). When the surge happens at the high tide or during the rising tide, L<0; when the surge happens at the low tide or during the receding tide, L>0.
Surge-tide nonlinearity (γ) Tidal effect γ (ϕ) Surge = L H + + tl t r Mean tidal level Tidal range Black curve: simulated tide Pink curve: observed tide Dots: average values of γ Blue curve: regression of γ
Surge-tide nonlinearity (γ) Tidal effect γ (ϕ) Surge = L H + + tl t Black curve: simulated tide Pink curve: observed tide Dots: average values of γ Blue curve: regression of γ r Mean tidal level Tidal range t P{ H < h} = P{ H + t + L < h} = P{ H + t + γ ( H + t ) t h} = Storm tide 2π P{ H Storm tide level 0 + r l < h t( ϕ) γ ( ϕ) t < 1 γ ( ϕ) r + tl 1 } dϕ 2π NCEP climate
Related processes: Sea level rise (SLR) (210 surge events (>70-year events) for the NCEP 1981-2000 climate) Nonlinear effect of SLR is small and thus neglected.
Related processes: Sea level rise (SLR) } { } { h s L t H P h H P f < + + + = < ϕ π ϕ γ ϕ γ ϕ π d s t t t h H P l r 2 1 } ) ( 1 ) ( ) ( { 2 0 + + < = Flood height (storm tide + SLR) Nonlinear effect of SLR is small and thus neglected. (210 surge events (>70-year events) for the NCEP 1981-2000 climate)
GCM flood height return level (assuming SLR of 1 m for the future climate ) Black: Current climate (1981-2000) Blue: A1B future climate (2081-2100) Red: A1B future climate (2081-2100) with R 0 increased by 10% and R m increased by 21% Lin et al. (2012)
Summary: We have developed a technique for downscaling global models or reanalysis data sets, using high resolution, atmospheric-ocean coupled TC and hydrodynamic models to assess surge risk in a changing climate. NYC is highly vulnerable to surge flooding. We show that the surge level for NYC will likely increase due to the change of storm climatology by a magnitude comparable to the projected SLR. The combined effects of storm climatology change and SLR may greatly shorten the surge flooding return periods in the future. The technique has been applied to other areas. Analyses show potentially large vulnerability in some places (like Dubai) where TCs have never been recorded, and larger- than-expected storm and surge risk in some other places (such as Carins and Tampa).