Evaluation of AOMIP Modeled Arctic Sea Ice Thickness (Using Observational ULS data) Mark Johnson 1, Andrey Proshutinsky 2, Yevgeny Aksenov 4, Igor Ashik 3, Beverly de Cuevas 4, Nikolay Diansky 5, Christian Haas 6, Sirpa Hakkinen 7, Ron Kwok 8, Ron Lindsay 9,Wieslaw Maslowski 10, An T. Nguyen 8, Jinlun Zhang 9 1 Institute of Marine Sciences, University of Alaska Fairbanks, Fairbanks, AK, USA 2 Wood Hole Oceanographic Institution, Woods Hole, MA, US 3 Arcticand Antarctic Research Institute, St. Petersburg, Russia 4 National Oceanography Centre, Southampton, Southampton, UK 5 Institute of Numerical Mathematics Russian Academy of Sciences, Moscow, Russia 6 University of Alberta, Edmonton, Canada 7 Goddard Space Flight Center, Greenbelt, MD, USA 8 Jet Propulsion Laboratory, Pasadena, CA, USA 9 Polar Science Center University of Washington, Seattle, WA, USA 10 Naval Postgraduate School, Monterey, CA, USA

Ice Thickness from models and ULS –Linear regression –Histogram –Differences (models-observations) –Correlations –Taylor Diagram (modified) –Model issues Ice Concentration – seasonality –methodology and validations

AWI IOS ULS Locations BGEP NPEO

( NPEO observed 3.87m removed) Comparison using monthly means from models and ULS obs

Model is too thin Model is too thick

Model – Observations Thickness

n=1 Correlations

30 cm-30 cm models>obsModels<obs

30 cm-30 cm These model-obs have differences > |30 cm|

We will look at the model vs observation correlations and the model minus observation values using a modified Taylor Diagram as follows

model – obs = 75 m model – obs = -30 m model – obs = 2m model – obs = - 2m model = obs model – obs = -75 m model – obs = 30 m Taylor Diagram (modified) rotation scaled to 2m model-obs thickness model thickness > obs model thickness < obs

model – obs = 75 m model – obs = -30 m model – obs = 2m model – obs = - 2m correlation=0.6 correlation=1 model = obs model – obs = -75 m model – obs = 30 m Taylor Diagram (modified)

model – obs = 75 m model – obs = -30 m model = obs model – obs = -75 m model – obs = 30 m Taylor Diagram (modified) model – obs = 2m model – obs = - 2m “good”

model – obs = 75 m model – obs = -30 m model = obs model – obs = -75 m model – obs = 30 m Taylor Diagram (modified) model – obs = 2m model – obs = - 2m “not so good”

Performance by model

n=1 “good” AWI3 “not so good” AWI1 AWI4 AWI11

“good” IOS4 IOS1, 2, 3, 4, 5, 6, 7, and 8 are underestimated BGEP A, B, C and D are overestimated

“good” IOS8 AWI2 AWI3 “not so good” AWI2 AWI7 BGEP A,B,C

“good” IOS8 AWI1 BGEPB BGEPC

“good” IOS2 IOS3 IOS8 AWI3 AWI5 AWI6

“good” IOS1 IOS2 IOS3 IOS8 AWI2 AWI3 AWI4 BGEPA BGEPB BGEPC

Performance by instrument

“good” UW - 4 ECCO2 – 3 NPS – 1 GSFC - 1 INMOM - 1

n=1 “good” UW - 3 ECCO2 – 3 NPS – 2 INMOM - 2 “not so good” INMOM - 2 ORCA - 2 n=1

“good” UW – 3 NPS – 2 “not so good” INMOM - 3

UWECCO2NPSGSFCINMOMORCA IOS 1 IOS 2 IOS 3 IOS 4 IOS 5 IOS 6 IOS 7 IOS 8 AWI 1 AWI 2 AWI 3 AWI 4 AWI 5 AWI 6 AWI 7 AWI 8 AWI 9 AWI 10 AWI 11 BGEP A BGEP B BGEP C BGEP D good: poor:53 n=1 correlation = -0.76

Where do the “good” data show up on the model vs observations?

( NPEO observed 3.87m removed)

Conclusions from ULS data Generally the models overestimate the ice thickness compared to ULS observations Models don’t have enough sea ice in the thin or FY ice range up to 1m thick Models have too much sea ice greater than 2m thick Models do better in the Beaufort than Fram Strait Questions –Why do the models have too much MY ice and not enough “new” ice? –Do the models not melt or advect away the thicker MY ice? –Do tides “open up” the sea ice allowing for more of the “thin” ice less than 1m thick? –What models have tides in them?

Seasonality of Sea-Ice Concentration Subsistence hunters along coastal Alaska have observed for years the start and end dates of “freeze-up” and “break-up” Algorithm to compute these event times using SSM/I has been developed with good agreement (Hajo Eicken) Seasonality can be computed using satellite record. Compare with model results?

Conclusions UW does well integrating thickness Most models agree with IOS data but not with AWI data TPD and Fram Strait export may play a key role Seasonality of sea-ice concentration appears to be an attractive validation tool.

CSFCECCO2INMOMNOCSNPSUW Domain b Resolution c Ice t regional 0.35 ⁰ ⁰ 720 s regional 15-22km 600 s regional 0.25 ⁰ 3600 s global 3-6 km 7200 s regional 9 km 3600s regional 6-75 km 1152 s Vertical coordinateσzzz Vertical levels Minimum depth25m5m6.065m Bering StraitRestoredNot restored Fully represented in global domain open Equation of stateMellor Jackett and McDougal, 1995 Jackett & McDougall (1995) UNESCO Vertical mixingMY2.5KPP, no double diffusion TKE (Gaspar et al.( 1990), Blanke & Delecluse (1993)) KPP Tracer advection Lin et al 1994 Piecewise parabolic 7 th order monotonicity- preserving (Direct space time with flux limiter) [Daru and Tenaud, 2004] TVD (Lévy et al. 2001) Central diff. Momentum advectioncenteredvector invariantEEN (Barnier et al. 2006) Central diff.

GSFCECCO2INMOMNOCSNPSUW Ice Physical parameterizations Salinity5Function of surface S64 Thickness categories d 2: ice and no ice 8 (7 for ice and 1 for open water) 112 Advection Centered mom. Upwind A+D Centered 2 nd order Prather, 2 nd order, 2 nd moment conserving Central diff. Dynamics e Generalized viscous Viscous plasticVP Teardrop plastic rheology, LSR solver Albedos

CSFCECCO2INMOMNOCSNPSUW Albedos Melting snow Cold snow – melting snow (clear sky, snow thickness dependent) 0.70 Cold ice (clear sky, ice thickness dependent) 0.75 Melting ice (clear sky, ice thickness dependent) 0.64 Ocean Surface Momentum Exchange Coefficients Atmos.-ice g 1.4E x 10^ x Surface BL Ice-OceanBL model5.4 x 10^-35.0 x Cw=0.0055

StartEndLatitude (degrees, minutes) Longitude (degrees, minutes) MooringInstrumentWater Depth Instrument Depth Data directory Aug-91Nov 'N12 40'WAWI411APL261002m48muls Aug-92Dec 'N11 43'WAWI412-2APL312362m50muls31-92 Aug-93Jul 'N07 38'WAWI414-2APL323425m70muls Jul-94Oct 'N12 59'WAWI410-2APL49413m73muls Aug-97Sep-9879N02WV10-1APL322600m58muls Sep-98Sep-9979N02 03'WV10-2APL472609m54muls Sep-99Aug-0079N02 03'WV10-3APL252582m53muls Oct-99Sep 'N10 15'WAWI419-1APL323229m63muls Aug-00Oct 'N02 03'WF10-4APL482554m67muls Sep-00Sep 'N10 12'WAWI419-2APL313160m65muls Sep-01Sep 'N10 12'WAWI419-3APL473160m82muls

Preliminary conclusions The major one is that it seems that the UW model is good and a bit better than others. ECCO2 models results (this is MIT model which An T Nguen runs at JPL are also good and better than the others except UW model. I will continue working with conclusions.