Transcript Slide 1
An evaluation of a hybrid satellite and NWPbased turbulent fluxes with TAO buoys ChuanLi Jiang, Kathryn A. Kelly, and LuAnne Thompson University of Washington Meghan Cronin NOAA/PMEL Outline 1. Motivation 2. Scatterometer “raises the bar” 3. TAO buoy comparisons 4. Heat flux map comparisons 5. Applications Motivation Intra-seasonal heat budget important in ENSO and climate change (McPhaden, 2002; Kessler et al. 1995; Zhang, 2001) Need accurate air-sea fluxes to force an ocean model NWP winds and heat flux products have systematic errors Satellite measurements provide accurate inputs for both momentum and turbulent heat fluxes Can QuikSCAT winds and microwave SST improve turbulent heat flux products? Jiang, Cronin, Kelly and Thompson, under revision for JGR Scatterometer “Raises the Bar” on Vector Wind Measurements Scatterometers revealed systematic 7o direction error in TAO buoys Difference between scatterometer winds and anemometer winds is ocean currents Scatterometer comparisons show importance of using a scalar average for wind speed Scatterometer winds wind vector relative to ocean surface U air U sea SHF ac pCH Ua Us Ta Ts TAO - QuikSCAT winds = currents (ADCP) Kelly, Dickinson, McPhaden, and Johnson, GRL, 2001 Scalar Averaging for Wind Speed • For LHF and SHF QuikSCAT winds converted to speed and then scalar averaged • TAO 10-minute winds vector averaged to obtain “daily” winds (for ARGOS transmission) • Comparisons with TAO10-minute winds show 4-day scalar average of QuikSCAT is more accurate than 4-day average of “daily” TAO wind SCALAR average winds for fluxes (i.e., compute wind speed from observations and then average or map) Passive Microwave SST (TMI) • microwave can see through clouds • 25km resolution • 40S-40N • MW/OI from Remote Sensing Systems Method Bulk algorithm: LHF a L CE U a U s qa qs SHF a c p C H Ua U s Ta Ts State variables used in COARE v3.0 algorithm Most accurate state variables determined by comparison with TAO buoys Turbulent heat flux products compared with TAO variables in COARE v3.0 algorithm State variable evaluation Relative wind speed U r Sea surface temperature SST Air specific humidity qair Air temperature T air State Variable Sources • 2 years: 2000 – 2001 • Average all variables to 4-day resolution (QuikSCAT mapping) • Scalar-average relative wind speed: Ur Uair Usea ocean current from altimeter (Kelly et al. 2004) • TAO Bulk SST (skin SST not available) “Truth” State Variable Sources State Variables QuikSCAT Ur 4-day 1 degree NCEP1 NCEP2 ERA40 TAO buoy 6 hourly Gaussian 6 hourly Gaussian 6 hourly 2.5 degree hourly daily daily daily daily q air daily daily daily daily Tair daily daily daily daily SST MW/OI 3-day .25 degree TAO buoys used in comparisons + 64 buoys for SST q air Tair 38 buoys for U r LHF Wind speed Bias = Product -TAO NCEP1 too weak NCEP2 better than NCEP1 ERA40 better than NCEP QuikSCAT best, Higher along 165E & 8N Histograms of wind speed in eastern Pacific best match NWP lacks high winds Histograms of wind speed near ITCZ lacks low wind speed weak zonal currents or rain contamination best match SST comparison Bias = product - TAO SDD = STD(product - TAO) NWP SST: warm in the cold tongue; cold off the equator MW/OI: consistently cold (but may be correct) Air specific humidity ERA40 has best humidity Dry along 165E NCEP2 worse than NCEP1 Better along 8N NCEP too dry in the east too wet in the west State variable evaluation summary SST (o C ) Tair ( o C ) Sources q air ( g / kg ) U r (m / s) Bias SDD Bias SDD Bias SDD Bias SDD NCEP1 -.1 .3 -.2 .5 -.0 .8 -1.3 .9 NCEP2 -.1 .3 .1 .5 -.4 .8 - .4 1.0 ERA40 .0 .3 -.1 .3 -.0 .5 - .5 .6 MW/OI -.1 .3 QuikSCAT Hybrid .0 MW/OI ERA40 ERA40 .5 QuikSCAT Sensitivity of LHF to state variable LHF(var(i) + all other TAO) - LHF(all TAO variables) Summary of sensitivity of LHF to state variables 3 Products NCEP1 NCEP2 ERA40 MW/OI 2 1 SST (o C ) Tair ( o C ) Bias SDD Bias 2.3 1.6 -0.5 3.7 11.6 -0.8 11.6 0.2 10.4 -0.4 8.5 q air ( g / kg) SDD Bias SDD 1.4 1.4 1.0 -6.3 -3.1 -4.8 18.3 15.0 13.0 18.4 2.5 13.8 11.0 4.4 8.3 QuikSCAT LHF errors: 1) humidity 2) wind 3) SST All errors in W/m2 U r (m / s ) Bias -4.0 SDD 6.9 Latent Heat Flux Products for evaluation against TAO/COARE Products SST (o C ) Tair ( o C ) q air ( g / kg ) U r (m / s) algorithm NCEP1C NCEP1 COARE NCEP2C NCEP2 COARE ERA40C ERA40 COARE Hybrid MW/OI ERA40 ERA40 QuikSCAT COARE LHF comparison NCEP1C underestimates NCEP2C overestimates ERA40C good Hybrid best in the east overestimates along 165E,8N LHF bias along 165E LHF bias from qair from Ur from SST ERA40: low humidity compensates for weak winds smaller bias Hybrid: low humidity + stronger winds too strong How do NWP products compare with using their state variables in the COARE algorithm? Summary of LHF comparison Using COARE LHF NWP products LHF NCEP1C NCEP2C ERA40C Hybrid Bias SDD Bias SDD Bias SDD Bias SDD 10.7 24.8 -8.9 26.6 -1.5 18.1 -5.8 16.2 NCEP1 NCEP2 ERA40 Bias SDD Bias SDD Bias SDD -4.7 26.8 -28.6 32.0 -13.6 18.1 Algorithm tuned to weak winds Same algorithm as NCEP1 COARE decreases bias Difference: Algorithm + State variables + Temporal resolution of input variables Map comparisons in the tropical Pacific Wind speed map comparison NWP winds are weaker than QuikSCAT SST map comparison NWP SST warmer in the cold tongue colder off the equator LHF map comparison Hybrid LHF 25 50Watt / m2 Larger than NWP/COARE Hybrid LHF is similar to NCEP1 off the equator Application of Hybrid product to intra-seasonal heat budget in ocean circulation model GOAL Role of downwelling Kelvin wave in ENSO variability. Method MODEL HIM Turbulent Heat flux Hybrid product Momentum flux QuikSCAT Solar radiation Corrected ISCCP Summary • QuikSCAT accuracy improves turbulent heat fluxes (scalar average) • LHF sensitive to specific humidity, wind speed, and SST • Differences in products from both state variables and bulk algorithm (NCEP1 vs. NCEP2) • Improvement in LHF from wind speed offset by error in air specific humidity • Problem areas for hybrid fluxes: ITCZ and warm pool