1. Air-sea flux distributions from satellite and models across the global oceans Carol Anne Clayson Woods Hole Oceanographic Institution Earth Observation for Ocean-Atmosphere Interactions Science 2014 Frascati, Italy 29 October 2014

2. Overall goal of research Air-sea interaction through surface fluxes of heat and moisture, combined with other weather properties, across variety of spatial and temporal scales Seeking to understand: The variability and extremes in air-sea fluxes of heat and moisture in context of water and energy cycle How distribution of fluxes varies of time, location, differing weather and climate states Using satellite data sets (ISCCP, SRB, SeaFlux, GSSTF, TMPA, HOAPS, GPROF) and MERRA

3. Current Satellite/Blended Datasets Goddard product: GSSTF3 Daily, 0.25°, input variables and turbulent fluxes; satellite plus Ta from reanalysis 1988- 2008; global oceans No current plans to update again – funding issues IFREMER version 3 Daily, 0.25°, input variables, turbulent fluxes; satellites plus Ta from reanalysis Currently available: 1992 (1999 with QuikSCAT) – November 2009; global oceans Japanese Ocean Flux datasets: J-OFURO2v2 – Input variables, fluxes, radiation, satellites plus Ta from reanalysis Daily, 1°, 1988 – 2005; global oceans Satellites, JMA model analyses HOAPS3.2 6-hourly, 0.5°, global oceans; input variables, precipitation; satellites July 1987 - December 2008 OAFlux Daily, 1°, global oceans; blended using reanalysis, in situ, satellites July 1985 – current (monthly available back to 1958)

4. Near-surface air temperature and humidity Roberts et al. (2010) neural net technique SSM/I only from CSU brightness temperatures (thus only covers 1997 - 2006) Gap-filling methodology -- use of MERRA variability – 3 hour Winds Uses CCMP winds (cross-calibrated SSM/I, AMSR-E, TMI, QuikSCAT, SeaWinds) Gap-filling methodology -- use of MERRA variability – 3 hour SST Pre-dawn based on Reynolds OISST Diurnal curve from new parameterization Needs peak solar, precip Uses neural net version of COARE Available at http://seaflux.orghttp://seaflux.org 1999 Latent Heat Flux 1999 Sensible Heat Flux SeaFlux Data Set Version 1.0

5. Average effect of diurnal SST on fluxes Clayson and Bogdanoff (2013)

6. Example distributions Latent heat flux 1999

7. 95 th Percentile

8. Extremes in LHF

9. Extremes in winds

10. Extremes in Qs-Qa

11. Trends...

12. Latent Heat Flux 95 th Percentile Values 1998 2007 Difference between 2007 and 1998 1998 2007 Difference between 2007 and 1998 Latent Heat Flux 95 th Percentile Values Changes in extremes in latent heat flux

13. Weather Regimes Example Use of ISCCP cluster weather states (Jakob and Tselioudis 2003) Tropical convection and MJO (Tromeur and Rossow, 2010; Chen and Del Genio, 2009) Datasets: ISCCP Extratropical Cloud Clusters (35N/S, 2.5°x2.5° 1985-2007, 3-hr) SEAFLUX (1998-2007,0.25°x0.25° 3-hr), LHF/SHF/Surface Variables Product Homogenization: Fluxes regridded and resampled to ISCCP 2.5x2.5 ISCCP 3-hr used to assign a daily class based on the most frequent cluster More convection Less convection

14. Decomposition of surface fluxes by weather state Weather regimes result in distributions of fluxes with different mean and extreme characteristics Associated with changes in means Both wind speed and near-surface humidity gradients are particularly well stratified, though the latent heat flux means are less so Indicates potential compensations LHF SHF Qs-Qa Ts-Ta Wind Speed

15. Tropics

17. Mid-Latitude North

19. Compositing methodology Conditionally sample data using weather state classification (WS1-WS8; most convective to least convective) Further sampled based on compositing index to evaluate low-frequency coupled variability Use NOAA Climate Prediction Center (CPC) indices for ENSO and MJO  Examining differences in means can be decomposed as changes in class mean (A), changes in RFO (B), and covariant changes (C) ABC

20. MJO Composites – Decomposition into Weather states Decompose LHF into weather state means and relative frequency of occurrence (RFO) Systematic variations of both weather state means and RFO with MJO index Both variations contribute to total impact of a given weather state on mean energy exchange associated with MJO evolution

21. Example: Climate regimes Composite MJO based on index strength not time-lagging All three regions typically show increased evaporation during convective phase and decreased evaporation during suppressed phase The Indo-Pacific region changes  more wind-driven Eastern Pacific changes  more near- surface moisture gradient changes But: EIO more coherent near- surface moisture changes than WP Convective NeutralSuppressed EIO 70E-90E WP 130E-150E EP 130W-110W

22. ENSO Composites by strength West Pacific (130E-150E) latent heating anomalies primarily driven by QSQA anomalies For MJO, near-surface wind speed was also anti-correlated but it was stronger than QSQA The East Pacific (130W-110W) LHF acts to damp the existing SST anomalies Unlike on MJO time scales, wind speed and QSQA are positively correlated EIO WP EP El Nino La Nina

23. MLD and surface flux effects on SST tendencies The mixed layer depth is an important contributor to the observed surface heat flux tendency pattern.

24. EIO and WP: deeper ML in convective; EP: slightly deeper ML in suppressed WP: LHF variability has roughly same effect on SST tendency throughout MJO. EP: LHF much higher effect on variability during convective phase EIO: Even shallower ML in suppressed phase, but still large LHF due to Qs-Qa difference: LHF variability strongest effect during suppressed phase Convective NeutralSuppressed EIO 70E-90E WP 130E-150E EP 130W-110W Deeper ML Shallower ML Deeper ML Shallower ML LHF variability roughly same effect LHF variability very different effect

25. Summary Cloud-based weather states can be used to provide improved understanding of surface energy flux variability Tropics: main contributor to latent heating is the trade cumulus regime: nearly highest mean LHF, most frequent weather state (clear sky has highest mean latent heat flux) Midlatitudes: main contributor to latent heating is the shallow BL cumulus (highest frequency and mean) Fair/foul weather: foul weather tends to have sharper peaks, fair broader distributions MJO variability is particularly well decomposed using ISCCP weather regimes from convective to neutral and suppressed states Different regions in the tropics show MJO and ENSO variability being driven by different processes Even when total LHF differences are equivalent, if winds versus Qs-Qa effects (with resulting MLD differences) occur, changes in LHF different effects on temperature tendency For example: fair weather cirrus vs. marine stratus vs. clear skies during suppressed conditions Both the weather state and the ML state affect resulting impacts on SST

26. Latent Heat Flux Specific Humidity Comparisons with CMIP4 models

27. Sea Surface Temperature Winds

28. Correlation of variability: satellites, CMIP4 LHF and wind LHF and humidity LHF and SST With Natassa Romanou

29. Extremes in LHF

30. Extremes in winds

31. Extremes in Qs-Qa

32. Intercomparing products by weather state While there are systematic mean differences in products, the anomalous changes between products (here, SeaFlux & OAFlux) are more closely aligned. The differences here can be related to specific types of weather regimes OAFlux shows a slight increase in the latent heat flux associated with deep convective conditions while SeaFlux shows a slight decrease. In broken stratocumulus conditions, SeaFlux indicates about a 20% change, nearly 2x that of OAFlux, again primarily from differences in near-surface moisture gradients