1. Implementation of the Updated Goddard Longwave and Shortwave Radiation Packages in WRF Toshi Matsui, Wei-Kuo Tao, and Roger Shi representing Goddard Mesoscale Dynamics and Modeling (GMDM) Group Toshi Matsui, Wei-Kuo Tao, and Roger Shi representing Goddard Mesoscale Dynamics and Modeling (GMDM) Group Goddard Space Flight Center Mesoscale Dynamics and Modeling Group

2. Goddard Microphysics Goddard Land Information System Goddard Chemistry and Aerosol Radiation Transport (GOCART) Goddard Radiation Surface emission and albedo Aerosol direct effect Cloud radiative forcing Aerosol indirect effect Land-atmosphere interactions Land-aerosol interactions WRF Dynamic Core Diabatic Heating (dT/dt) Role of Goddard Radiation in WRF Roger Shi

3. Goddard radiation packages Goddard radiation package (original name CLIRAD) has been developed for two decades at NASA Goddard by Ming-Dah Chou and Max J. Suarez for use in general circulation models (GEOS GCM), regional model (MM5) and cloud-resolving models (GCE). References Chou M.-D., and M. J. Suarez, 1999: A solar radiation parameterization for atmospheric studies. NASA Tech. Rep. NASA/TM-1999-10460, vol. 15, 38 pp Chou M.-D., and M. J. Suarez, 2001: A thermal infrared radiation parameterization for atmospheric studies. NASA/TM-2001-104606, vol. 19, 55pp WavelengthSW (Solar)LW (thermal) Flux solutionTwo-stream adding methodSchwarzchild equation # of bandsUV&PAR(8 bands) Solar-IR(3 bands) 10 bands Optical approximationDelta-Eddington approximation (for scattering and transmission) Henyen-Greenstein function (for scattering), One/two-parameter scaling, modified k-distribution (for absorption) Optical parametersH 2 O, O 2, O 3, CO 2, condensates (cloud water, cloud ice, snow, rain, and graupel), aerosols (sulfate and precursors, dust, black carbon, organic carbon, sea salt) H 2 O, O 3, CO 2, trace gases (N 2 O, CH 4, CFC11, CFC12, CFC22), condensates (cloud water, cloud ice, snow, rain, and graupel), aerosols (sulfate and precursors, dust, black carbon, organic carbon, sea salt), AccuracyHeating rate error within 5% accuracy in comparison with a LBL model. Cooling rate error within 0.4K/day in comparison with a LBL model.

4. New (2006) Goddard SW radiation in comparison with old (1997) Goddard SW radiation in WRF 1) Optical depths for condensates (1st-order effect) NewRad has a strict threshold of cloud optical depth (=0.0001) for cloud flags in order to account for thin-cloud radiative effects. (OldRad has a loose threshold (=0.05), and does not account for thin-cloud radiative forcing.) NewRad accounts for optical properties of rain, ice+snow, and liquid cloud droplets. Ice cloud effective radius (25~125micron) depends on ambient temperature. (OldRad accounts for only ice and liquid cloud droplets. Ice cloud effective radius is fixed value (=80micron). ) 2) Radiative Transfer (2nd-order effect) NewRad has a correct two-stream adding approximation in diffuse transmissivity. (OldRad uses incorrect diffuse transmissitivty. This is a critical bug in the code. ). NewRad uses delta-Eddington approximation for reflection and transmittance of direct and diffuse radiation. (OldRad also uses delta-Eddington approximatatin for direct radiation, but it uses equations in Sagan and Pollock [JGR, 1967] for diffuse radiation.) 3) Molecular absorption (3rd-order effect) NewRad weights the molecular absorption coefficients by cosine of solar zenith angle. (OldRad uses the same molecular absorption coefficients without considering cosine of solar zenith angle.) NewRad accounts for water vapor absorption. (OldRad does not account for water vapor absorption.) NewRad accounts for O 2 and CO 2 absorption above and below cloud-top level. Below cloud-top level, the flux reduction rate depends on the ratio of clear-sky and cloudy-sky net radiation. (OldRad accounts for O 2 and CO 2 absorption only above cloud-top level.) NewRad setup CO 2 concentration for Y2000 (=336ppm). (OldRad set up 300ppm.)

5. Average CPU time in OldRad, NewRad and NewRadFast Test bed Old (OldRad) and new (NewRad) Goddard SW radiation were tested in all-sky conditions of 45x45x30- grid (x-y-z coordinations) domains to compare CPU time. Overcast option A logical option (overcast) deals with either overcast cloud (cloud fraction=0 or 1; true) or broken cloud (cloud fraction 0~1; false). overcast=false requires two-stream solutions in different combination of clear-cloud sky cases, which result in much longer computational time. overcast OldRad NewRad NewRadFast.true. 1.6sec 4.0sec -> 1.5sec 1.0sec.false. 4.5sec 5.1sec -> 2.0 sec 2.0sec (Toshi’s optimization) i) 1-dimensionalized radiative transfer to skip nighttime computation, ii) removed redundant routines, and iii) added a new logical option (fast_overcast), and “fast_overcast =.true.” that uses a pre-computed look-up table for F cloud /F clear as a function of cloud albedo. Fast, Faster, Fastest!

6. Comparison in SW flux and heating rate between NewRad, OldRad, and NewRadFast Results NewRad-OldRad differences in surface downwelling shortwave radiation are up to  40W/m 2 due mostly to upgrade (1). It has large discrepancy in heating profile up to  2K/day. NewRad-NewRadFast differences in surface downwelling shortwave radiation are up to  1W/m 2. It has discrepancy in heating profile up to  0.5K/day. Y-distance=120km

7. Options in Goddard LW radiation code and CPU time Goddard LW logical options “high=.true.”computes transmission functions in the CO 2, O 3, and the three water vapor bands with strong absorption using look-up table, while “high=.false.” use k-distribution methods for this (faster). “trace =.true.” accounts for absorption due to N 2 O, CH 4, CFCs, and the two minor CO 2 bands in the LW window region, while “trace =.fase.” does not account (faster). A combination of “high=.true.” & “trace=.true.” is most accurate. Exp NamehightraceCPU time (sec) RHT.true. 2.80 (170%) RH.true..false.2.13 (129%) RT.false..true.2.19 (133%) R.false. 1.65 (100%) Results Each option (high or trace) costs about 0.6sec, thus the no-options experiment (R) save 1.2sec CPU time in comparison with the full-option experiment (RHT).

8. Comparison in LW flux and heating rate between RHT, RH, and RT. Results Downwelling longwave radiation between RHT and RT are similar. LW cooling profiles between RHT and RH are largely different near the cloud top. LW cooling profiles between RHT and RT are slightly different at TOA. If one include the stratosphere in model, the RHT-RT difference would become larger [Chou and Suarez 2001]. Y-distance=120km

9. Aerosol Direct Effect Goddard radiation module is liked to the GOCART module, which compute 11-SW-band and 10-LW-band aerosol optical properties for a given aerosol masses (sulfate, sea salt, dust, OC, and BC). The module deals with IO process in GOCART aerosol masses for initial/boundary conditions. module_ra_goddard.F module_gocart_offline.F module_radiation_driver.F GOCART global aerosol masses module_gocart_online.F

10. Aerosol Direct Effect Off-line GOCART-induced diurnally averaged SW and LW aerosol radiative forcing in d01.

11. Satellite Data Simulation Unit (SDSU) for evaluating WRF and supporting NASA’s mission Driver & Interface Module Visible-IR Optical properties Microwave Optical properties WRF output Active Sensors CALIPSO ICESAT, and ground- based LIDAR Passive Sensors AVHRR,TRMM VIRS, MODIS, GOES, Multifilter Rotating Shadowband Radiometer (MFRSR), and etc. Active Sensors TRMM PR, GPM DPR, CloudSat CPR, Millimeter Wavelength Cloud Radar (MMCR), W-Band (95 GHz) ARM Cloud Radar (WACR), and etc. Passive Sensors SSM/I, TRMM Microwave Imager, AMSR-E, AMSU, and MHS, ground-based Microwave Radiometer (MWR), Microwave Radiometer Profiler (MWRP), and etc.

12. SDSU-WRF supports NASA Global Precipitation Mission (GPM) Radar reflectivity and microwave brightness temperature are computed in off-line using the WRF- simulated meteorology and hydrometeors field via satellite-data simulation unit (SDSU). Simulated Tb and reflectivity are used to support algorithm development for future NASA Global Precipitation Mission (GPM) satellites.