RTTOV-91 Science and Validation Plan Annex-A: RTTOV9 Cloud validation Author O Embury C J Merchant The Univerity of Edinburgh Intitute for Atmo. & Environ. Science Crew Building King Building Edinburgh EH9 3JN Draft Verion: 20 March 2007
Introduction Thi work wa undertaken a an extenion to an NWP SAF Aociated Scientit Miion initiated in November 2005, and hould be read in conjunction with the reulting report RTTOVCLD verion 8: Independent Aement (Merchant, Embury and Old, 2006). The objective i to validate new cloudy radiance code, RTTOV9, by cro-comparion of RTTOV8 cloudy radiance imulation againt and independent cattering model. DISORT A with the previou independent aement of RT8C, the DISORT computer code i ued to calculate control radiance. Due to the enhancement introduced in RTTOV9, incorporating DISORT a the radiative tranfer olver wa more traightforward than in the previou tudy. RTTOV9 now calculate all the parameter neceary for cattering calculation o DISORT can effectively be ued a an alternative to the rttov_integrate ubroutine. Modification to the different part of RTTOV9 to facilitate the comparion are lited in further detail below. Scattering parameter and code modification RTTOV9 make ue of four cattering parameter: aborption optical depth, cattering optical depth, backcatter parameter, and phae function. RTTOV9 ue the phae function only for olar radiance calculation, o the phae function i only tored for near infrared channel. DISORT ue the full phae function for all calculation; therefore new cattering coefficient file were neceary, including the phae function for all channel. Furthermore DISORT require that the phae function i expreed a Legendre polynomial coefficient rather than tabulated a a function of cattering angle. The optical propertie of aerool and water cloud ued in RTTOV9 are taken from the Optical Propertie of Aerool and Cloud (OPAC) package. Thu, for conitency, we extracted phae function data for the thermal channel from the OPAC package. The cattering coefficient file were alo altered to contain the aymmetry parameter, rather than the backcatter parameter. The aymmetry parameter i not actually ued in the current verion of the DISORT code, but wa ued for early tet which aumed a Henyey-Greentein phae function. rttov_opdpcattir The ubroutine rttov_opdpcattir wa modified to extract the phae function for all channel rather than jut the olar-affected channel. Code previouly ued to calculate the azimuthally averaged phae function wa removed a it wa no longer required and relied on the phae function being tabulated a a function of cattering angle. Finally the phae function moment were aved in the tranmiion_catt_ir tructure for paing to the rttov_integrate ubroutine. rttov_integrate Thi ubroutine wa replaced with a wrapper around the DISORT computer code. A with the RT8C comparion it i neceary to 'correct' the RTTOV9 calculated optical depth to equivalent nadir optical depth. With RTTOV9, however, there i no need to interpolate optical depth / radiance to model preure level a all calculation are performed on the ame fixed preure level. The optical propertie of each layer are calculated from the ga, aerool and cloud optical propertie a: τ = τ + τ + τ + τ + τ e τ ω = ga a cld + τ τ e cld a aer cld aer a aer
η η = i τ cld cld i cld τ + η τ + τ aer aer i aer where i i the i th phae function moment. Profile et A with the RT8C validation, the 60L-SDr profile et (Chevallier 2001) wa ued to provide profile of cloud water content. However, fixed profile of temperature, water vapour and trace gae were ued for all imulation. The example RTTOV9 profile (prof_101lev.dat) included with the code wa ued for thi purpoe. Unlike RT8C which required input profile on two different preure grid, the new code ue only one grid. In the verion of the code being teted thi wa limited to the 100 layer ued by the coefficient file. Input cloud profile to RTTOV9 can be pecified a one of ix different cloud type (five water cloud type or ice cloud) different layer can contain different cloud type, but only one cloud type i allowed on a given layer. RTTOV9 cloud profile were generated for each of the 67 60L-SDr profile including cloud liquid water data. The entire profile wa aumed to be a ingle cloud type, although the cloud type choen wa varied for different imulation. Profile of mixed cloud type are not conidered here. However, a the cloud type i imply ued to determine the optical propertie the preented reult here can be ued a a guide to mixed cloud condition. Under thee condition, the RTTOV error i expected to be no wore than that for the cloud-type modelled leat well Clear-ky comparion A cloud free comparion wa performed to verify the DISORT model wa functioning correctly. BT difference pectra are hown in Figure 1 for zenith angle of 0 and 60 degree. The agreement i very good difference are le than 0.1K for urface viewing channel and an order of magnitude maller for ounding channel. The difference between RTTOV and DISORT are due to differeing urface reflectivity aumption: RTTOV aume pecular reflection, while Lambertian reflection wa aumed for the DISORT imulation. Thi mean the reflected downward radiance in the DISORT imulation i independent of zenith angle, while the RTTOV reflected radiance increae with zenith angle (a emiivity decreae). If a urface emiivity of 1.0 i ued, then all RTTOV DISORT difference are reduced to ~0.001 K A the urface emiivity i the primary ource of bia between the RTTOV and DISORT model in the clear-ky cae, a urface emiivity of 1.0 wa ued for all cloudy-ky comparion.
Figure 1 Comparion of RTTOV9 and DISORT calculated clear-ky pectra. Water cloud There are 67 profile in the reduced Chevallier dataet containing cloud liquid water. For the purpoe of thi comparion they were aigned to one of three cloud type: tratu continental, tratu maritime, or cumulu maritime. The cumulu continental clean and continental polluted were not teted; however reult from Matricardi (2006) can be ued to infer their behaviour from reult preented here for other water cloud type. Figure 2 how the radiative impact of two different cloud profile auming the tratu continental cloud type, and the difference between the RTTOV and DISORT calculation. The maximum BT decreae in thee imulation i ~40K roughly twice that reported in Matricardi (2006) due to the larger range of cloud profile conidered. However the RTTOV error are conitent with thoe reported by Matricardi le than 1 K for thermal infrared and le than 5 K in the near infrared. Figure 3 how the RTTOV error againt cloud optical depth for two channel correponding to wavelength of 11.85 micron and 3.75 micron (choen to be cloe to the AATSR 12 and 3.7 micron channel ued in the RTTOV8 cloudy validation report). The thermal infrared channel how excellent agreement at nadir the bia i ~0.1 K for thick cloud (optical thickne greater than ~3), while at a zenith angle of 60 degree the bia reache only 0.5 K. For the near infrared channel, where cattering dominate, the biae are larger reaching ~3.5 K at nadir and 6 K at zenith angle of 60 degree. Figure 4 through 7 how equivalent plot for the maritime tratu and maritime cumulu cloud type. The maximum error i le for thee cloud type than tratu continental. Figure 8 how optical propertie of the three cloud type, howing that tratu continental ha both higher backcatter coefficient and ingle cattering albedo than the maritime cloud, in the near infrared, meaning that cattering i more ignificant to the cloudy radiance. The larger RTTOV-DISORT biae are aociated with higher value of the ingle cattering albedo ( ) and backcatter parameter (b). Thi follow from the (aborptivity) caling approximation ued in RTTOV9 to define effective extinction optical depth: ~ τ = τ + bτ e a Where τ = ωτ and ubcript e, a, and refer to extinction, aborption and cattering repectively. e Although in mot cae, the caling approximation lead to an overetimation of the ToA radiance, there are alo cae where the radiance i underetimated. Thi i particularly noticeable for the
thermal channel where the optical depth i between 0.1 and 10.0. All thee cae are aociated with very high cloud (cloud preent between 600 and 400 mbar) wherea profile with lower level cloud (lower than 600 mbar) give poitive RTTOV-DISORT biae. Thi i a reult of uing (iotropic) thermal emiion to approximate the cattered radiance which, in DISORT, i (correctly) non-iotropic. The water cloud type ued in RTTOV8 i cloet to maritime cumulu they would be virtually inditinguihable if plotted on Figure 8. Figure 9 how RT8C-DISORT biae for the water cloud type. Clear-ky tranmiion are calculated uing the tandard 43 level RTTOV8 coefficient, then interpolated to the 60 ECMWF model level for cloudy calculation. Comparing Figure 9 and Figure 7 we ee that RTTOV9 ha greatly improved the imulation in the thermal infrared biae at nadir have been virtually eliminated while at higher zenith angle they have been reduced from ~1 K to ~0.3 K. In the near infrared the maximum bia (correponding to optical depth of 10+) ha not been ubtantially altered. However biae at lower optical depth (1 to 10) are improved compared againt RT8C. Furthermore in both cae there i ignificantly le catter in the RTTOV9 plot than the correponding RT8C plot. Thi i primarily due to the different level and interpolation cheme employed by the model. In RT8C, Planck radiance are only calculated on the 43 fixed RTTOV level, then interpolated to the model level ued for the cloudy calculation. RTTOV9, however, perform no uch interpolation and therefore doe not uffer from uch interpolation artefact. Figure 10 how the biae when RT8C i compared againt RTTOV9. Again the water cloud type i ued in RT8C and maritime cumulu in RTTOV9. The outlying point with cloud optical depth ~0.4 i due to an interpolation artefact when the 60-level cloud profile wa interpolated to the 100 RTTOV9 level. Note that at low cloud optical depth the difference in clear-ky BT due to the updated pectrocopy and larger number of variable gae i apparent. Figure 2 Radiative impact of tratu continental cloud type (left) and difference between RTTOV and DISORT calculation (right) for two different cloud profile.
Figure 3 RTTOV-DISORT bia againt cloud optical depth for tratu continental cloud type. Optical thickne i total cloud extinction at pecified wavenumber. Figure 4 A Figure 2 for tratu maritime cloud type. Figure 5 A Figure 3 for tratu maritime cloud type.
Figure 6 A Figure 2 for cumulu maritime cloud type. Figure 7 A Figure 3 for cumulu maritime cloud type. Figure 8 Optical propertie of the three cloud type
Figure 9 RT8C DISORT biae for water cloud type. Clear-ky calculation performed on 43 RTTOV8 preure level; cloudy-ky calculation performed on 60 ECMWF model level. Figure 10 RT8C RTTOV9 difference. RT8C run uing water cloud type; RTTOV9 run uing cumulu maritime cloud type. Ice cloud DISORT imulation were not run for ice cloud a full phae function data were not available. However, the reult in Matricardi 2006 are conitent with the water cloud comparion here. Cirru cloud have a very low backcatter parameter, therefore we except only mall difference between RTTOV and DISORT (Matricardi report typical error of le than 0.2 K). Furthermore Matricardi note that RTTOV9 tend to underetimate the radiance from cirru cloud while cloud and aerool are overetimated. We have identified a imilar behaviour for water cloud above ~600 mbar. A the RTTOV9 caling approximation repreent cattered radiance with thermal emiion it follow that the biae will be dependent on the cloud temperature. Concluion RTTOV9 i a ignificant improvement on RT8C for imulating cloudy radiance compared to DISORT (full cattering model) run with equivalent optical propertie. Thi i particularly true for thermal infrared wavelength, where biae een in RT8C are all but eliminated. Partly, thi improvement will arie from the fact that, with RTTOV9, all calculation are performed on a conitent et of 100 level. Nonethele, cattering i ignificant at thermal wavelength and the very low biae indicate that the parameteriation of cattering i effective.
At near-infrared wavelength, biae are ignificantly improved up to cloud optical depth of about 10. At greater optical depth, the biae found relative to DISORT were imilar in magnitude to thoe with RT8C, but are, perhap, rather more conitent between cloud. A pattern of over-etimation of radiance for warm cloud and under-etimation for cold cloud i evident, which fit with the method of parameteriation of cattering, in which the aborption optical depth (and thu the thermal emiion) i caled independently of incident radiance. Future uggetion One le obviou, but ignificant, advantage of the RTTOV9 code compared with RT8C i the eae with which an alternative radiative tranfer equation olver (i.e. DISORT) could be incorporated. With RT8C it wa eaier to write a eparate computer program which called RTTOV8 to calculate the clear-ky tranmiion, then calculated the cloud optical propertie before calling DISORT. However, with RTTOV9 it wa poible to make the call to DISORT directly from RTTOV9 itelf, after making, comparatively, minor modification to ome of the optical propertie which RTTOV9 tore internally. If the variou RTTOV module were to have a more generic interface then it would be poible to implement new / alternative RTE olver without modifying the ret of the RTTOV code.