1、A practical guide to IR and MW radiative transfer using the RTTOV model and GUI,What is atmospheric radiative transfer?,Study of the of the propagation of electromagnetic radiation through the atmosphere which involves interactions with atmospheric constituents (gas molecules, aerosols, clouds, hydr
2、ometeors) and the surface. From a data assimilation perspective an RT model is the observation operator for assimilating passive visible/near-infrared, infrared and microwave satellite radiances into NWP models. RT models take NWP fields (p, T, q, trace gas profiles and surface variables) as input a
3、nd calculate top-of-atmosphere (TOA) satellite-seen radiances.,Definitions,Wavenumber: Radiance: Energy flux per unit solid angle per wavenumber in direction All matter with a temperature T0 Kelvin radiates energy. The radiance emitted by a black-body at wavenumber is given by the Planck function:Br
4、ightness temperature:,Radiation-particle interactions,(1) Absorption: radiation attenuation of by energetic modification (heat or chemical reaction),(1),(2) Emission: isotropic increase in radiation by molecular excitation due to absorption (Kirchhoffs law: emissivity = absorptivity),(3) Scattering:
5、 radiation attenuation by deviation of radiation from original direction; also increase in radiation by deviation of radiation into direction under consideration,What is RTTOV?,RTTOV = Radiative Transfer for TOVSTOVS = TIROS Operational Vertical SounderTIROS = Television Infrared Observation Satelli
6、te(RTTOV has been around for 25 years),What is RTTOV?,A fast radiative transfer model for passive VIS, IR and MW nadir-viewing instruments. Funded by EUMETSAT through the NWP SAF, developed by Met Office, Mto-France and ECMWF. Direct, TL, AD, K models. Applications: data assimilation, reanalysis, si
7、mulated imagery, 1D-VAR, .RTTOV v11: 700 users RTTOV v10: 600 users,RTTOV inputs,Vertical profiles of p, T, q Other optional trace gases: O3, CO2, CO, N2O, CH4 Viewing geometry: zenith and azimuth angles Surface variables: skin T, surface pressure, 10m wind u/v Surface emissivity (optional),Transmit
8、tance,Transmittance is related to optical depth by = e-(optical depth),Clear-sky RT equation,The computationally challenging term to calculate accurately is the transmittance or equivalently the optical depth where = e-(optical depth),*TOA = top of atmosphere,Weighting functions,Weighting function:T
9、he upwelling emission is an integral of the Planck function weighted by w(z).The largest contribution comes from the region where w(z) is largest i.e. where changes most rapidly with height.,Transmittance varies monotonically with height z.We can write the upwelling emission term as:,Polychromatic c
10、hannels,Passive IR/MW sensor channels are not monochromatic.,Ideally we would solve the RT equation at many wavelengths and integrate the resulting radiances over the channel spectral response function (SRF).In practice we integrate transmittances over the SRF and solve the RT equation once per chan
11、nel.,Line-by-line (LBL) models,LBL models embody the physics behind the absorption processes = accurate, but slow.,Line-by-line (LBL) models,LBL models embody the physics behind the absorption processes = accurate, but slow.Key idea: RTTOV parameterises off-line LBL calculations of optical depths to
12、 enable very rapid optical depth calculations for each instrument channel.,RTTOV optical depth calculation,83 diverse atmospheric profiles each at 6 zenith angles = 498 training profiles.,RTTOV optical depth calculation,83 diverse atmospheric profiles each at 6 zenith angles = 498 training profiles.
13、 Divide atmosphere into 53* layers defined by 54 fixed pressure levels.,*For hi-res sounders we also produce coefficients for 100 layers/101 levels.,RTTOV optical depth calculation,83 diverse atmospheric profiles each at 6 zenith angles = 498 training profiles. Divide atmosphere into 53* layers defi
14、ned by 54 fixed pressure levels. Calculate database of LBL optical depths for each layer at high spectral resolution for each training profile.,*For hi-res sounders we also produce coefficients for 100 layers/101 levels.,RTTOV optical depth calculation,Define a set of atmospheric “predictors” derive
15、d from input profile variables = there are separate sets of predictors for the optical depth due to mixed gases, water vapour and each additional trace gas.,RTTOV optical depth calculation,Define a set of atmospheric “predictors” derived from input profile variables = there are separate sets of pred
16、ictors for the optical depth due to mixed gases, water vapour and each additional trace gas. Integrate the LBL optical depths in each layer over each instrument channel SRF for every training profile.,RTTOV optical depth calculation,Regress layer optical depths onto predictors (pi) for each channel
17、= coefficients (ci) which are stored in a file for each instrument,* strictly speaking these are “pseudo” optical depths (RTTOV science and validation reports give more details),Output radiances and BTs,Optional surface emissivity calculation,RTTOV flow diagram,internal RTTOV calculations,Implicatio
18、ns for accuracy,Sources of error: Use of polychromatic optical depths,Implications for accuracy,Sources of error: Use of polychromatic optical depths Optical depth parameterisation (regression),Implications for accuracy,Comparison of TOA BTs from a simple forward RT model (upwelling emission plus su
19、rface term with unit emissivity) run with:,LBL channel-integrated optical depths RTTOV optical depths (from predictor regression),Implications for accuracy,Sources of error: Use of polychromatic optical depths Optical depth parameterisation (regression) Discretisation of atmosphere into homogenous l
20、ayers and associated interpolation,Implications for accuracy,Sources of error: Use of polychromatic optical depths Optical depth parameterisation (regression) Discretisation of atmosphere into homogenous layers and associated interpolation Input profiles values (including zenith angle) lying beyond
21、the limits of the training set,Jacobian (K) model,This calculates the derivatives of the simulated radiances or BTs with respect to each profile variable. For example:,It tells us how sensitive the satellite-seen radiance is to each individual profile variable.,RTTOV capabilities,Clear-sky visible/n
22、ear-IR, IR and MW radiances Internal sea surface emissivity and reflectance models Land surface emissivity and reflectance atlases Aerosol- and cloud-affected IR radiances Cloud- and precipitation-affected MW radiances Simulated Principal Components for hi-res IR sounders and more.,How to get RTTOV,Freely available, simply register here: http:/nwpsaf.eu/deliverables/rtm/index.htmCoefficient files are available here: http:/nwpsaf.eu/deliverables/rtm/rttov11_coefficients.htmlRTTOV forum: http:/www.nwpsaf.eu/forum/,Thanks for listening. Any questions?,
