Efficient All-Weather (Cloudy and Clear) Observational Operator for

Satellite Radiance Data Assimilation 

October 2004 – September 2006

Dr. Manajit Sengupta, Dr. Tomislava Vukicevic and Dr. Thomas H. Vonder Haar

CIRA/CSU

 Dr. K. Franklin Evans

CU

GOAL

To build an efficient OBSERVATIONAL OPERATOR with the following features

(1) Capability to handle all visible and infrared satellite channels

(2) Capability to handle all types of weather (cloudy and clear)

(3) Can be used with any data assimilation system (1DVar, 3DVar, 4DVar, EnKF)

(4) Can use output from any NWP model

Schematic of the Observational Operator

NWP Model Output

pCRTM: Compute gaseous transmission

Anomalous diffraction: Compute hydrometeor optical properties

RT Models: Compute radiance

Observations Compute cost function

Fo

rwar

dA

djo

intAdjoint of RT Models

Adjoint in pCRTM Adjoint of ADT

Features of the Observational Operator

Required Inputs to Forward Model

Profiles of Pressure(P), Temperature (T), WV Mixing ratio (r)

Surface Temperature(Ts), Surface Type

Hydrometeor Mixing Ratio (rc) Number Concentration (Nc) for various hydrometeor types (currently 7) (Cloudwater, pristine ice, aggregates, graupel, hail, snow and rain)

Solar zenith angle (date and time of day)

Satellite and channel, location of grid point, satellite position for computing zenith and azimuth with respect to sun

Components of the Forward Model

pCRTM: Inputs: P, T, r, Ozone(O)

Output: gaseous transmittance(t)

used to compute gaseous absorption optical depth ( a)

Anomalous Diffraction Theory: Inputs: rc and Nc,

Assumptions: mass- area-dimensional power law, gamma distribution, tunneling factor

Outputs: Optical depth (c), Single scatter albedo (o), Asymmetry parameter (g)

Radiative Transfer: SHDOMPPDA (visible, near-IR) or Delta-eddington (IR)

Inputs: T, cumulative a, c, o, legendre-coefficients using HG phase function using g

Output: Radiance, Reflectance (visible channels)

Adjoint modelling

Adjoints constructed using Transformation of Algorithms in Fortran (TAF) software available from FastOpt

Radiative Transfer: ADJ SHDOMPPDA or ADJ EDDINGTON

Inputs: radiance_ad

Outputs: T_ad, a_ad, c_ad, o_ad, g_ad, Ts_ad

Adjoint pCRTM: Inputs: t_ad from a_ad

Outputs: P_ad, T_ad, r_ad, O_ad

Adjoint MADT: Inputs: c_ad, o_ad, g_ad

Outputs: rc_ad and Nc_ad P_ad, T_ad

Status of Proposed Upgrades to the Observational Operator

Previous Proposed

Research version of OPTRAN and its adjoint

pCRTM forward and adjoint (implemented)

Anomalous diffraction theory

SHDOM with tangent linear

Coupled to RAMS output Standalone version (being implemented)

SHDOMPPDA has forward and adjoint (being implemented)

Exact scattering: Mie theory for cloudwater and Ice scattering tables (under development)

SHDOMPPDA: A Radiative Transfer Tool for Assimilation

Goal: Flexible, fast, and accurate forward and adjoint models for cloudy scattering radiative transfer

Based on SHDOMPP forward model Solves solar and/or thermal emission RTArbitrary accuracy depending on angular and spatial resolution

parameters chosen SHDOM modified for plane-parallel RT Source function represented with spherical harmonics on an

optical depth grid Solution method: source function iteration using spherical

harmonics and discrete ordinates Automatic adaptive layers to minimize error

Building the SHDOMPPDA Adjoint

Philosophy: minimize hand coding of adjoint by modifying forward model to persuade TAF compiler to generate a decent adjoint

Solution iterations with adapative grid

Solution iterations with fixed grid,compute series acceleration parameters

Solution iterations with fixed grid,use series acceleration parameters

Divide solution iterationroutine into three

Generate tangent linear andadjoint with TAF operatingonly on last solution routine

Trick TAF into avoiding extensive loop recomputations by using separate arrays for iterated variables and putting in unneeded STORE directives

Testing SHDOMPPDA

Tests made for solar and thermal RT with multiple particle species

Tangent linear compared with finite difference of forward model; < 1% differences (single precision)

Adjoint tested with tangent linear by comparing inner products: (dy H dx) = (dx HT dy) to machine precision for many random

input vectors dx (profiles of LWC, rmm

, T; surface

temp/albedo) and output vectors dy (radiances).

Status, Expectations and Limitations

Final product will be state of the art Observational Operator with complete visible and infrared capabilities in both clear and cloudy scenarios

Capable of handling different hydrometeor types and their combinations

Dependent on availability of CRTM coefficients and therefore can handle visible channels only if no gaseous absorption is assumed.

Will depend on the availability of input variables from NWP model

Sensitivity of GOES Sounder Channels in presence of clouds

Ch 11

Ch 12

Sensitivity of Water Vapor channels in ice clouds

Mid-level Moisture

Upper-level Moisture

Sensitivity of Temperature channels in ice clouds

Ch 5

Ch 13

Low-level temperature

Conclusions

Modeled ice clouds significantly improved by GOES IR observations

Modeled liquid clouds not improved as imager IR observations have little to no information below ice clouds

Need water vapor information below ice cloud to improve forecast of multi-layer clouds

Sounder channel have sensitivity below certain ice clouds and can potentially provide additional information.

References

Vukicevic, T., M. Sengupta, A. S. Jones, T. Vonder Haar, 2005: Cloud Resolving Satellite Data Assimilation: Information content of IR window observations and uncertainties in estimation; J. Atmos. Sci, under review.

Vukicevic, T., T, Greenwald, M. Zupanski, D. Zupanski, T. Vonder Haar, and, A. Jones, 2004: Mesoscale cloud state estimation from visible and infrared satellite radiances. Mon. Wea. Rev. , 132, 3066-3077.

Greenwald, T. J., T. Vukicevic, and, L. D. Grasso and T. H. Vonder Haar, 2004: Adjoint sensitivity analysis of an observational operator for cloudy visible and infrared radiance assimilation. Q. J. R. Met. Soc. 130, 685-705.

Greenwald, T. J., R. Hertenstein, and T. Vukicevic, 2002: An all-weather observational operator for radiance data assimilation with mesoscale forecast models. Mon. Wea. Rev., 130, 1882-1897.