CLOUD AND MODEL SETUP

• testing domain size: 19 vertical layers, 10 x 10 grid boxes in the horizontal • new solver is spinned up using 100 time steps in a Rayleigh atmosphere • after 101 time steps, we suddenly introduce

• a simple rectangular two-layer 3D cloud between 900 hPa and 800 hPa height and • another simple, horizontally shifted rectangular one-layer 3D cloud between 750

hPa and 700 hPa height

ADVECTIVE TREATMENT OF RADIATION

• in one radiation time step, photons are only transported to neighboring grid boxes

• ingoing fluxes of the previous time step as well as the transport coefficients of the corresponding grid box are used to compute updated outgoing fluxes

• in every radiation time step, we iterate once through the grid boxes • number of iterations through grid boxes per radiation time step thus equals

that in a 1D independent column approximation used nowadays

Schematic illustration of the advective treatment of radiation using a point source at top of atmosphere:

MOTIVATION

Development of a Fast 3D Radiative Transfer Solver for NWP Models

R. Maier1, B. Mayer1, C. Emde1 & A. Voigt2

1) Meteorologisches Institut München, Ludwig-Maximilians-Universität München 2) Institut für Meteorologie und Geophysik, Universität Wien

contact: richard.maier@physik.uni-muenchen.de

REFERENCES Emde et al. (2016): The libradtran software package for radiative transfer calculations (version 2.0.1). Geoscientific Model Development, 9(5):1647-1672 Jakub and Mayer (2015): A three-dimensional parallel radiative transfer model for atmospheric heating rates for use in cloud resolving models — The TenStream solver, Journal of Quantitative Spectroscopy and Radiative Transfer, 163:63-71 Mayer and Kylling (2005): Technical note: The libRadtran software package for radiative transfer calculations - description and examples of use. Atmos. Chem. Phys., 5: 1855-1877

CLASSICAL RADIATIVE TRANSFER

• atmosphere is split into independent vertical columns • every column is split up into different layers • layer interfaces are called levels • one-dimensional radiative transfer:

at every atmospheric level, just up- and downward irradiance is calculated

• irradiances are calculated from scratch in every model time step

• up- and downward irradiances at level i are in general dependent on all other irradiances above and below it

TENSTREAM SOLVER (JAKUB AND MAYER, 2015) JU

IMPLEMENTATION INTO LIBRADTRAN

The new solver that is treating radiation like advection was implemented into the libRadtran radiative transfer library to compare the new model with other, well-established solvers using the exact same model input and parameterizations.

simple 2D visualization of the streams TenStream uses to represent 3D radiation in

the case of sun shining in from the right

• 3D radiative transfer solver using three direct and ten diffuse streams to represent 3D radiation

• transport coefficients precomputed using Monte Carlo methods

Definition of the streams: ingoing diffuse stream (6 in 2D, 10 in 3D) outgoing diffuse stream (6 in 2D, 10 in 3D) ingoing direct stream (2 in 2D, 3 in 3D; shown for sun shining in from the right) outgoing direct stream (2 in 2D, 3 in 3D; shown for sun shining in from the right)

IMPLICATIONS AND CONVERGENCE BEHAVIOR

DRAWBACKS

• horizontal transport of energy is neglected

• three-dimensional extension of the algorithm is computationally very expensive, since the irradiance in every model grid box is in general dependent on all other grid boxes

CONSEQUENCE

need for a fast 3D radiative transfer solver

Time Step 1 We start with a point source shining in from the right at top of atmosphere.

Time Step 2 Outgoing direct and diffuse fluxes are calculated using the corresponding grid boxes’ transport coefficients and ingoing fluxes of the previous time step.

Time Step 3We repeat the procedure from time step 2 by iterating through all the grid boxes and calculating updated outgoing fluxes.

• first order radiative effects are considered immediately, but it takes some time for the individual model layers to communicate with each other

• new solver converges towards TenStream fluxes and heating rates in the limit of a large number of iterations

• rate of convergence primarily dependent on the distance between different clouds and the information that has to be exchanged over that distance