The Monte Carlo atmospheric radiative transfer model McArtim: Introduction and validation of Jacobians and 3D features

A new Monte Carlo atmospheric radiative transfer model is presented which is designed to support the interpretation of UV/vis/near-IR spectroscopic measurements of scattered Sun light in the atmosphere. The integro differential equation describing the underlying transport process and its formal solution are discussed. A stochastic approach to solve the differential equation, the Monte Carlo method, is deduced and its application to the formal solution is demonstrated. It is shown how model photon trajectories of the resulting ray tracing algorithm are used to estimate functionals of the radiation field such as radiances, actinic fluxes and light path integrals. In addition, Jacobians of the former quantities with respect to optical parameters of the atmosphere are analyzed. Model output quantities are validated against measurements, by self-consistency tests and through inter comparisons with other radiative transfer models.     

A Monte Carlo method for 3D thermal infrared radiative transfer

A 3D Monte Carlo model for specific application to the broadband thermal radiative transfer has been developed in which the emissivities for gases and cloud particles are parameterized by using a single cubic element as the building block in 3D space. For spectral integration in the thermal infrared, the correlated k-distribution method has been used for the sorting of gaseous absorption lines in multiple-scattering atmospheres involving 3D clouds. To check the Monte-Carlo simulation, we compare a variety of 1D broadband atmospheric fluxes and heating rates to those computed from the conventional plane-parallel (PP) model and demonstrate excellent agreement between the two. Comparisons of the Monte Carlo results for broadband thermal cooling rates in 3D clouds to those computed from the delta-diffusion approximation for 3D radiative transfer and the independent pixel-by-pixel approximation are subsequently carried out to understand the relative merits of these approaches.     

A Monte Carlo approach for the calculation of polarized light: application to an incident narrow beam

Monte Carlo approaches to compute multiple scattering of polarized light are examined. A Backward Monte Carlo (BMC) method is developed to solve the Stokes vector of the multiple scattered light for an inhomogeneous scattering medium with boundaries. A generalized form of the BMC method in vector notation is proposed. This method can determine the scattered light with sufficient accuracy in both intensity and polarization compared to the same calculation using the doubling-adding method for a plane parallel medium.For application to a narrow incident beam and an inhomogeneous medium, a modified BMC method is developed, borrowing a concept from the Forward Monte Carlo (FMC) method for the first scattering events. Furthermore, a modification of the total scattering matrix, i.e., the combination of the derived scattering matrix with its time inverse, is discussed. This BMC method can be used successfully for model calculations of lidar and other laser measurements of polarized light.     

Numerical test of an inverse polarized radiative transfer algorithm

A procedure is tested with which to determine the single-scattering albedo from polarization measurements of the angle-dependent intensity at two locations within, or on the boundaries of, a homogeneous finite or infinite atmosphere that scatters radiation according to the Rayleigh law with true absorption.     

Discrete ordinate and Monte Carlo simulations for polarized radiative transfer in a coupled system consisting of two media with different refractive indices

We present an algorithm for polarized radiative transfer in a vertically stratified system consisting of two plane-parallel media with different refractive indices. It is based on the discrete ordinate method and includes multiple elastic scattering, thermal radiation, Fresnel reflection and transmission, incident parallel-beam or isotropic radiation at the top of the upper medium and bidirectional reflection at the bottom of the lower medium. Comparisons with results from Monte Carlo simulations show that the discrete-ordinate code provides accurate results for all four elements of the Stokes vector (I, Q, U, and V) at a speed that is orders of magnitude faster.     

Three-dimensional radiative transfer modeling using the control volume finite element method

In the present study, a three-dimensional algorithm for the treatment of radiative heat transfer in emitting, absorbing and scattering media is developed. The approach is based on the utilization of control volume finite element method (CVFEM) which, to the knowledge of the authors, is applied at the first time to 3D radiative heat transfer in participating media. The accuracy of the present algorithm is tested by comparing its predictions to other published works. Comparisons show that CVFEM produces good results. Moreover, this approach permits compatibility with other numerical methods used for computational fluids mechanics problems.     

A linearized vector radiative transfer model for atmospheric trace gas retrieval

We present a plane parallel radiative transfer model for polarized light, that provides the intensity vector as well as the derivatives of the four Stokes parameters with respect to atmospheric trace gas profiles. These derivatives are essential for retrieval of height resolved trace gas information from satellite measurements of backscattered sunlight. The model uses the Gauss-Seidel iteration technique for solving the radiative transfer equation. For the first time, the forward-adjoint radiative perturbation theory is applied for the linearization of a radiative transfer model including polarization. The accuracy of the model is better than 0.025% for all four Stokes parameters and better than 0.03% for the derivatives.     

Influence of particle orientation on retrieving cirrus cloud properties by use of total and polarized reflectances from satellite measurements

The influence of ice crystal orientation was investigated on retrieving cirrus optical thickness (τ) and aspect ratio of ice crystals (Q) from satellite measurements using the total and polarized reflectances at a wavelength of . We considered columnar and plate like hexagonal ice crystals whose long axes are randomly oriented in the horizontal plane (2D model) with some amplitude of oscillation expressed by a Gaussian distribution function with the standard deviation of σ.The retrieved τ and Q values significantly depend on the assumption of σ, in particular for the plate type. Furthermore, the relationship between σ and the retrieved values depends on the solar, satellite, and target geometries. In our case study, for one target area, τ value retrieved using the 2D model with σ=5° was approximately twice larger than that using the 2D model with σ=20°, while the retrieved Q value was not significantly influenced by σ. For another target area, the τ(Q) retrieved using the 2D model with σ=5° was about 1.5 (1.8) times larger than that retrieved using the 2D model with σ=20°.     

A new 1D approximation for the solution of 2D radiative transfer problems

The simplified M-1D algorithm (M stands for modified) for the calculation of horizontal distribution of reflected brightness coefficient in 2D regions with large homogeneous pixels is presented. This algorithm is based upon modified 3^?N−1 1D-transport equations (where N is the number of large pixels) instead of one 2D-transport equation, usually used in such problems. The method does not rely on empiric assumptions on both optical properties of atmosphere or diffuse radiation intensity. Numerical results demonstrating the accuracy of the presented algorithm for simulating brightening and shadowing effects in a vicinity of the jump of optical properties given. The accuracy of the M-1D approximation strongly depends on the geometry and illumination conditions. However, it remains below 15% and can reach 1% for all cases studied and is strongly higher than the accuracy of usually employed independent pixel approximation. Time reduction via replacing 2D problem via modified 1D problem is about 17 times for all cases considered in this paper.     

Multiple-scaling methods for Monte Carlo simulations of radiative transfer in cloudy atmosphere

Two multiple-scaling methods for Monte Carlo simulations were derived from integral radiative transfer equation for calculating radiance in cloudy atmosphere accurately and rapidly. The first one is to truncate sharp forward peaks of phase functions for each order of scattering adaptively. The truncated functions for forward peaks are approximated as quadratic functions; only one prescribed parameter is used to set maximum truncation fraction for various phase functions. The second one is to increase extinction coefficients in optically thin regions for each order scattering adaptively, which could enhance the collision chance adaptively in the regions where samples are rare. Several one-dimensional and three-dimensional cloud fields were selected to validate the methods. The numerical results demonstrate that the bias errors were below 0.2% for almost all directions except for glory direction (less than 0.4%) and the higher numerical efficiency could be achieved when quadratic functions were used. The second method could decrease radiance noise to 0.60% for cumulus and accelerate convergence in optically thin regions. In general, the main advantage of the proposed methods is that we could modify the atmospheric optical quantities adaptively for each order of scattering and sample important contribution according to the specific atmospheric conditions.     

Prediction of radiative heat transfer in 3D complex geometries using the unstructured control volume finite element method

In this paper, a 3D algorithm for the treatment of radiative heat transfer in emitting, absorbing, and scattering media is developed. The numerical approach is based on the utilization of the unstructured control volume finite element method (CVFEM) which, to the knowledge of the authors, is applied for the first time to simulate radiative heat transfer in participated media confined in 3D complex geometries. This simulation makes simultaneously the use of the merits of both the finite element method and the control volume method. Unstructured 3D triangular element grids are employed in the spatial discretization and azimuthal discretization strategy is employed in the angular discretization. The general discretization equation is presented and solved by the conditioned conjugate gradient squared method (CCGS). In order to test the efficiency of the developed method, several 3D complex geometries including a hexahedral enclosure, a 3D equilateral triangular enclosure, a 3D L-shaped enclosure and 3D elliptical enclosure are examined. The results are compared with the exact solutions or published references and the accuracy obtained in each case is shown to be highly satisfactory. Moreover, this approach required a less CPU time and iterations compared with those of even parity formulation of the discrete ordinates method.     

Stochastic radiative transfer model for mixture of discontinuous vegetation canopies

Modeling of the radiation regime of a mixture of vegetation species is a fundamental problem of the Earth's land remote sensing and climate applications. The major existing approaches, including the linear mixture model and the turbid medium (TM) mixture radiative transfer model, provide only an approximate solution to this problem. In this study, we developed the stochastic mixture radiative transfer (SMRT) model, a mathematically exact tool to evaluate radiation regime in a natural canopy with spatially varying optical properties, that is, canopy, which exhibits a structured mixture of vegetation species and gaps. The model solves for the radiation quantities, direct input to the remote sensing/climate applications: mean radiation fluxes over whole mixture and over individual species. The canopy structure is parameterized in the SMRT model in terms of two stochastic moments: the probability of finding species and the conditional pair-correlation of species. The second moment is responsible for the 3D radiation effects, namely, radiation streaming through gaps without interaction with vegetation and variation of the radiation fluxes between different species. We performed analytical and numerical analysis of the radiation effects, simulated with the SMRT model for the three cases of canopy structure: (a) non-ordered mixture of species and gaps (TM); (b) ordered mixture of species without gaps; and (c) ordered mixture of species with gaps. The analysis indicates that the variation of radiation fluxes between different species is proportional to the variation of species optical properties (leaf albedo, density of foliage, etc.) Gaps introduce significant disturbance to the radiation regime in the canopy as their optical properties constitute major contrast to those of any vegetation species. The SMRT model resolves deficiencies of the major existing mixture models: ignorance of species radiation coupling via multiple scattering of photons (the linear mixture model) or overestimation of this coupling due to neglecting spatial clumping of species (the TM approach). Thus, based on the former experience with mixture modeling, this study establishes an advanced theoretical basis for future mixture applications.     

Retrieval of water cloud characteristic from active sensor data using the analytical solution of radiative transfer equation

An analytical forward model and numerical algorithm for retrieving the parameters of water cloud of earth atmosphere from optical measurements carried out by satellite-based lidars is presented. The forward model, based on the analytical solution of the radiative transfer equation, is used to fit the temporal profile of the laser light pulses backscattered from the cloud layers. The cloud parameters extracted from the analysis at each position on earth include the transport mean free path, the average radius of water drops, the density of drops, the scattering length, the scattering cross section, the anisotropy factor, and the altitude of top level of major clouds. Also estimated is the possible thickness of cloud layers. The efficacy of the approach is demonstrated by generating parameters of water cloud using the data collected by NASA's cloud-aerosol lidar and infrared pathfinder satellite observations (CALIPSO) satellite when it passed through North America on August 7, 2007.     

A simple new radiative transfer model for simulating the effect of cirrus clouds in the microwave spectral region

The upwelling atmospheric radiation in the millimeter wave spectral range is influenced by the presence of cirrus clouds. A plane parallel radiative transfer model which can take into account the effect of multiple scattering by ice particles in the cirrus has been developed and is used to simulate the brightness temperatures as they would be measured by a satellite instrument. The model uses an iterative procedure to solve the radiative transfer equation. The formulation of the model is such that it can easily be adapted to treat the full specific intensity vector instead of just the scalar total intensity. A convergence test for the model is explained and two cirrus cloud scenarios are simulated. The results illustrate the linearity of microwave radiative transfer for not too strong cirrus clouds in this frequency region.     

Efficient unbiased variance reduction techniques for Monte Carlo simulations of radiative transfer in cloudy atmospheres: The solution

We present five new variance reduction techniques applicable to Monte Carlo simulations of radiative transfer in the atmosphere: detector directional importance sampling, n-tuple local estimate, prediction-based splitting and Russian roulette, and circum-solar virtual importance sampling. With this set of methods it is possible to simulate remote sensing instruments accurately and quickly. In contrast to all other known techniques used to accelerate Monte Carlo simulations in cloudy atmospheres - except for two methods limited to narrow angle lidars - the presented methods do not make any approximations, and hence do not bias the result. Nevertheless, these methods converge as quickly as any of the biasing acceleration techniques, and the probability distribution of the simulation results is almost perfectly normal. The presented variance reduction techniques have been implemented into the Monte Carlo code MYSTIC (“Monte Carlo code for the physically correct tracing of photons in cloudy atmospheres”) in order to validate the techniques.     

Radiation symmetry test and uncertainty analysis of Monte Carlo method based on radiative exchange factor

Using Monte Carlo method, the paper investigates the radiative heat transfer in participating media. Based on the radiative exchange factor, an uncertainty analysis of Monte Carlo method is undertaken and the corresponding mathematical expressions are deduced to predict its accuracy. Furthermore, randomness properties of pseudorandom number generators are investigated, and a model to test radiation symmetry is adopted to validate the performance of some generators. The paper studies the effects of energy bundle numbers, discretization schemes, emission location, optical thicknesses, wall emissivity and CPU time on the numerical accuracy. In addition, the simulation results are proved to give a reference for using Monte Carlo method, which is applicable for calculation of the radiative exchange factor.     

Three-dimensional vector radiative transfer in a semi-infinite, Rayleigh scattering medium exposed to a polarized laser beam: spatially varying reflection matrix

Three-dimensional vector radiative transfer in a semi-infinite, Rayleigh scattering medium exposed to a polarized, Gaussian laser beam directed perpendicular to the surface is studied. The focus of this investigation is the 4×4, spatially varying reflection matrix that can be used to determine the normally backscattered radiation when the polarization of the incident radiation is specified. An inverse integral transform is used to construct the spatially varying reflection matrix from the generalized reflection matrix found in a previous study. The elements of this matrix depend on location specified by optical radius and azimuthal angle. The azimuthal variation is found by performing part of the inverse transform analytically, while the radial variation is described by five functions that are calculated numerically via an inverse Hankel transform. Benchmark numerical results for these five functions are presented, and the effects of beam radius and particle concentration are discussed. Expressions that describe the behavior of the reflection functions at small and large optical radii are developed, and comparisons are made to the one-dimensional and scalar situations. The scalar approximation fails to predict the three-dimensional effects produced by the polarized beam, and even when the incident radiation is unpolarized, the error in the scalar reflection function can be as high as 20%.     

Modified finite-volume method based on a cell vertex scheme for the solution of radiative transfer problems in complex 3D geometries

A modified finite-volume method based on a cell vertex scheme was applied to solve radiative transfer problems within a participating medium of complex three-dimensional shaped domain. The computational spatial domain of interest was divided into four-node tetrahedron elements with unstructured meshes while the adopted formulation was combined with a closure relation based on an exponential scheme. The studied medium was assumed to be grey, non-scattering and was bounded by black surfaces. Our results were then compared with those found in other articles on the subject. The approach shows a very good level of performance for wall heat transfer evaluation. Accurate results were obtained on coarse computational meshes and solution errors were found to decrease with grid refinement.     

Three-dimensional vector radiative transfer in a semi-infinite, Rayleigh scattering medium exposed to spatially varying polarized radiation: generalized reflection matrix

Three-dimensional vector radiative transfer in a semi-infinite medium exposed to spatially varying, polarized radiation is studied. The problem is to determine the generalized reflection matrix for a multiple scattering medium characterized by a 4×4 scattering matrix. A double integral transform is used to convert the three-dimensional vector radiative transfer equation to a one-dimensional form, and a modified Ambarzumian's method is then applied to derive a nonlinear integral equation for the generalized reflection matrix. The spatially varying backscattered radiation for an arbitrarily polarized incident beam can be found from the generalized reflection matrix. For Rayleigh scattering and normal incidence and emergence, the generalized reflection matrix is shown to have five non-zero elements. Benchmark results for these five elements are presented and compared to asymptotic results. When the incident radiation is polarized, the vector approach used in this study correctly predicts three-dimensional behavior, while the scalar approach does not. When the incident radiation is unpolarized, both the vector and scalar approaches predict a two-dimensional distribution of the intensity, but the error in the scalar prediction can be as high as 20%.     

Assessment of the re-ordered wide band model for non-grey radiative transfer calculations in 3D enclosures

The objective of the present study is to evaluate variations of the re-ordered wide band model for non-grey radiative transfer calculations in 3D enclosures using the discrete ordinates method. First, the performance of various angular and spatial discretisation schemes of the discrete ordinates method is investigated. Then, several formulations, averaging procedures, and scaling methods of the re-ordered wide band model are tested, and the results are validated against those of a statistical narrow band model. The grey gases formulation using three optimised absorption coefficient is found to be the most efficient method.