Application of a Monte Carlo solar radiative transfer modelin the McICA framework

Large‐scale atmospheric models (LSAMs) that utilize the Monte Carlo Independent Column Approximation (McICA) have, thus far, paired McICA only with two‐stream approximations (TSAs) of the radiative transfer equation. In this study, the short‐wave TSA is exchanged for a Monte Carlo (MC) photon transport model. More than 44 000 domains of cloud properties retrieved from A‐Train satellite data, each measuring 256 km in length, were used to assess the noise characteristics of TSA‐ and MC‐based McICA models. It appears as though application of an MC algorithm in McICA will be both beneficial and tractable for LSAMs. This is because known levels of acceptable radiative noise produced by TSA‐based McICAs can be achieved with small numbers of MC photons. The greatest concern with the TSA McICA has been noise associated with heating rates for cloudy layers. But with as few as 500–1000 photons per simulation, the MC McICA reduces cloudy layer heating rate errors by typically ∼20%. Furthermore, since MC models can utilize detailed descriptions of cloud particle scattering phase functions and TSAs use only corresponding asymmetry parameters, TSA‐based McICAs, on average, overestimate all‐sky top‐of‐atmosphere reflected flux density at small solar zenith angles θ 0 by ∼3 W m −2 and underestimate it at large θ 0 by ∼1 W m −2 ; vice versa for surface net flux density. Systematic biases such as these are important when attempting to balance an LSAM's energy budgets and when making detailed estimates of radiative forcings due to anthropogenic activities.