The project is organized in 4 major tasks.
Task 0: coordination of the project
Task 1: “Transition-sampling and computer-graphics”: Data analysis and modeling for revisiting of the radiative-forcing formulation. Sampling strategies and variance reduction techniques. Spectral database modeling, processing and storage for efficient line-sampling strategies.
Task leader: Mathias Paulin
Summary : This task deals with the optimization of Galtier et al’s Monte Carlo algorithm, including sensitivity estimates. In this context, when we refer to a modeling effort, this never means that we change the physical model itself. We only model the multi-dimension integral in order to reduce the variance of the corresponding statistical estimate. The choice that will be finaly made of one model or another will not change the estimated quantity: it will introduce no bias. Only the convergence speed (the number of required samples) and the computation requirement (the computation time associated to each sample) will be impacted. This task contains three subtasks:
Task 1.1 Modeling and iterating over a high-dimensional integration domain.
Task 1.2 Multiple Importance Sampling.
Task 1.3 Storage and processing of spectral database for efficient Radiative forcing.
Task 2: “Radiative forcing from 4D fields”: Validation of the radiative forcing computed by the codes developed in task 3. Building a statistical 3D geometry of clouds from GCMs outputs.
Task leader: Jean-Louis Dufresne
Summary: This task gathers the climatic dimension of the project, from deciding and evaluating the successive test-cases, to the detailed picturing of how these new simulation potentials may be used by the climate-change community. First, the code developed in Task 3.1 will be validated and assess for a few numbers (hundred) of atmospheric profiles with clear sky conditions (no clouds). This will be done in the frame of the RFMIP project which has a dedicated research themes on testing radiative codes (https://rfmip.leeds.ac.uk/rfmip-irf/). The atmospheric profiles are produced by general circulation models (GCMs) with a typical horizontal resolution of 50 to 200 km and a vertical resolution of a few tens to a few hundreds meters. The goal of Task 2.2 is to provide the necessary information to a statistical description of the 3D geometry of clouds for each atmospheric profile from the information provided by the GCMs. At the end of the project, the last task (T-2.3) is to perform, analyse and promote the computation of the radiative forcing over the whole globe, over climatic durations, considering the 3D radiative effects. This task contains three subtasks:
Task 2.1 : Validation and analysis of the radiative code in clear sky conditions or with the plan parallel assumption
Task 2.2 : Building a 3D structure of clouds using GCMs outputs
Task 2.3 : Performing and promoting the 4D computation of the radiative forcing at global scale
Task 3: Software development: Development and assessment of a radiative tool for the climate community. Recoding of Galtier’s codes using the star-engine library and adding sensitivity evaluation. Interfacing with GEISA/HITRAN. Interfacing with GCM outputs. Implementation of the sampling strategy of Task 1.3.
Task leader: Richard Fournier
Summary: This task deals with the coding and testing of a succession of Monte Carlo solvers with gradually enhanced calculation potentials and interfaces. In a first phase (Tasks 3.1 and 3.2) the algorithm of Galtier et al will be implemented as is, that is to say independently of the better-understanding and technical solutions resulting of Task 1. The only algorithmic change will be to add the evaluation of sensitivities (algorithm already available, see preliminary work in paragraph I-b-4). In the second phase (Task 3.3), the objective will be precisely to upgrade the software according to Task 1-conclusions. This division guaranties that in any case, although not optimum, a new software will be distributed to the community and that this software will meet the principal objective (evaluating radiative forcing at the global scale, integrated over climatic durations).
According to this development strategy, five versions of the radiative-transfer code will be delivered, with graduated application and distribution potentials. The first version impacts the project strongly as both early Task 1 and early Task 2 depend on its achievement. But this first version is essentially the recoding of available softwares in a well known environment and the corresponding development-time is accurately predictable. This task contains three subtasks:
Task 3.1 : Rewriting M. Galtier’s PhD code
Task 3.2 : Accessing spectroscopic and climatic data
Task 3.3 : Including the validated outputs of Task 1.3