Developing a model to capture all aspects of a complex dynamical system is an immense task, and each model will have deficiencies in some areas, such as global climate models having difficulty in capturing tropical intraseasonal variability such as the Madden--Julian oscillation. Besides complex models, it is possible to create simplified, low-dimensional models to capture specific phenomena while ignoring many aspects of the full system. Here, we propose a strategy to allow complex models to communicate with simplified models throughout a simulation. The communication allows one to leverage the strengths of each model, without needing to change their dynamics, to mitigate model error. Furthermore, to ensure ease of implementation in complex systems, the strategy is based on common data assimilation techniques that are normally used to combine models and realworld data. This strategy is investigated here in a test case that is nonlinear, non-Gaussian, and high-dimensional (approximately 105 degrees of freedom), and the multiple models have different state spaces. In particular, it is an idealized tropical climate model in three spatial dimensions. The multimodel communication strategy is seen to mitigate model error and reproduce statistical features akin to those of the truth model when the communication is sufficiently frequent. In these tests, the low-dimensional model contributes only two degrees of freedom, which suggests that, in some systems, large amounts of model error can possibly be reduced by focusing on a small set of model components.

In climate predictions, clouds are the leading source of uncertainty. This is partly because, to simulate the fluid dynamics of climate over the entire globe, a large grid spacing must be used, so clouds are a subgrid-scale parametrization rather than a resolved feature. Here, a framework is investigated with finer grid spacing of O(1) or O(10)km so that some clouds are not subgrid-scale; instead, clouds evolve on the numerical grid. This cloud evolution is achieved using stochastic modeling. Hence, the framework is idealized in the sense that the full fluid dynamics of cloud circulations is still not resolved, and simplified vertical structures are used. Nevertheless, the fluid dynamics model includes evolving clouds that interactively adjust in size, shape, lifetime, and regional coverage. In addition, different cloud types are included with different roles in the climate system, including deep convective clouds and also boundary-layer clouds such as shallow cumulus and stratocumulus clouds. Other basic aspects of the idealized climate system are planetary-scale circulations (e.g., Walker circulation) and radiation. With these ingredients (evolving clouds, planetary-scale circulations, and radiation), the framework has the potential for idealized investigations of climate change with interactive cloud-radiative feedback of individual clouds. Here, the formulation of the model equations is presented, and numerical simulations are shown to illustrate the model dynamics and climate change.

This paper presents the authors' efforts on developing a tool that can help the process of parallelizing a serial code, at the level of MPI-based parallelization.This is a traditionally interesting topic, even more interesting at the current stage of migrating C1 GMDD Interactive commentPrinter-friendly version Discussion paper from homogeneous multi-core clusters to heterogeneous clusters with GPUs or other many-core accelerators.This is overall a quite interesting work.The authors have also done a good job to provide not only a tool but also demonstrations in a number of different applications or kernels.My major question is about the comparison with other similar efforts in the field.Parallelization, especially MPI-based parallelization, has been around for many decades.Groups from both computer science and application domains have existing projects that try to derive languages, compiler, tools to support better and easier parallelization.Therefore, an introduction of such a tool should come with a comprehensive overview of existing efforts.Also, for demonstrating the efficiency and performance of the proposed tool, comparisons should be made on both the parallel performance achieved, and the extra coding efforts needed.In the current paper, we only see results of the parallel code in the proposed tool, but not sure how good it is when compared to other similar tools, or languages, such as the Unified Parallel C project from Berkeley.

This paper addresses the issue of parallelizing a serial code used in geophysical applications.An open-source module named EZ Parallel is produced for this purpose to automatically with minimum invasion and knowledge of the user parallelize a typical