Computational fluid dynamics: Its carbon footprint and role in carbon reduction

Abstract

Turbulent flow physics regulates the aerodynamic properties of lifting surfaces, the thermodynamic efficiency of vapor power systems, and exchanges of natural and anthropogenic quantities between the atmosphere and ocean, to name just a few applications of contemporary importance. The space-time dynamics of turbulent flows are described via numerical integration of the non-linear Navier–Stokes equation—a procedure known as computational fluid dynamics (CFD). At the dawn of scientific computing in the late 1950s, it would be many decades before terms such as “carbon footprint” or “sustainability” entered the lexicon, and longer still before these themes attained national priority throughout advanced economies. The environmental cost associated with CFD is seldom considered. Yet, large-scale scientific computing relies on intensive cooling realized via external power generation that is primarily accomplished through the combustion of fossil fuels, which leads to carbon emissions. This paper introduces a framework designed to calculate the carbon footprint of CFD and its contribution to carbon emission reduction strategies. We will distinguish between “hero” and “routine” calculations, noting that the carbon footprint of hero calculations—which demand significant computing resources at top-tier data centers—is largely determined by the energy source mix utilized. We will also review CFD of flows where turbulence effects are modeled, thus reducing the degrees of freedom. Estimates of the carbon footprint are presented for such fully and partially resolved simulations as functions of turbulence activity and calculation year, demonstrating a reduction in carbon emissions by two to five orders of magnitude at practical conditions. Besides generating a carbon footprint, the community's effort to avoid redundant calculations via turbulence databases merits particular attention, with estimates indicating that a single database could potentially reduce CO2 emissions by approximately O(1) × 106 metric tons.