You build, improve, and maintain the numerical models that produce weather forecasts, climate projections, and atmospheric research simulations. The work is at the intersection of atmospheric physics, applied mathematics, and scientific computing. You take what is known about how the atmosphere behaves — equations governing fluid motion, radiation transfer, cloud microphysics, boundary layer turbulence, chemical reactions — and encode that knowledge into software that runs on some of the most powerful computers on the planet.
The modeling challenge is both scientific and computational. The atmosphere operates across scales that span nine orders of magnitude — from molecular-scale cloud droplet formation to planetary-scale circulation patterns. No computer can resolve all of these scales simultaneously, so every model involves parameterizations — simplified representations of processes that happen at scales the model can't resolve. Getting these parameterizations right is one of the most consequential scientific challenges in the field, because every forecast and every climate projection depends on them.
This is one of the most computationally intensive scientific endeavors that exists. Global weather models run on supercomputers that rank among the world's fastest. The code bases are large, complex, and often written in Fortran — a language that is unfashionable in the broader software world but remains dominant in atmospheric modeling because of its performance characteristics on high-performance computing architectures.
Kitsune can talk through anything on this page — whether it might suit you, what to do next, questions this page doesn't answer. Everything here is yours to read either way.
The connection to "weather" in the public sense is more distant than you'd expect. An atmospheric modeler might go months without looking at an actual weather forecast in a professional context. The work is about the model — its physics, its numerics, its performance characteristics — not about today's weather. If you want to forecast tomorrow's weather, you become an operational forecaster; if you want to build the tool that makes tomorrow's forecast possible, you become a modeler.
The legacy code challenge is real. Some of the world's most important weather and climate models contain code that has been continuously developed for decades. Working with, understanding, and improving legacy scientific code is a significant part of the job. Modern software engineering practices are penetrating the field, but slowly.
The career sits at a strange intersection of scientific prestige and public invisibility. The models these people build underpin every weather forecast in the world and every climate projection that informs policy. The modelers themselves are almost entirely unknown outside the field. If recognition matters to you, this role will test that.
Bachelor's in physics, mathematics, computer science, atmospheric science, or engineering, followed by a graduate degree (master's or PhD) with a focus on numerical modeling, computational atmospheric science, or a related area. Strong programming skills are essential — Python, Fortran, C/C++, and experience with high-performance computing environments. Positions at national weather modeling centers (NCEP, ECMWF, Met Office), research labs (NCAR, GFDL, NASA GSFC), and universities. The field also employs people with backgrounds in applied mathematics and computational fluid dynamics who develop expertise in atmospheric applications.
The ML weather-model revolution is this role transforming; Fortran-physics expertise joined/displaced by ML modelling; hybrid physics-ML is the frontier.
Strong demand for ML-fluent modellers; future is hybrid physics-ML; pure legacy-parameterization work narrows.
People drawn to Atmospheric Modelerare often drawn to these — in the order they're closest. The ones marked sit in a different field entirely.