Environmental Science and Engineering Seminar

In this talk I summarize some of the benefits we have found from introducing stochasticity into weather and climate models. This includes not only improvements in forecast skill, but also improvements in the climatology of the models. Some improvements are highly nonlinear - e.g. increasing the persistence time of circulation regimes. In recent years my group has been working on a consequence of stochasticity - accurate low-precision (16-bit) numerical schemes with stochastic rounding. I look forward to the day when stochasticity can be represented in hardware.