Find the pre-print on arXiv of this work with my fab supervisors, @PeterAGWatson and @laurence_ai, Elizabeth Kendon of @metoffice and @sumanravuri. And look out for more that we’ll be doing e.g. using it with flood models and multi-variate output. https://t.co/ygB6ThXb8v

In our paper we show these are realistic samples of rainfall that match the distribution of the climate model, with evidence of good performance even for up to ~1-in-100 year extremes.

We have developed a diffusion model to emulate the CPM, taking coarse-resolution climate model fields as inputs. We call this a CPM Generative EMulator (CPMGEM). We can produce multiple realistic samples of high-res daily mean rainfall fields for a single set of coarse inputs.

To adapt to a changing climate in the UK, we need to know how rainfall is going to change at a local (~1-10km) scale e.g. for flood modelling. But running climate simulations at this resolution is very expensive.

Excited to announce a pre-print of my first paper as part of my PhD at @BristolUni! We’ve created a generative ML emulator of a UK convection-permitting climate model (CPM). It’s able to produce simulations of high-res precipitation at far lower computational cost.

Pfft… NIPS sold out years ago. Props to artisanal, small-batch conferences like AISTATS, UAI, and COLT for keeping it real.

Nothing like some home networking to wake you up in the morning. Now which wire gets me to the Internet? http://t.co/JuAy2ByswW