Abstract

Many of today’s global crises, such as the 2015 migration crisis in Syria and the 2020 COVID pandemic, have a sudden evolution that complicates the preparation of a community response. Simulation-based forecasts for such crises can help to guide the development of mitigation policies or inform a more efficient distribution of support, but only if they are delivered in time and sufficiently robust. In this talk I will focus on recent advances in the parallel Flee model and our recent emergency forecasting work for Ukraine in UNHCR-funded research efforts, where among other things we had to adapt models and deliver an outage-driven displacement forecast with two months advance notice. The SEAVEA toolkit for Verification, Validation and Uncertainty Quantification, which can be used by researchers in any domain, has enabled us to do these kinds of activities. With it, we have been able to automatically validate our models against a wide range of conflicts, investigate the sensitivity of the results to key assumptions and deliver ensemble forecasts accounting for a range of scenario developments. In addition, we have been able to greatly speed up the time-to-forecast through workflow automation and effective use of HPC pilot frameworks such as QCG-PilotJob.