Quantifying uncertainty in simulations of the West African Monsoon with the use of surrogate models

Simulating the West African monsoon (WAM) system using numerical weather and climate models suffers from large uncertainties, which are difficult to disentangle due to highly non-linear interactions between different components of the WAM. We propose a new approach to this problem by emulating a full-blown numerical model, the ICON model of the German Weather Service, through statistical surrogate models. The ICON model was run during the rainy seasons in four years in a nested limited-area mode. The uncertainty contributions of six selected model parameters were investigated. To this end, we employed a sampling strategy to obtain model parameter combinations for a manageable number of ICON model runs. Surrogate models were then constructed to describe a relationship between the model parameters and selected Quantities of Interest (e.g. characteristics of the African and Tropical easterly jets or the Saharan heat low) to employ sensitivity and parameter studies. For better interpretation a local parameter analysis based on the output fields was conducted using the same setup. Results reveal the complex nature of the WAM system and indicate for which parameters (and thus processes) uncertainties need to be reduced to lower the spread in the outputs. Among the considered parameters, the entrainment rate and the terminal fall velocity of ice show the greatest effects, where larger values lead to a decrease of cloud cover and precipitation, and to an intensification of the Saharan heat low, despite distinct regional differences. The evaporative soil surface also shows a significant effect, mostly on temperature and cloud cover.
The 28th IUGG General Assembly (IUGG2023) (Berlin 2023)