• English 
  • Spanish 

Case study reproducibility in a convection-permitting WRF multi-physics ensemble: the role of internal variability

In the framework of the CORDEX FPS on Convective phenomena at high resolution over Europe and the
Mediterranean (FPS-CEM, Coppola et al., 2018), a convection-permitting multi-model ensemble was used to
simulate high-impact weather events over the Alps. This experiment resulted in noticeable discrepancies
between models in representing selected heavy precipitation events. The groups using the Weather and
Research Forecasting (WRF) model organized a multi-physics ensemble, suited to identify the processes
behind those discrepancies. In this work we analyze the uncertainty arising from internal variability in this
multi-physics ensemble at one-month and one-year timescales. To distinguish the uncertainty due to the use
of different parameterizations from that of the internal variability, a set of simulations with perturbed initial
conditions was performed. We measured quantitatively the uncertainty arising from both sources using
inter-member variances. For circulation variables, the results suggest that uncertainties from multi-physics
and internal variability have comparable magnitude, exhibiting an annual cycle with higher values in summer
than in winter. The spatial distribution of the uncertainties show similar patterns, with higher values over the
northeastern part of the domain. These patterns are in agreement with previous studies which conclude that
internal variability increases where the inflow of the boundary information is less dominant: that is, in
summer when the boundary forcing is not able to overcome the local-scale processes, and far from the
westerly flow coming from the north Atlantic. The behaviour of uncertainty also depends on the variable.
Surface variables are more affected by parameterized processes (soil physics, boundary layer, clouds, etc.),
hence the uncertainty associated to the parameterizations has more decisive role for these variables than for
circulation variables.