Incorporating convective cell evolution into convective storm modelling

Climate change is intensifying short-duration, high-intensity rainfall events in many regions worldwide, highlighting the need to improve rainfall modelling for such storms to support effective stormwater management. A recurring challenge is that many existing rainfall modelling tools fail to account for the evolution of convective cells, potentially leading to under- or over-estimates of rainfall extremes and their hydrological impacts.

To address this challenge, this study presents a spatial-temporal rainfall generator that explicitly incorporates convective cell evolution. In this approach, storm arrivals are modelled by a point process, with storm cells represented by clusters of rainfall objects, of which each is characterised by specific intensity and geometric properties. Whereas most existing generators assume constant cell properties throughout a storm, the properties of our convective cells evolve with time –a more realistic representation of cell lifecycles exhibiting growth and decay. This design also naturally captures the birth of new cells and the dissipation of existing ones during storm events.

The parameters of the generator are derived from an analysis of 167 convective storm events observed in the Birmingham area (UK) between 2005 and 2017. These events were identified and tracked using the enhanced TITAN storm tracking algorithm (Munoz et al., 2018), which extracts convective cell paths and their key properties (e.g., rainfall intensity, spatial extent, storm and cell duration, and movement). The resulting dataset was then used to calibrate a copula-based convective cell lifecycle generator (Tseng et al., 2025), serving as the core mechanism for introducing cell evolution into the rainfall modelling framework.

Preliminary results suggest that our generator not only reproduces the observed standard statistics but also more effectively preserves rainfall extremes than existing generators that assume constant cell properties. In addition, by offering a more realistic representation of cell dynamics and improved spatial-temporal rainfall structures, our generator has the potential to yield more accurate hydrological responses.