How key features of early development shape deep convective systems.

Deep Convective Systems (DCSs) reaching scales of 100-1000 km play a pivotal role as the primary precipitation source in the tropics. Those systems can have large cloud shields, and thus not only affect severe precipitation patterns but also play a crucial part in modulating the tropical radiation budget. Understanding the complex factors that control how these systems grow and how they will behave in a warming climate remain fundamental challenges. Research efforts have been directed, on one hand, towards understanding the environmental control on these systems, and on the other hand, towards exploring the internal potential of systems to develop and self-aggregate in idealized simulations. However, we still lack understanding on the relative role of the environment and internal feedbacks on DCS mature size and why. The novel high-resolution global SAM simulation from the DYAMOND project, combined with the TOOCAN Lagrangian tracking of DCSs and machine learning tools, offers an unprecedented opportunity to explore this question. We find that a system's growth rate during the first 2 h of development predicts its final size with a Pearson correlation coefficient of 0.65. Beyond this period, growth rate emerges as the strongest predictor. However, in the early stages, additional factors-such as ice water path heterogeneity, migration distance, interactions with neighboring systems, and deep shear-play a more significant role. Our study quantitatively assesses the relative influence of internal versus external factors on the mature cloud shield size. Our results show that system-intrinsic properties exert a stronger influence than environmental conditions, suggesting that the initial environment does not strictly constrain final system size, particularly for larger systems where internal dynamics dominate.