Downscaling of space-time rainfall using a Bernoulli-lognormal multiplicative framework

We present an analytical framework for the space-time downscaling based on Bernoulli-lognormal (BLN, traditionally known as beta-lognormal) multiplicative cascades. Considering recent results about the analytical parametrization of the BLN generator, we derive the explicit relation for obtaining fine-scale statistics directly from the coarse-resolution inputs while preserving the space-time dependence structures characteristic multi-scale extreme precipitation. The method is implemented in an automated workflow on Google Earth Engine, which enters precipitation data in real time and dynamically updates the multifractal parameters to generate high-resolution space-time synthetic fields. We evaluate the performance of the scheme by comparing the disaggregated fields with independent observations. The results indicate that the procedure provides a robust approach for the downscaling of precipitation in hydrometeorological applications and supports improved occurrence probability estimation and uncertainty quantification for extreme events.