Localized Radar-Based Nowcasting of Convective Rainfall

Accurate short-range rainfall forecasts, known as nowcasts, are crucial for providing early warnings
of extreme precipitation and flooding, especially in urban areas with high population density.
Traditional nowcasting techniques, such as extrapolating radar echoes from constant altitude
plan position indicator (CAPPI) or lowest-angle plan position indicator (PPI), often struggle
to predict the development and dissipation of convective storms effectively. To address these
limitations, methods like RadVIL, which uses mass balance equations of Vertically Integrated
Liquid (VIL), and Spectral Prognosis (SPROG), which employs an autoregressive (AR) model,
have been developed. Recent advancements include Autoregressive Nowcasting using VIL
(ANVIL), which models the growth and decay of VIL using an autoregressive integrated (ARI)
process, decomposing VIL into multiple spatial scales and applying a separate ARI model to
each scale. Another advancement called SPROG-Localized (SPROG-LOC) extends the SPROG
approach, which is the deterministic version of Short-Term Ensemble Prediction System (STEPS),
by estimating spatially localized parameters of the AR process, improving the accuracy of rainfall
forecasts. Building on these recent methods, we propose a novel approach that combines ANVIL
and SPROG-LOC, termed SLANVIL (SPROG-Localized Autoregressive Nowcasting using VIL).
This integrated method leverages the strengths of both techniques, aiming to improve nowcasting
performance, particularly in scenarios with large, non-uniformly distributed precipitation areas
and isolated convective features. While we are currently in the process of obtaining preliminary
results and its probabilistic extension, we will present a first validation of our method for lead
times up to 2 hours, comparing its forecast skill at various precipitation thresholds and spatial
scales to established and operational nowcasting methods, such as pySTEPS-BE.