A Geostatistical Framework for Quantifying the Temporal Evolution and Predictability of Rainfall Fields

Abstract

Abstract A geostatistical framework for quantifying the temporal evolution and predictability of rainfall fields for time lags between 5 min and 3 h is proposed. The method is based on the computation of experimental space–time variogram maps of radar reflectivity fields. Two new metrics for quantifying temporal innovation and predictability based on minimum semivariance values at different time lags are proposed. The method is applied to high-resolution composite radar reflectivity maps over the United States to study the evolution of 25 convective and 25 stratiform events during the warm season of 2014. Results show that the temporal innovation can be modeled as the sum of two exponential functions of time lag, with approximately 50% of the total innovation occurring over the first 60 min. The median predictable scales for convective events are on the order of 1.6 km at 5 min, 5 km at 15 min, and 12.7 km at 1 h. Furthermore, the optimal time lag for predicting future innovation, taking into account measurement uncertainty and natural variability, appears to be between 30 and 60 min.