Adjusting thresholds of satellite‐based convective initiation interest fields based on the cloud environment

Abstract

The Time‐Space Exchangeability (TSE) concept states that similar characteristics of a given property are closely related statistically for objects or features within close proximity. In this exercise, the objects considered are growing cumulus clouds, and the data sets to be considered in a statistical sense are geostationary satellite infrared (IR) fields that help describe cloud growth rates, cloud top heights, and whether cloud tops contain significant amounts of frozen hydrometeors. In this exercise, the TSE concept is applied to alter otherwise static thresholds of IR fields of interest used within a satellite‐based convective initiation (CI) nowcasting algorithm. The convective environment in which the clouds develop dictate growth rate and precipitation processes, and cumuli growing within similar mesoscale environments should have similar growth characteristics. Using environmental information provided by regional statistics of the interest fields, the thresholds are examined for adjustment toward improving the accuracy of 0–1 h CI nowcasts. Growing cumulus clouds are observed within a CI algorithm through IR fields for many 1000 s of cumulus cloud objects, from which statistics are generated on mesoscales. Initial results show a reduction in the number of false alarms of ~50%, yet at the cost of eliminating approximately ~20% of the correct CI forecasts. For comparison, static thresholds (i.e., with the same threshold values applied across the entire satellite domain) within the CI algorithm often produce a relatively high probability of detection, with false alarms being a significant problem. In addition to increased algorithm performance, a benefit of using a method like TSE is that a variety of unknown variables that influence cumulus cloud growth can be accounted for without need for explicit near‐cloud observations that can be difficult to obtain.