Exploring the Use of Public Weather Station Data for Operational Weather Forecast Verification

Abstract

In recent years, the availability of crowd-sourced weather measurements has increased substantially. Yet, despite offering an insight into the weather where people live, these measurements are not currently being utilized by public weather services in the operational objective verification of forecasts. Here, we explore the use of crowd-sourced temperature observations from the Weather Observations Website (WOW) to verify and compare the performance of the Met Office's replacement post-processing system, known as IMPROVER, against the old system. It is found that, even after quality control, the WOW data still has up to five times the number of sites compared to the official surface network. The overall errors are marginally worse than using the official network; for example, the Mean Absolute Error is approximately 0.2 K larger for IMPROVER verified with WOW over SYNOP sites. However, 95% of the errors at all quality-controlled WOW sites are less than or equal to 2.5 K, and 70% of the errors are less than or equal to 1 K, indicating a good level of consistency with the forecasts. The sensitivity of the results to quality control depends on the choice of error metric. Finally, given the degree of consistency, quantity, and location of good-quality WOW data, it is recommended that crowd-sourced data continue to be used as an operational verification truth source in conjunction with the official surface network.