Validation of ECMWF Ensemble Forecasts of Sea Surface Latent and Sensible Heat Fluxes in the Pacific Arctic Against Saildrone Observations

Abstract

Estimates of sea‐surface latent and sensible heat fluxes using in situ observations are very rare in the Arctic Ocean. Saildrone Explorer uncrewed surface vehicles (saildrones) were deployed in the Bering, Chukchi, and Beaufort Seas from May–October 2019. Sea‐surface fluxes are estimated using surface state variables (temperature, humidity, and wind) observed by the saildrones applied to a bulk algorithm. In this study, the observationally based estimates of sea‐surface fluxes and the surface state variables are compared with those in ECMWF ensemble forecasts. Errors in the forecasts are mostly random and within the observed standard deviation. There are, however, sporadic and large error spikes (>3 standard deviations). Sea‐surface temperature and surface air temperature and humidity are systematically underestimated in the forecasts. From the perspective of bulk flux calculations, errors in latent heat fluxes are mainly due to errors in air‐sea differences in humidity, and errors in sensible heat fluxes are mainly due to those in air‐sea differences in temperature. Contributions from errors in wind speed are secondary. Differences in the flux algorithms used in the forecast and observations contribute only slightly. This study illustrates that even limited in situ observations from uncrewed mobile platforms may yield useful information on the accuracy of ensemble forecasts that otherwise is unavailable in regions without other in situ observations. Issues related to statistical treatment of data from moving sources are addressed through a simple application of an ergodicity test to the saildrone data.