Hindcasting Changes in Upper-Ocean Thermal Structure on Short Timescales with Data from MILE and Ocean Weather Station Papa

Abstract

The skill of mixed-layer hindcasts of short (12?120 h) duration was investigated using data from Ocean Weather Station Papa for the years 1960?68, and from the Mixed Layer Experiment (MILE), which was conducted about 4O km southwest of Papa in late summer of 1977. The hindcasts were initialized and validated with observed temperature profiles and forced with surface wind stresses and heat fluxes calculated from meteorological observations. Mean and rms hindcast errors for sea surface temperature (SST) and mixed-layer depth (MLD) were compared with errors for persistence and climatology. Hindcast skill was calculated as the percent improvement of the hindcast rms error over the persistence rms error. The hindcast skill was significantly higher for the MILE data than for the Papa data. Hindcast skill with the Papa data was generally higher in spring and summer than in fall and winter. The range of hindcast skill for hindcasts of 12?36-h duration was 39%?48% for SST and 28%?37% for MLD for the MILE data versus 14%?19% for SST and 22%?33% for MLD for the Papa data for the spring and summer. Smoothing the MILE temperature observations with a 2-h running mean resulted in an increase in hindcast skill of about 3% due to the reduction in small-scale noise. The persistence rms MLD error was found to exceed the climatological error after about 2 days for both the MILE hindcasts and for the Papa spring and summer hindcasts. Hindcasts initialized from climatology in spring and summer showed skill for MLD similar to that for hindcasts initialized from an observed profile for hindcast durations longer than about 2 days.