1/1
ECMWF YOPP dataset 2017-2020
Version 2 2020-10-13, 17:38
Version 1 2020-10-13, 17:28
figure
posted on 2020-10-13, 17:38 authored by Peter BauerPeter Bauer, irina sandu, linus magnusson, richard mladek, manuel fuentesFigure 1: Schematic of the YOPP dataset production framework: The initial conditions and ensemble forecasts are generated by ECMWF’s operational forecasting system based on ensembles of data assimilation and models. Atmosphere, land and ocean waves are separate from the ocean and sea-ice circulationin the data assimilation system, while the forecasts are fully coupled. Only the unperturbed control experiment output is ingested in the YOPP dataset from day 0 to day 15, and complemented by physical process tendency output from day 0 to day 2.
Figure 2: Mean error as a function of lead time for 700hPa temperature forecastsfor 2018 averaged over the Arctic. The ECMWF HRES forecast (red) andthe ENS control forecast used in the YOPP dataset (black). Verification is performed against the operational ECMWF analysis (included as time step 0).
Figure 3: Time series of model tendencies for temperature averaged over thefirst 24 forecast hours (top) and 24-hour temperature forecast errors (bottom),averaged between 850 hPa and 500 hPa. The top panel also includes the evolu-tion of the temperature over 24 hours. The bottom panel includes the forecasterror calculated from the 24-hour forecast output (black) and from the sum ofthe tendencies minus the evolution (grey dashed). A 30-day running mean isapplied to the time-series. All data is averaged inside a box between 75N-85N latitude and 110E-170E longitude.
Figure 4: 24-hour temperature forecast errors, averaged between 850 hPa and 500 hPa for the period 1 July to 1 October 2018. Darker colours (top rightcolour bar) indicate statistical significance at the 95% level.
