POAMA-2 Skill compared with other international models

Just a few plots on a page that provide a sample of how POAMA-2 compares with other international models.

Accuracy score for above median rainfall

POAMA 2 ECMWF MF UKMO
POAMA 2 Hit Rate ECMWF Hit Rate MF Hit Rate UKMO Hit Rate
Colour bar

Accuracy score for above median rainfall forecast (fraction of hits and correct negatives compared to total forecasts) for P2.4, ECMWF (IFS), UKMO (HadGEM2), MF (ARPEGE/OPA) models. Lead time 1 month, forecast initialised one month before the start of the verification season. Cross validated skill measured for 1980-2005 hindcast period.

Anomaly correlation for above median rainfall

POAMA 2 ECMWF MF UKMO
POAMA 2 Anomaly correlation ECMWF Anomaly correlation MF Anomaly correlation UKMO Anomaly correlation
Colour bar

Anomaly correlation of ensemble mean rainfall forecast for P1.5, P2.4, ECMWF (IFS), UKMO (HadGEM2), MF (ARPEGE/OPA) models. Lead time 1 month, forecast initialised one month before the start of the verification season. 1980-2005 hindcast period.

Reliability error for above median rainfall

Reliability error

Reliability error (component of Brier Skill Score) for above median rainfall forecast for P1.5, P2.4, ECMWF (IFS), UKMO (HadGEM2), MF (ARPEGE/OPA) models. Lead time 1 month, forecast initialised one month before the start of the verification season. Cross validated skill measured for 1980-2005 hindcast period.

Attributes diagram for above median rainfall

POAMA 2 ECMWF MF UKMO
POAMA 2 Attributes ECMWF Attributes MF Attributes UKMO Attributes

Attributes diagram for above median rainfall forecast for P2.4, ECMWF (IFS), UKMO (HadGEM2), MF (ARPEGE/OPA) models. MAM, JJA, SON and DJF at lead time 1 month and 4 months combined. Lead time 1 month means the forecast was initialised one month before the start of each verification season. Cross validated skill measured for 1980-2006 hindcast period. The y-axis is the relative observed frequency, the x-axis is the forecast probability. The solid line shows perfect reliability. The shaded area indicates a skilful forecast. The size of the data point indicates the relative fraction of grid points occurring in each forecast probability bin.