Temperature

BCRMSE

    • For both the 00 and 12 UTC initializations, a diurnal trend is displayed in the distribution of BCRMSE values, which generally range between 2.2 – 4.0 °C, depending on the temporal aggregation.
    • For the annual, summer and spring aggregations, the largest errors occur for forecasts valid at 00 UTC, while the smallest forecast errors are seen for forecasts valid at 06 and 09 UTC. For the fall and winter aggregations, the largest errors are seen for forecasts valid at 15 UTC, while the smallest errors are valid at 03 UTC.
    • An increase in BCRMSE is noted with forecast lead time. In many cases, for the annual aggregation, this increase is statistically significant (SS) for forecasts valid at the same time, 24 hours apart (e.g., the 6-hour versus the 30-hour lead time BCRMSE values from the aggregated 12 UTC initializations).

Bias

    • In general, a SS cold bias is seen for a majority of forecast lead times with only a few exceptions; forecasts valid at 09 and 12 UTC during the annual and fall aggregations encompass zero and forecasts valid between 06 and 12 UTC for the winter aggregation have a SS warm bias.
    • The amplitude of the bias does not generally increase with forecast lead time, for any temporal aggregation.

Dew Point Temperature

BCRMSE

    • All temporal aggregations reveal an increase in BCRMSE values with lead time. A diurnal trend is noted in all but the fall aggregation.
    • For both the 00 and 12 UTC initializations, BCRMSE values generally range from 2.4 – 4.5 °C. For the winter aggregation, the largest errors are associated with forecasts valid at 12 UTC, while for all other temporal aggregations the largest error is seen at 21 UTC. The smallest errors are noted for forecasts valid at 03 UTC for the winter aggregation, 09 UTC for the annual and spring aggregations and 15 UTC for the fall and summer aggregations.

Bias

    • Generally, there is a SS high bias for forecasts valid between 18 and 00 UTC for both the annual and spring aggregations initialized at both 00 and 12 UTC. A SS high bias is also noted between 15 – 00 UTC valid times initialized at 00 UTC and 21 UTC valid times for forecasts initialized at 12 UTC for the summer aggregation. The winter aggregation has a SS high bias for nearly all forecasts, regardless of initialization time.
    • A SS low bias is noted for forecasts valid between 03 – 12 UTC for the summer aggregations and for forecasts valid at 15 UTC for the fall aggregation.
    • In most cases, the confidence intervals (CIs) for forecast valid times not mentioned above encompass zero and, thus, are not significantly different from an unbiased forecast.

Wind

BCRMSE

    • For both initializations, the BCRMSE values range from approximately 1.9 – 2.5 m/s, with a distinct diurnal trend noted for all temporal aggregations except fall and winter, and increasing error with lead time.
    • For all temporal aggregations, the largest BCRMSE values are seen at 00 UTC, regardless of initialization time, while the smallest errors are around the 09 – 12 UTC valid times for the 00 UTC initializations and the 15 UTC valid time for the 12 UTC initializations.

Bias

    • A SS high wind speed bias is noted for nearly all forecast lead times, regardless of initialization hour or temporal aggregation.
    • For both the 00 and 12 UTC initializations, a very large diurnal cycle is noted with SS larger high bias during the overnight hours compared to the daytime hours.

3-hour Precipitation Accumulation

GSS

    • Regardless of initialization time or forecast lead time examined, the highest GSS values are seen at the lowest precipitation threshold and steadily decrease to near-zero for higher thresholds. The number of observed events (base rate) by threshold has a similar trend. The largest CIs are noted for the winter aggregation, which also has the lowest base rate.

Frequency Bias

    • For both the 00 and 12 UTC initializations, a SS high bias is generally noted for all thresholds below 0.25”, regardless of forecast lead time or temporal aggregation. For the annual and summer aggregations, above the 0.25” threshold, a transition towards a SS low bias is noted for the highest thresholds. For the other seasonal aggregations, the higher thresholds generally encompass one.

24-hour Precipitation Accumulation

GSS

    • A general decrease in GSS with increasing threshold and lead time, similar to that seen for 3-hour precipitation accumulations, is noted for all temporal aggregations except winter. Again, a similar trend is noted in the base rate, with very few observed events of 3” accumulations in 24 hours.
    • For the winter aggregation, the median GSS value for the 24- and 36-hour lead times remains higher for higher threshold values, however the CIs associated with those values are very large, indicating a low confidence.

Frequency Bias

    • While the CIs for the fall and winter aggregations frequently encompass one for many of the thresholds at the different initialization hours and lead times, the other temporal aggregations exhibit a fairly consistent high bias for all but the highest thresholds, which encompass one.
    • In general, as the threshold increases, the width of the CIs increase, indicating more spread and less confidence in the median bias value, likely related to the decreasing base rate.

Temperature

BCRMSE

    • Regardless of forecast lead time or temporal aggregation, a minimum in the BCRMSE values occurs in the mid-levels (around 400 and 500 hPa) and increases to a maximum at both the near-surface and upper levels.
    • A SS increase in BCRMSE values between the 12-hour and 48-hour forecast lead times is noted. Errors are less than 2.2 °C for all lead times.

Bias

    • For all temporal aggregations, a SS cold bias is seen for all lead times at 850 and 700 hPa transitioning to a SS warm bias at and above 500 hPa levels. The largest magnitude bias is generally at the 200 or 100 hPa levels.

Dew Point Temperature

BCRMSE

    • As the pressure level decreases from 850 to 500 hPa and the forecast lead time increases from 12 hours to 48 hours, the BCRMSE values increase for all temporal aggregations except winter, where the median value decreases from the 700 to 500 hPa level.

Bias

    • A SS moist bias is noted for all forecast levels and lead times for all temporal aggregations except summer, which encompasses zero for all lead times at 850 hPa and has a SS cold bias at 700 hPa for the 24-, 36- and 48-hour lead times.
    • Similar to the results for BCRMSE with pressure level and lead time, the highest bias is seen at the 500 hPa level 48-hour forecast lead time.

Wind

BCRMSE

    • A general increase from the minimum BCRMSE values at the lowest levels (typically at the 850 or 700 hPa) to a maximum at the 300 or 200 hPa level for all forecast lead times and temporal aggregations, is noted.
    • Overall, the errors generally range between 2 and 6.5 m/s, with a SS increase in BCRMSE values between the 12- and 48-hour lead times seen for several levels.

Bias

    • In general, there is a SS low wind speed bias for the 12-hour lead time at all levels above 850 hPa, regardless of the temporal aggregation. As the forecast lead time increases, bias values decrease and encompass zero at several levels for the spring, fall and winter aggregations only.
    • The largest (low wind speed) bias is observed at 200 hPa for all temporal aggregations.