Temperature

BCRMSE

    • A strong diurnal trend is seen for each seasonal aggregation, but becomes somewhat averaged out for the annual aggregation. The distribution of BCRMSE values range from 2.0 - 4.5 C among all temporal aggregations. An increase in the median value of BCRMSE with forecast lead time is also noted.
    • For the summer and spring aggregations, the largest errors occur for forecasts valid at 21 UTC and the smallest errors are seen for forecasts valid at 15 UTC. The fall and winter aggregations exhibit the largest error for forecasts valid around 9 UTC and the smallest at 00 UTC. This leads to a weaker diurnal signal in the annual aggregation, however, it trends more towards the summer and spring aggregations in terms of when the maximum and minimum errors occur.

Bias

    • A consistent statistically significant (SS) warm bias is noted for forecasts valid from 03 - 18 UTC in the annual aggregation. A distinct diurnal trend is seen, where the largest warm bias occurs with forecasts valid at 15 UTC. For forecasts valid at 21 UTC, the confidence intervals (CIs) encompass zero and, thus, are classified as unbiased forecasts. A similar trend is noted among a majority of forecast lead times regardless of seasonal aggregation examined.

Relative Humidity

BCRMSE

    • A strong diurnal cycle is noted in the distribution of both the summer and spring aggregations, however this trend is much smaller for the winter aggregation and nearly non-existent for the fall aggregation. A general increase of BCRMSE values with forecast lead time is noted, regardless of temporal aggregation, and the BCRMSE values range from 11 - 18 %.
    • For the spring, summer and annual temporal aggregations, the largest errors occur for forecasts valid around 00 UTC, while the smallest errors occur for forecasts valid around 12 UTC.

Bias

    • Typically, forecasts valid between 9 and 15 UTC have a SS low bias for all temporal aggregations except winter, which has a SS high bias for forecasts valid around 18 through 03 UTC. In general, the CIs for the other forecast valid times encompass zero.

Wind

BCRMSE

    • For all temporal aggregations, the largest errors are noted for forecasts valid at 21 UTC, while the smallest errors are typically seen for forecasts valid around 09 to 12 UTC. The distribution of BCRMSE values range from 3.0 - 5.0 m/s.

Bias

    • A consistent SS high bias is noted at all forecast lead times for both 00 and 12 UTC initializations, with values ranging from 0.6 - 2.5 m/s, depending on temporal aggregation.

3-hour Precipitation Accumulation

ETS

    • ETS values decrease with increasing threshold for both 00 and 12 UTC initializations, all forecast lead times and all temporal aggregations. The CIs are generally larger during the fall and winter aggregations when compared to the spring and summer aggregations. In addition, CIs for forecasts valid at 12 UTC are larger than those valid at 00 UTC.

Bias

    • In general, all of the temporal aggregations have a SS high bias at the lowest thresholds, which transitions towards a SS low bias at the highest thresholds for the summer and annual aggregations, especially at longer forecast lead times. The mid-range thresholds for all aggregations and highest thresholds for the fall, winter and spring aggregations typically encompass one, and thus are not significantly different from an unbiased forecast.

24-hour Precipitation Accumulation

ETS

    • Similar to that seen for the 3-hour accumulations, ETS values decrease with increasing threshold for both initializations and all forecast lead times. The winter aggregation tends to have SS higher ETS values for the mid-range thresholds compared to the summer aggregation due to a more gradual decrease with increasing threshold.

Bias

    • A SS high bias is noted at the 0.01 and 0.10" thresholds for both initializations and a majority of forecast lead times for all temporal aggregations. Beyond that, the CIs typically encompass one for all thresholds at each of the forecast lead times, with the exception of summer which tends to have a SS low bias at the mid-range thresholds.

Temperature

BCRMSE

    • As forecast lead time increases, BCRMSE values also increase, where errors at the 60 hour lead time are SS higher than those seen at the 12 hour lead time. The BCRMSE values range from 0.8 - 2.6 C.
    • The smallest errors are seen between 300 - 500 hPa, while the largest errors are observed between 200 - 100 hPa for a majority of temporal aggregations and lead times.

Bias

    • For the majority of temporal aggregations at the 12-hour lead time, a general SS cold bias is noted for levels at and below 700 hPa and a SS high bias noted above 400 hPa, with the exception of 150 hPa, which encompasses zero along with 500 hPa. The largest (warm) bias occurs at 100 hPa. As the lead time increases, more CIs encompass zero, especially for the summer aggregation.

Relative Humidity

BCRMSE

    • For all forecast lead times, BCRMSE values increase from 850 - 500 hPa, where values range from 8.5 - 32%, depending on the forecast lead time and temporal aggregation.

Bias

    • A SS moist bias is displayed at all levels for all forecast lead times during the spring and annual aggregations. The 850 hPa level encompasses zero for all lead times for the fall aggregation and for select lead times for the summer and winter aggregations.

Wind

BCRMSE

    • In general, the largest BCRMSE values occur between the 200 - 300 hPa levels for all forecast lead times and temporal aggregations, while the smallest errors generally occur at the top- (100 hPa) and bottom- (850 hPa) most levels. The values range from approximately 3.0 - 16 m/s.

Bias

    • When a SS bias is seen, it is generally low for all temporal aggregations, only a few exceptions apply (for the summer, winter and spring aggregations). The levels at which these SS biases occur depend on the forecast lead time.