Temperature

BCRMSE

    • The general trend is an increase in BCRMSE with increasing lead time for both the 00Z and 12Z initializations. Diurnal signatures are also observed for all temporal aggregations.
    • With the exception of fall and winter, diurnal signatures reach a peak of 2.5-3.0 degrees C between 2100-0000 UTC, while minimum errors of 2.0-2.6 degrees C is noted between 1200-1500 UTC. Both the fall and winter aggregations have diurnal peaks at 1200 UTC. The errors for winter are generally higher than all other temporal aggregations with the distribution of the diurnal trend ranging between 2.5-3.5 degrees C.

Bias

    • A statistically significant cold bias is generally observed for all aggregations with a clear diurnal signature. The smallest bias is seen at 1500 UTC, while the largest cold bias is between 0000-0900 UTC.
    • The winter aggregation has the largest cold bias for both initializations between 1.0-2.0 degrees C while the other aggregations generally don't see a cold bias larger than 1.0 degrees C. During the summer a warm bias, generally between 0-0.5 degrees C, is seen from 1500-0000 UTC.

Dew Point Temperature

BCRMSE

    • For both the 00Z and 12Z initializations, the distribution of BCRMSE generally increases with an increase in lead time for all temporal aggregations. A diurnal trend is noticeable for all aggregations from 2100-0000 UTC.
    • Winter generally has the largest errors and as well as the largest increase in error with time. Both fall and winter have the smallest diurnal trends, while spring has the largest diurnal signature. The largest errors occur between 1500-0000 UTC and the smallest errors occur from 0300- 1500 UTC.

Bias

    • A strong diurnal signature is apparent for all aggregations and initializations. Regardless of initialization and aggregation, there is a peak in bias at the 21-00 UTC lead times and minimums between the 9-15 UTC lead hours.
    • A ss wet bias is generally seen between 18-6 UTC up to 1.5 degrees C and depending on aggregation and initialization. The aggregation with a higher and more frequent wet bias at both initialization times is spring and the highest dry bias of 0.9 degrees C is in winter, however, fall and summer see the most frequent dry biases.

Wind

BCRMSE

    • There is a general increase in BCRMSE with increased lead time for both the 00Z and 12Z initializations and all aggregations. For both initialization times, a weak or no diurnal signature is generally identified, with the exception of the annual and summer aggregations which have a strong diurnal signature peaking between 21-03 UTC and 09-12 UTC showing the lowest errors.
    • The winter and spring aggregations generally have high error through all lead times ranging from 1.9-2.4 m/s. The 12Z initilizaion generally has a lower error for most aggregations and lead hours than the 00Z initialization.

Bias

    • A strong diurnal signature is apparent for all aggregations and initializations. A ss significant fast wind speed bias is observered for all aggregations and initializations for all lead hours except for the 12Z spring and summer aggregations for the 06-12 and 06-09 lead hours respectivly, where the bias is neutral.
    • The peak wind speed biases always occur during the overnight hours from 18-00 UTC and range in magnitude from 1.0-1.3 m/s, depending on aggregation. On the other hand, the minimum bias occur during daylight hours from 18-00 UTC and range in magnitude from 0.0-0.7 m/s, depending on aggregation.

3-hour Precipitation Accumulation

GSS

    • In general there is a decrease in GSS towards zero with increasing thresholds. This is coincident with a decrease in the base rate, which is related to the number of observed events. A decrease in GSS is also noticed with increasing lead times. The highest GSS values are found during the winter (0.39 at the lowest threshold) while the summer exhibits the lowest values (0.2 at the lowest threshold).

Frequency Bias

    • In general for all aggregations except winter there is an over-estimate in the forecast for thresholds between 0.05 and 0.35 inches and an under-estimate for the lowest and highest thresholds, where winter is dominated by forecasts that are unbiased. Forecasts valid at 00Z tend to have higher biases than the forecasts valid at 12, with the exception of winter. For all aggregations and forecast hours, there is an increase in the spread of the CI with increased threshold which means a lower confidence in the bias value and appears to be correlted to the decrease in the base rate.

24-hour Precipitation Accumulation

GSS

    • Similar to the 03-hours accumulations, for all aggregations, there is a decrease in GSS with increasing threshold as well as a decrease in GSS with increasing lead times. The summer aggregation has the lowest base rate and GSS thatn all other aggregations with values going to near-zero.

Frequency Bias

    • In general, there is an over-estimate in the forecast for all aggregations and forecast hours, except thresholds of >0.01 and >3 inches where the results tend to be unbiased. For winter and fall there are more thresholds (>1) where there is no significant difference from the unbiased forecast at the 12Z valid hours. Summer is quite diferent from the other aggregations, with a continual increase in bias with increasing threshold(up to as much as 3 for the highest threshold) resulting in an over-forecast for all thresholds except >0.01 for the 36 and 48 hour forecast. Similar to the 3-hour accumulation, there is a larger spread of CI's with increasing threshold resulting in a lower confidence of the bias value which is likely correlated to the simultaneous decrease in the base rate.

Temperature

BCRMSE

    • For all aggregations and lead times, there is an increase in BCRMSE with increaased lead time. A pronounced minimum is found at the 500-400 hPa layer with magnitudes ranging from 0.45 degrees C in the initial forecasts to 1.4 degrees C in the 48-hour lead time. Maximum BCRMSE are observed near the surface and at 200-100 hPa, depending on the aggregation.
    • Winter generally exhibits the highest error for all lead times while summer generally has the lowest error. Above 200 hPa, there is a shift from BCRMSE increasing with increased height to BCRMSE decreasing with increased height, however, there is still increases with increased lead times at these levels.

Bias

    • All aggregations and lead times are similar, with the exception of the summer aggregation. In general ther is a ss cold or neutral bias from 850-400 hPa. Above that there is a shift to ss warm biases up to 200 hPa. A significant dip to ss cold biases is seen at the 150 hPa level, with a bounce back up to a ss warm bias at 100 hPa, with the exception of the initial lead time which generally remains cold at that level.
    • Unlike the rest of the aggregations, the summer exhibits ss warm biases from the near-surface upt to 500 hPa. The 400-300 hPa level above shifts to ss cold or neutral biases, with a shift back to ss warm biases aloft.There is still a decrease in bias at 150 hPa, however, it only becomes less warm rather than shifting to neutral or cold. The exception for the summer aggregation is the initial lead time which has the same profile as the other aggregations.

Dew Point Temperature

BCRMSE

    • For all aggregations and lead times, there is a general increase in BCRMSE with height from 850 to 500 hPa with errors decreasing with height above that. A noticable increase in BCRMSE is also observed with an increase in lead time.
    • The minimum erroe occurs near the surface at 850 hPa, ranging from 2.3 degrees C at the initail forecast hour to 7.5 degrees C at the 48-hour lead. The maximum, observed from 700-500 hPa ranges from 5.4 to 9.5 degrees C.

Bias

    • From the near surface at 850 hPa to 500 hPa, there is a general increase in bias with a decrease in pressure for all temporal aggregations and lead times.
    • At the 850 hPa level, errors are either neutral or small ss moist with values ranging from -0.4 to 0.8 degrees C while aloft at 500h higher moist biases are observed ranging from 0.7 to 3.1 degrees C. The fall aggregation often has lower biases and spring with the highest.

Wind

BCRMSE

    • For all aggregations and lead times, there is a general increase in BCRMSE with height up to a maximum located at 300-200 hPa, with the exception of the initial time which has a minimum at 500 hPa and maximum within 150-100 hPa.
    • Minimum errors range from around 2 to 4 m/s with maximums up to 6.5 m/s. An increase in error with lead times is noticable at most levels for all aggregations. Higher BCRMSE are generally seen in the winter and spring aggregations and fall and summer have lower values.

Bias

    • For the inital forecast hour, a ss slow wind speed bias is found below 200 hPa with a minumum bias occurring between 500-200 hPa. For all other aggregations, a transition from ss fast or neutral speeds at the surface to ss slow speeds at 200 hPA where the minimum occurs is observed. Above 200 hPa, there is generally another transition back to ss fast or neutral speeds.
    • Minimums in forecast error tend become more negative with increased forecast lead time and range in magnitude from -0.3 m/s at the initial hour down to -1.7 m/s for the 48-hour lead. Maximum bias values generally found at 100 hPa typically range between 0 to 1 m/s.