Temperature

BCRMSE

    • For both the 00Z and 12Z initializations, the BCRMSE generally increases with forecast lead time. For most temporal aggregations, a diurnal signature is observed with the distributions ranging from around 2-3 C.
    • For the 00Z initiation, the diurnal signature peaks at 21 UTC with a minimum generally observed only at 15 UTC for all aggregations except fall and winter. While the fall aggregation is generally low with no diurnal trend, the winter aggregation is continually high with pronounced peaks at 12 UTC. For the 12Z initialization, diurnal peaks for all aggregations occur between 21-00 UTC and minimums occur between 6-15 UTC with the exception of winter which remains high. Both winter and spring have the highest error.

Bias

    • A diurnal trend is present for all aggregations, however, the maximums and minimums for both the winter and spring aggregations appear to be at opposite times as the other aggregations for both the 00Z and 12Z initializations. Either a ss neutral (CI's encompassing 0) or a ss cold bias is generally observed for both initializations and aggregations.
    • Summer sees a ss warm bias from 9-12 UTC for both the 00Z and 12Z initializations ranging up to 0.6 degrees C. Statistically significant cold biases are observed for all lead times in winter and nearly all lead times for fall with magnitudes reaching to 1.0-1.4 degrees C.

Dew Point Temperature

BCRMSE

    • For all temporal aggregations, the BCRMSE increases with increased lead time for both the 00Z and 12Z initializations. Diurnal signatures are observed for all seasons with the distribution of the values ranging from 2-4 C, depending on the season.
    • The winter and spring aggregations see the largest forecast error for all lead times, with an error generally above 3 C. The diurnal maximums are observed between 21-00 UTC, while minimums are generally seen between 3-15 UTC. The exception is winter, which does not generally exhibit clear diurnal minimums and instead remains higher, especially for the 12Z initialization.

Bias

    • Diurnal signatures are present for both initializations with varying magnitudes depending on the temporal aggregation. A general moist bias is observed for all aggregations with the exception of fall and winter for the latter half of the valid forecast hours. Diurnal maximums exist between 21-00 UTC with magnitudes ranging from 0.3 to 1.5 degrees C and minimums within 12-15 UTC ranging from -0.8 to 1.0 degrees C.
    • For most of the temporal aggregations a decrease in the wet bias or change to dry bias is noted with an increase in forecast lead time, especially the fall and winter aggregations. For both initializations, the summer aggregation shows a ss wet bias for all forecast hours, while fall and winter exhibit ss neutral (CI's encompassing 0) or dry biases for a majority of the lead times with the exception being the beginning lead times for the 12Z initializations.

Wind

BCRMSE

    • In general, the BCRMSE generally increase with forecast lead time for both the 00Z and 12Z initializations. A diurnal trend is observed for most aggregations ranging from 2.5-4.5 degrees C.
    • For both initializations, the winter aggregation is the only one not exhibiting a diurnal variation with a high forecast error throughout the forecast lead times. The other aggregations show a diurnal peak at 21-00 UTC and minimum at 9-12 UTC

Bias

    • For both the 00Z and 12Z initializations and for all aggregations there is a very strong diurnal trend as well as a ss high wind speed bias for all lead times.
    • The diurnal maximums occur during the nighttime hours from 9-15 UTC and range from 0.7-1.4 m/s and the diurnal minimums occur during the day from 21-00 UTC and range between 0.2-0.7 m/s. Fall and winter have the smallest diurnal range while spring and summer have the largest diurnal range in error.

3-hour Precipitation Accumulation

GSS

    • For all temporal aggregations and times, there is a decrease in GSS with increasing threshold, so that the highest GSS values are found at the lowest threshold of >0.01 and the lowest GSS values are found at the highest threshold of >1. A simultaneous decrease in the base rate, which is the number of observed occurances, is observed. The values of the base rate are generally higher for the 00Z valid times than the 12Z valid times, however the GSS values do not share this trend. GSS values for all aggregations and initializations also generally decrease with increaased lead time. The winter season sees the highest GSS values while summer has the lowest GSS values.

Frequency Bias

    • For the annual, fall and summer aggregations at the 00Z valid times, there is a general high bias for all but the highest thresholds, which are commonly unbiased. For the 12Z valid times for the same aggregations, the thresholds below 0.1 are also not significantly different from an unbiased forecast. The spring aggregation exhibits unbiased values mainly for the first two and last two thresholds for all valid times with over-estimations present at the other thresholds. On the other hand, the exhibits higher biases for most valid times up to the >0.35 threhold, and unbiased greater than the > 0.35 threshold. For all aggregations, the 00Z initialization and 12-hour forecast does not follow the trends of the other times and instead is generally not significantly different from an unbiased forecast. A noticable trend is the increase in spread of the CI's with increasing threshold which means a lower confidence in the bias values and also appears to correlate with the decrease in the base rate.

24-hour Precipitation Accumulation

GSS

    • For all aggregations, there is a decrease in GSS values with an increase in threshold. A small decrease in GSS as well as base rate is also noticed as lead times increase. Summer has the lowest GSS and base rate values compared to the other aggregations, similar to the 03-hour accumulation.

Frequency Bias

    • A general over-estimate of the forecast is observed for most of the thresholds and aggregations. The main exception is winter, which over-estimates at the lower thresholds with the bias decreasing at the higher thresholds where there is no longer a significant difference from the unbias forecast. For most of the spring and summer times, the bias increases with increasing threshold (up to nearly 4 for 36-hour summer), whereas the other aggregations generally see a decrease in bias for the highest threshold.Similar to the 3-hour accumulation, the CIspread increases with increasing threshold indicating a reduced confidence in the bias value which also may relate to the simultaneous decrease in the base rate.

Temperature

BCRMSE

    • A minimum in BCRMSE is present at the 500-300 hPa levels for all aggregations and forecast hours with values generally increasing going towards the highest and lowest pressure levels. Errors increase with increasing forecast times.
    • At the initial forecast hour the minimum is near 0.5 degrees C and the maximum ranging around 1.2 degrees C depending on aggregation. For the 48-hour forecast time the minimum has increased to the 1.0-1.4 range and the maximum from 1.6-2.2 degrees C depending on temporal aggregation. Winter generally has the highest error at all levels and forecast lead times, while summer tends to have lower errors than the rest of the aggregations.

Bias

    • For all aggregations except summer, there is a ss cold or neutral bias at the lower 850-400 hPa levels for all lead times, with the exception of spring which shows a ss warm bias at 850 hPa for the 36 and 48 hour lead times. Above 400 hPa, ss warm biases are more frequent. However, there is sudden change to ss cold, ss neutral or less warm bias at 150 hPa and at 100 hPa, the bias varies depending on lead time and aggregation.
    • During the summer, a ss warm bias is seen at 700-500 hPa, a ss cold bias at 400-300 hPa and a ss warm or neutral bias aloft and near surface for all lead times except for the initial forecast time. The initial lead time shows similar trends to the other aggregations.

Dew Point Temperature

BCRMSE

    • For all forecast lead times and temporal aggregations, the lowest errors are near the surface and the highest errors are found at 700-500 hPa. The initial forecast lead time has the lowest errors, increasing with increase forecast lead time.
    • For the initial forecast hour, the minimum ranges from 2.8-3.7 degrees C near the surface with maximums between 5.5-6.5 degrees C aloft. Summer generally has the lowest error at low levels (850-700 hPa) and the highest error at higher levels (500-400 hPa), while winter is the exact opposite.

Bias

    • In general, an increase in bias with decreasing pressure is observed from 850-500 hPa. A ss moist bias is most common for all temporal aggregations and lead hours. The main exception is during winter, where a ss neutral bias is generally observed near the surface, with ss moist biases aloft. For most aggregations, there is an increase in error with increased lead time.
    • Of all aggregations, spring exhibits the highest bias for all lead times.

Wind

BCRMSE

    • In general, the error generally increases with increasing height and lead time. The exceptions are decreasing errors from 850-500 hPa for the initial forecast lead time and decreasing errors above 300-100 hPa, depending on forecast lead time.
    • The minimum errors occur at the initial forecast lead time ranging from 1.7-2.0 m/s at 500 hPa with the maximum errors of 5.4-6.2 m/s at the 300-200 hPa level for the 48 hour forecast time.

Bias

    • In general, there is a ss slow or neutral wind speed bias below 150 hPa with a general decrease from 850-200 hPa for all lead times and temporal aggregations. A sudden increase in bias from ss slow speeds to neutral or fast speeds takes place from 200 to 100 hPa. An increased in biase with increasing lead time is also observed.
    • The minimum slow wind bias is found at 200 hPa with magnitudes ranging from -1.2 to -0.3 m/s depending on aggregation and lead. The maximum fast wind biases are found both near the surface and top pressure levels ranging from -0.3 m/s (12Z winter) and 0.4 m/s (12Z fall). Winter generally sees the slowest wind bias at mid-levels 500-200 hPa.