Temperature
BCRMSE
• For both configurations, a
diurnal trend with a general increase in 2 m temperature BCRMSE is seen for all aggregations over the CONUS for both the 00 and 12 UTC initializations. A majority of PS pair-wise differences favor AFWAOC, typically at and around times valid 15 - 18 UTC. During the summer aggregation at times valid 03 - 09 UTC, NoahMP is a PS better performer than AFWAOC.
Bias
• When considering 2 m temperature bias, AFWAOC has a diurnal signal, regardless of initialization time or temporal aggregation; maximum errors occur at times valid between 21 - 03 UTC, with exact timing dependent on temporal aggregation. A minimum in errors (i.e., typically the smallest cold bias) is seen at times valid around 12 - 15 UTC. NoahMP, on the other hand, displays a bimodal distribution, with peak bias values valid at 18 and 00 UTC over the CONUS domain for the annual aggregation. All differences for 2 m temperature bias are PS, with a weak diurnal signal in which configuration is favored. For times valid between 18 - 21 UTC, if there are differences, NoahMP is the better performer; for times valid between 12 - 15 UTC, if there are differences, AFWAOC is the favored configuration.
Dew Point Temperature
BCRMSE
• Similar to 2 m temperature, a diurnal signal superimposed within a gentle increase in errors with time is present in 2 m dew point temperature BCRMSE for all temporal aggregations, both initializations, and for both configurations. A slightly amplified diurnal signal with larger errors around times valid 18 - 00 UTC is noted for NoahMP, except in the winter where median values of BCRMSE are consistently higher for NoahMP than AFWAOC. Pair-wise differences favoring AFWAOC are seen for valid times between 15 - 03 UTC. Any PS pair-wise differences favoring NoahMP are seen in the summer and spring temporal aggregations, and generally between times valid 09 - 12 UTC.
Bias
• The two configurations have the largest divergence in 2 m dew point temperature bias values in the overnight into morning hours (i.e., valid 03 - 12 UTC); this pattern is observed in all but the winter aggregation. Typically at these valid times NoahMP has higher magnitude values than AFWAOC with the favored configuration dependent on temporal aggregation. When pair-wise differences are noted, all are PS. Most PS differences favor AFWAOC; however, there are PS pair-wise differences that favor NoahMP at some forecast lead times, depending on initialization and temporal aggregation.
Wind
BCRMSE
• For 10 m wind speed BCRMSE, both configurations display a weak diurnal signal with a general increase with forecast lead time for all temporal aggregations and both initialization times. The errors are largest at times valid between 21 - 00 UTC, while the smallest errors are seen at times valid near 12 UTC. No PS pair-wise differences are seen
Bias
• A prominent diurnal signal in bias is seen for all temporal aggregations and both initializations for 10 m wind speed bias, with highest errors seen at times valid 03 - 12 UTC and with lowest errors at times valid 15 - 00 UTC. With only a few exceptions in the afternoon and early evening hours, a high wind speed bias is observed regardless of initialization and temporal aggregation; this high bias generally grows with forecast lead time. While a number of SS pair-wise differences are observed, none are PS.
3-hour Precipitation Accumulation
GSS
• Regardless of configuration, initialization, or forecast lead time, median GSS values decrease as the threshold increases from 0.01" to 1.00". Several SS pair-wise differences are noted but are dependent on initialization, forecast lead time, and precipitation threshold, with most favoring NoahMP.
Frequency Bias
• In general, a high bias for valid times of 00 UTC is present in all but the highest thresholds, while the forecasts are unbiased for valid times at 12 UTC during the spring and summer. No SS pair-wise differences are noted, regardless of initialization hour, forecast lead time, or threshold.
24-hour Precipitation Accumulation
GSS
• In general, for both configurations, initializations, and forecast lead times, GSS decreases as threshold increases. Scattered pair-wise differences are observed with the favored configuration dpendant on temporal aggregation, lead time, and threshold; no consistent pattern is noted.
Frequency Bias
• Generally, regardless of configuration, initialization hour, temporal aggregation, or forecast lead hour, a high bias is present at most thresholds, with exception to the lowest and highest thresholds. No differences are noted for either initialization for any forecast hours or thresholds.
Temperature
BCRMSE
• For both configurations, regardless of temporal aggregation or forecast lead time, a minimum in BCRMSE values is found between 500 and 300 hPa, with the largest error occurring at the lower and upper-levels. In general, the largest differences between the two configurations are seen at 850 and 700 hPa. This is reflected in the SS/PS pair-wise differences, with a majority of the differences occurring at and below 700 hPa and favoring Noah-MP.
Bias
• While the AFWAOC has a cold bias at the lowest level that transitions to a warm bias above for the annual aggregation, NoahMP displays a warm bias at all levels. A cold bias at the lowest levels is noted for both configurations for other temporal aggreagtions, especially fall and winter. At 850 hPa, the favored configuration depends on the temporal aggregation; AFWAOC is favored in the summer and spring, while NoahMP is favored in the fall and winter. All of the noted differences at that lowest level are PS. While consistent SS pair-wise differences favoring NoahMP are noted for the annual, summer and spring aggregations at mid levels (500 - 300 hPa), the fall and winter aggregations show very few SS pair-wise differences at those levels.
Dew Point Temperature
BCRMSE
• Both configurations display an increase in dew point BCRMSE as forecast lead time increases and pressure decreases in all temporal aggregations, except the winter aggregation, where at most forecast lead times, a slight decrease in median BCRMSE from 700 hPa to 500 hPa is observed. A few SS/PS pair-wise differences are noted, all of which favor NoahMP; however, the details depend greatly on temporal aggregation and forecast lead time.
Bias
• For most temporal aggregations, 850 hPa displays a wet bias at the 12-h forecast lead time, with dew point temperature bias transitioning to a dry bias as forecast lead time increases; at 700 and 500 hPa, for all forecast lead times, a wet bias is present. Most SS/PS pair-wise differences favor for the winter and spring aggregations, while it is level and lead time dependant for the annual and summer aggregations.
Wind
BCRMSE
• For both configurations, BCRMSE generally increases from a minimum at 850 hPa to a maximum around 300 - 200 hPa, with decreasing error further aloft. As forecast lead time increases, the errors also show a tendency to increase. Overall, both configurations have very similar distributions of BCRMSE; this is reflected when examining the statistical significance, where only a few scattered SS pair-wise differences are noted. No differences were considered PS.
Bias
• At 850 hPa, both configurations, for all temporal aggregations, generally have a negative bias (i.e., winds are too light) that transitions towards a high bias with forecast lead time. A negative bias is noted at nearly all forecast lead times and seasonal aggregations at 200 hPa; at levels above 200 hPa, forecasts become either unbiased or display a positive bias. A few SS pair-wise differences are seen, depending on forecast lead time and vertical level; none are PS.
