Bias and Error Graphed Against Time of Day

Bias and Error Graphed Against Time of Day

This page has graphs similar to those on other pages but uses time of day as the horizontal axis. Some meteorological conditions occur on a diurnal cycle so an analysis of error and bias as they occur throughout the day and for different times of the year could lead to insights about the weaknesses in a Planetary Boundary Layer (PBL) scheme, sometimes called an Atmospheric Boundary Layer (ABL) scheme. My research has not yet lead me to making changes to the PBL schemes in WRF, but I thought it worthwhile to show data that might help me or someone else identify needed changes for future research.

This location is in southern Idaho in U.S.A., so the time offset is UTC - 7. In the U.S.A it is common to describe time based on Greenwhich Mean Time (GMT) which is the same as Coordinated Universal Time (UTC).

This graph shows the wind speeds as measured by a sodar during different times of day and for different times of the year. It shows a general diurnal cycle with some variation for different times of the year. The time of year variations peak and trough at different times of day, probably because of changing sunrises and sunsets causing changes in the times of highs and lows of the day.

This graph gives an indication of how many data points went into each average for each time of day. The data comes from the same 525 WRF runs mentioned elsewhere, but each run only provides 7 hours of coverage, so each time during the day ends up with about 140 data points. And, as also mentioned elsewhere, times are ignored when usable sodar readings were not available.

This graph shows how bias varies with time of day. Both RUC and WRF biases vary diurnally but with a slightly different pattern. And as with other pages in this section, RUC has a signicantly higher bias, which means it predicts winds to be significantly slower than what WRF predicts and what the sodar measures.

This graph shows the mean of the total error in the WRF and RUC predictions. Some of the error seems to follow the pattern of wind speed and some of the error follows the pattern of bias. But some of the error seems independent of those, so meteorological conditions other than raw speed seem to be affecting the PBL scheme's handling of wind predictions.

And if I remove the bias from two graphs up from the error in the above graph, we get the best possible bias correction we could hope for using bias based on time of day. And as with other pages in this section, such a correction helps WRF a little and RUC a lot.