MRF MOS TEMPERATURE AND DEW POINT DEVELOPMENT
February 3, 2000
The predictands for the MRF-based MOS max/min guidance were values corresponding to a
local daytime maximum and nighttime minimum temperature. The daytime is defined as 7 AM
to 7 PM Local Standard Time (LST); nighttime from 7 PM to 8 AM LST. An algorithm is used
to calculate max/min from METAR observations. The predictands for the 6-h temperature and
dew point guidance were observations reported at specific times each day. Predictands are valid
at projections 18 to 192 hours from after model run time (00 UTC).
- •MRF model predictors: temperatures, thicknesses, dew points, mean relative humidity,
model precipitation, u- and v- wind components, wind speeds and vertical velocities on
isobaric surfaces; pressure, K-index, and lapse rates. The predictors were offered as values at
the specific time of the forecast and as 12-h time-averaged fields.
- •Geoclimatic predictors: sine and cosine of the day of the year and twice the day of the year.
- •Observed predictors: 06 UTC temperature and dew point at projections valid 18 to 36 hours
after 00 UTC.
- •Frequently chosen predictors: 1000-mb temperature, 2-m temperature, low level
thicknesses, 1000-mb dew point and the 1000-500-mb mean relative humidity. The 2-m
temperature was the favored predictor for the daytime max and the 18 UTC and 00 UTC
temperature and dew point. However, the temperature at a particular layer (usually 1000-mb)
was most common for the nighttime min and the 06 UTC and 12 UTC temperature and dew
point. At the extended projections, the climatological terms were more frequently used.
- •Cool season (October through March): 1992-97 MRF re-analysis forecast data (every 5th
day), 1997-98 and 1998-99 MRF cool season data.
- •Warm season (April through September): 1992-96 MRF re-analysis forecast data (every 5th
day), 1997, 1998 (except June 16 through July 21) and 1999 MRF warm season data.
- •For seasonal transition, a two week overlap at the beginning and end of each season was
- •Last 15 days of each month in the 1998-99 cool season and the 1999 warm season.
- •Test forecasts were evaluated in terms of mean algebraic and mean absolute error.
- •Daytime max and nighttime min were more accurate than the old MRF MOS out to 108
hours and comparable in skill thereafter (out to 192 hours).
- •Temperature and Dew point guidance improved significantly over direct model output and
climatology at all projections.
Additional results can be found on our web site at http://www.nws.noaa.gov/tdl/synop .