Gridded Model Output Statistics (MOS)
Coop Tests for Maximum and Minimum Temperature Forecasts
In response to several complaints from the field concerning discrepancies between the max/min MOS grids
and the MOS point values of max/min at corresponding locations, we have been investigating the effect of Cooperative
stations located near METAR sites.
Cooperative sites contribute a large amount of detail to gridded MOS Max/Min forecasts, especially in data-sparse
regions. However, the 24-hour reporting period used for coops can be difficult to match with the appropriate daytime
maximum and nighttime minimum forecasts at METAR and mesonet sites.
We propose removing coop sites within 10 km of METAR sites. This will allow us to make the best use of coop sites
where we need their added detail, but keep the grid forecasts nearer to the our daytime maximum and nighttime minimum
definition where possible. Click here for a comma-delimited table listing all coop sites being considered for removal.
During the process of selecting coops and testing, we identified 12 coop sites that our objective land/water assignment
method incorrectly identified as ocean stations (due to their location on small land features not well represented in our masks). We adjusted
the land/water designation of these coops in our tables to reflect that they are more representative of land observations.
Click here for a comma-delimited table listing the coop sites that were adjusted. These adjustments
are not reflected in the case study maps below, with the exception of the differences in the Atlantic Ocean off the southeast US coast,
where stations were both adjusted and removed from the MaxT/MinT analysis.
The map below displays all stations used in the GMOS MaxT/MinT analysis, divided by station type (please refer to
the legend for station symbology). Coops proposed for removal are represented by red "X"s.
MOS and GMOS verification were used to assess areas where the grids contained larger forecast error than stations
at the same point. The difference in GMOS and MOS mean MAE for July 2010 MinT is displayed as stars. Larger stars represent a greater
difference in MAE. MAE differences greater than 2 degrees Fahrenheit are colored red; differences greater than 5 degrees Fahrenheit are
colored dark red.
Case Study Images
Several case studies were prepared to evaluate the differences in the grids when coops located near METAR sites were removed from the analysis.
Forecast
| Valid Time
| Difference Map
|
20110224 00z GFS Daytime MaxT |
20110225 00z |
|
20110224 00z GFS Nighttime MinT |
20110225 12z |
|
20110227 00z GFS Daytime MaxT |
20110228 00z |
|
20110227 00z GFS Nighttime MinT |
20110228 12z |
|
20110227 00z GFS Daytime MaxT |
20110301 00z |
|
20110227 00z GFS Nighttime MinT |
20110301 12z |
|
20110227 12z GFS Nighttime MinT |
20110228 12z |
|
20110227 12z GFS Daytime MaxT |
20110301 00z |
|
20110227 12z GFS Nighttime MinT |
20110301 12z |
|
20110227 12z GFS Daytime MaxT |
20110302 00z |
|
5-Case Mean Daytime MaxT |
|
|
5-Case Mean Nighttime MinT |
|
|
|