The current version of the GFS was implemented a little over a month ago.  Since then forecasters have observed several problems.

1) High 2 m dewpoint bias in the midwest.  The new GFS had a change in grassland and cropland properties to reduce a summertime warm dry bias in the Great Plains;  the fix aggravated a springtime wet bias in the northern Great Plains. 

2) Widespread light convective rain.  The GFS has long had a high bias in the lightest amounts,technically known as socialist rain.  That bias was reduced in January 2015 when the 13 km was implemented; at the same time the warm dry summertime bias in the Great Plains appeared.  In reducing the warm dry bias in the May 2016 implementation we increased the bias in the lightest amounts.

3) Rain maxima over the Great Lakes near the coastline at a time when the lakes are cold and have low cape values.  Elsewhere the GFS has shown a preference for rain just off shore-near Florida, the Yucatan penisula, Panama, the west coast of India.  But those tend to be regions where caper is higher over the ocean than the land.

4) High values of cape over the southern Great Plains, especially at 0Z, more so in the last week or so.  ECMWF appears occasionally to have a low bias in CAPE in some of the same cases.  The high bias in GFS sfc CAPE is exagerratied in AWIPS when the internal calculation is used rather than the value from the GFS.

The MEG has discussed these problems.  Thanks to TJ Turnage, Jeff Craven and others for bringing them to our attention.

Jongil Han has developed changes to the covnection in the GFS that will make it harder to trigger convection; it is being tested currently.  The changes may well address some of these problems.