Accurate sub-seasonal to seasonal (S2S) weather forecast between 2 weeks and several months is crucial to making informed decisions regarding changes in the risk of extreme weather events, resource management, agriculture, forestry, public health, energy, etc. However, significant gaps exist between the needs of society and what forecasters can produce, especially over longer lead times. Using three goodness-of-fit metrics, this study examined the ability of the SOMs-generated CFSv2 to forecast the correct (observed) circulation pattern, as opposed to the actual observed gridded field over a 90-day forecast period. Mean sea-level pressure (MSLP), near-surface wind (wnd10m), 850-mb temperature (t850), and 700-mb geopotential heights (z700) from the North American Regional Reanalysis were used to categorize the synoptic-scale circulation for three regions (East, West, and Gulf) across North America from January 1979 – December 2016. Expectedly, forecast skill generally decreased from the first day down to the skill of climatology (after 10 -15 days) and also varied regionally, seasonally, and between variables. The forecasts for the winter and summer seasons outperformed others, while t850 and z700 forecasts outperformed wnd10m, except in the west region. More importantly, this study found that the SOMs-generated CFSv2 forecasts improve upon the skill of the raw CFSv2 forecast near the one 1 – 2 weeks lead time. This study thus demonstrates the potential utility of a SOMs-based forecasting method in medium-range weather forecasts.