For state-level yield analyses, correlations were computed at 7-day
intervals between the 1-month composited index standardized anomalies
(Eq. (2)) and yield anomalies (Eq. (3)). To identify optimalwindows
during the growing season where an index is most predictive of atharvest
yield, a two-dimensional correlation space was computed for
each index, crop and region. The 1-month index composites were further
averaged over a variable window prior to correlation with yield
anomalies. In 2-D plots of these analyses, the x-axis represents the
end-date of the index averaging window, and the y-axis represents
the length of the window.