The summary plots in Fig. 6 show time series extracted from Fig. 5a
(combined region “All”) for a 4-week averaging window, identifying
windows of maximum index sensitivity for each crop on average over
the eight agricultural states targeted in this study and demonstrating
reasonable agreement in timing between indices at the national scale.
These correlations were computed by combining all state-level yieldindex sample pairs, decreasing the threshold of significant correlation
at p b 0.05 to approximately 0.2 depending on number of years with
yield data reported in the CONAB dataset. In general, soybean correlations peak between February and April, the first corn planting between
January and February, the second between May and July, and cotton between March and May. In each case, there is a statistically significant
correlation several weeks prior to the nominal harvest date. For soybean, cotton and the first corn season, peak correlations are higher for
ESI than for LAI anomalies, indicating that value is accrued by combining
LST and VI inputs to ALEXI. For these crops, ESI is also better correlated
with state-scale yield anomalies than is precipitation. Peak correlations
are more similar between all indices for the second corn crop, which
may relate to high variability in safrinha crop yields due to less reliable
climate conditions in the second season.