Update 2012: Predicting yield from Landsat Reflectance Indexes: Results
The scatter obtained when relating Landsat LAI and current and past year yield was consistent with Lampinen et al. (2009, 2012) and showed that the yields for the orchards included in this study were close to the maximum potential yield for the range of satellite measured LAI (Fig. 1)
The Growing Degrees Hour was found closely related to other orchard tree phenological changes by Richardson et al. (1974). This formula was used to calculate the GDH as it accumulated from January 1st (Day of Year 1) to August 31 (DOY 243, most years). The spring warming is compared to increasing LAI in the first 140 days of the year (Fig. 2).
Fig. 2. The LAI is function (dependent) on the spring warming. For 2006 to 2011, we see a very high correlation between the satellite estimate LAI and the Growing Degree Hours. Displayed is the change in canopy in 2010 from mostly flowers DOY 55 (February 24), loss of flowering and leaf initiation DOY 63 (March 4th), early leaf out DOY 73 (March 14th), and moderate leaf out DOY 83 (March 24th). The canopy is approximately half of maximum LAI.
Although current year LAI is similar to previous year LAI, we concluded that it is not possible to predict current year LAI from previous year, though correlated. The previous year’s LAI accounts for only 44% of the current variability among the orchards (Fig. 3).
However, we can predict current year yield by knowing previous year's maximum LAI and rate of warming in the spring (Fig. 4. left panel). Knowing previous year satellite LAI estimates alone would account for 41% of the variability in yield, and estimate the yield within +/-550 lbs/ac (Fig. 4).
The best correlations were between previous year Landsat LAI and yield when the GDH accumulation between end of January (DOY 30) and the end of March (DOY 90), with the DOY, were included in the regression model. More than 50% of the variability among the orchard yield could be explained by the combination of these variables, and estimate the current yield within +/-500 lbs/ac (Fig. 4 left panel).
From this research we can conclude that it may be possible to predict yield early enough so the amount of fertilizer applied will match the harvest removal using early greenup rate and warming temperatures to yield. We do know that previous year yield is not a predictor, but previous year LAI may be a good predictor.