PHENOLOGY AND YIELD ASSESSMENT IN MAIZE SEED CROPS USING SENTINEL 2 VIs’ TIME SERIES

Vegetation indexes (VIs) Time-Series (TS) data can be used to deliver in-season yield forecasts and to extract key phenological metrics. The present work addresses the TS analysis of different VIs, obtained from Sentinel 2 data for the 2015-17 growing seasons, for yield assessment and for the determination of the beginning of the reproductive stage (silking) of maize seed crops cultivated in the Po plain. Yields (expressed as green ears weight and grain weight) and silking dates (DoY-Day of Year) were collected for 17 maize seed crops (2 crops for the 2015 growing season, 11 for 2016, 4 for 2017) and correlated with different temporal parameters extracted from VIs’ field average TS by fitting different functions (Gaussian, Asymmetric Gaussian, Double Logistic, Double Sigmoid and Growth function). Extracted parameters included large integrals and small integrals between the fitted curve and the Greenness baseline as determined by the tangent at the first inflection point or by an arbitrary VI threshold. Results showed best linear correlations of yields with the small integrals of EVI (Enhanced Vegetation Index) field average TS, fitted by the Gaussian function and with an arbitrary EVI threshold in the 0.15-0.25 range. The date of maximum Green-up at the first inflection point of the Double Logistic fitting was the best silking date estimator, regardless of the VI used.