On correlation of maize and wheat yield with NDVI: Example of hungary (1985–1998) Original Research Article

The aim of the study is to quantify statistical relations between the Normalized Differential Vegetation Index (NDVI), derived from NOAA/AVHRR multi-channel irradiance, and yield of wheat and maize commonly covering 23.5% of Hungary. The 14 years period, 1985–1998 is used, reserving the later vegetation seasons for independent validation, in the future. The yield reporting units are the 19 administrative counties, characterized by 50–90 km of linear measure and diverse vegetation. Cluster analysis of the yield series is performed to identify possible outliers, but there are no qualitatively separating outliers found among the 14 years and 19 counties for the investigated plants. The same procedure is performed for the average proportion of land-use types and for the NDVI series with the aim of finding coherent groups of counties to unify them into larger, cumulative samples. However, these analyses did not yield the necessary similarities, hence, no further spatial integration was performed. The applied NDVI series are filtered against possible remained atmospheric disturbances, whereas the yield data are standardized against the, actually decreasing, linear trends. The relationships between weekly composite NDVI data are residual yield percentages are rather different for maize and wheat: wheat yield is closely related to the early spring NDVI, whereas for maize yield only the much later, near-harvest periods exhibit some informative value. Since the obtained correlations are rather similar in the neighboring weeks and the weekly NDVI series are strongly auto-correlated, themselves, a four-weekly integration of NDVI is performed before the final estimation of the yield from the NDVI. This integration is performed in four different ways, recommended by literature sources, with no real differences of the results, that promises stability of the correlation, despite the short samples. This statistical predictability of wheat yield residuals with about three months time lead, can be interpreted as follows: Mixed vegetation of the counties indicate the spring restart of wheat development, which conditions determine a substantial part of yield variability.