Estimation of wheat production by the integration of MODIS and ground data

This paper presents a methodology capable of combining Moderate Resolution
Imaging Spectroradiometer (MODIS) Normalized Difference Vegetation Index
(NDVI) imagery and ancillary data to estimate durum wheat production in
Tuscany (Central Italy). First, the phenological stages of winter wheat are simulated
by the use of an agro-meteorological model (Syrius 4.1). Next, MODIS
NDVI images at 250 m spatial resolution are utilized to identify fields likely
grown with winter wheat. The multitemporal NDVI profiles of these fields are
then integrated with Syrius 4.1 outputs through a previously developed procedure
in order to simulate wheat above-ground biomass and harvest index. This allows
the computation of wheat yield, which, combined with relevant cropped area,
provides provincial wheat production estimates. The methodology is tested using
ground and MODIS data taken over four Tuscany provinces where winter wheat is
widely cultivated. The accuracy of all estimated variables (wheat cropped area,
yield and production) is finally evaluated against provincial statistical data. The
results of this experiment indicate that the accuracy of wheat cropped area estimation
and yield simulation is variable, but interannual production variations are
reproduced well for all provinces.