The main aims of this work were i) to test the ability of using multispectral variability and high spatial information from different remote sensing images to recognize land use changes by means of a new hybrid classification method and ii) to quantify and evaluate the temporal variation of main crop rotations.

The test area covered the 132 square kilometres of the Capitanata plain in Southern Italy (Apulia Region), including mostly agricultural landscape (wheat, sugar beet, asparagus, vineyard, olive grove). The comparison was made between a land use classification with a series of SPOT 5 images acquired in May 2007 and a set of WorldView-2 images acquired in April and July 2010. This new data set underlined better performances in terms of ground (2 m) and spectral resolution (eight bands: Blue, Coastal (blue), Green, Yellow, Red, Red edge, NIR 1, NIR 2).

The approach was based on a hybrid classification implemented in T-Map software by the spin-off company SI2G. The results showed an overall accuracy of more than 82%, were displayed on maps and used to calculate some indicators of crop sequences: This allowed of a quantification of more frequent rotation types (monoculture, rotations every 2 or 3 years) to promote ways of managing crop sequences compatible with environmental protection. Another indicator was the water requirement. According to the crops and the fields cropped, the total yearly average irrigation requirement was evaluated to correctly plan the irrigation water distribution.

Citation
Malinverni E S, M Rinaldi & S Ruggieri, 2012. Agricultural crop change detection by means of hybrid classification and high resolution images. EARSeL eProceedings, 11(2): 132-154