Time series analysis of EO-1 hyperion data for yield estimation at an agricultural site

A time series of EO-1 Hyperion images was acquired over the 2001/2002 summer growing season at the Coleambally Irrigation Area in Australia. This data, in conjunction with other remotely sensed and ancillary datasets collected over the site, is being used to develop a range of applications in precision agriculture. This paper evaluates the potential of Hyperion hyperspectral data for providing improved yield prediction. A number of hyperspectral indices were derived from the processed Hyperion images and compared with rice yield data over several paddocks. Results show that new hyperspectral indices such as the first derivative value and wavelength position of the red edge inflection point are more sensitive to within-paddock variations in yield than commonly used indices such as NDVI. It was also shown that the relationship between indices and yield varied among the paddocks. The influence of standing water during the early growth stages and early senescing and evergreen weeds in mid to late season on the spectral signal poses challenges in the development of yield prediction models for irrigated rice.