USE OF HIGH-RESOLUTION MULTISPECTRAL IMAGERY TO ESTIMATE SOIL AND PLANT NITROGEN IN OATS (Avena sativa)

Abstract:

Precision agriculture requires high spatial resolution in the application of the inputs to agricultural production. This requires that actionable information about crop and field status be acquired at the same high spatial resolution and at a temporal frequency appropriate for timely responses. In this study, high-resolution imagery was obtained through the use of a small, unmanned aerial vehicle, called AggieAir^TM, which provides spatial resolution as fine as 15 cm. Simultaneously with AggieAir flights, intensive ground sampling was conducted at precisely determined locations for plant and soil nitrogen among other parameters. This study investigated the spectral signature of oats and formulated a machine learning regression model of reflectance response between the multi-spectral bands available from AggieAir (red, green, blue, near infrared, and thermal), plant nitrogen and soil nitrogen. A multivariate relevance vector machine (MVRVM) was used to develop the linkages between the remotely sensed data and plant and soil nitrogen at approximately 15-cm resolution. The results of this study are presented, including a statistical evaluation of the performance of the model.