Assessing Filtering and Smoothing of Active Sensor Data for the Management of N Applications in Corn.

Improved efficiency of Nitrogen (N) fertilizer applications is an issue of importance for the agricultural community. The environmental and economic considerations of optimizing N input to yield ratios warrant research in topics including optimal timing of fertilization, effectiveness of variable rate applications for improving Nitrogen Use Efficiency (NUE), and abilities of scientific methods to detect the onset and spatial variability of N stress. Remote sensing techniques have been shown to be capable of assessing the N status of crops. As with other remote sensing methods, active reflectance sensors do not require the labor intensive methods that soil or plant sampling entail. In comparison to passive reflectance measurements such as aerial or satellite imagery, active remote sensing has practical advantages in the ability to be conducted regardless of natural illumination conditions and in its capability to combine the acquisition of N status data with the application of the fertilizer, both of which improve the likelihood of completing the N prescription/ application process within the narrow temporal window available to producers. A unique issue with active remote sensing is that the data stream produced by the sensor is highly variable, particularly at earlier growth stages with partial canopy cover, due to the comparatively small field of view of the sensor, high frequency of data acquisition, and the constant motion of the platform. Data must be smoothed to avoid an erratic application of fertilizer. Gaps in the canopy may lead to smoothed data that prescribes excessive N application within N sufficient areas. Active sensor data was collected over N treatment plots at growth stages from V5 to VT from 2007 to 2015. The potential for filtering active sensor data points that contain minimal crop information and the optimal bin size for smoothing data to capture spatial variability of canopy N status is investigated.