Low ground-cover filtering to improve reliability of the nitrogen reflectance index (NRI) for corn N status classification

Recent advancements in low-cost remote sensing equipment and techniques have proven that crop parameter classification can be practical in agricultural situations. One problem still remaining is how to account for in-field variability of mixed ground-cover areas. This study focused on finding a non-invasive method of screening low ground-cover areas before assessing crop nitrogen (N) status. Four years of plot data were used where intense ground sampling was coupled with nadir-viewing remotely sensed data from a ground-based remote sensing system. Data were collected over irrigated corn (Zea mays L.) with both light and dark soil backgrounds to develop a relationship between leaf area index (LAI) and remotely sensed data. An index based on subtracting the red from the green reflectance (green - red) was shown to be strongly related to LAI. A concurrent assessment of the N reflectance index (NRI) over the four-year study indicated that the corn N status classification improved with increasing LAI. Using the (green - red) index and a screening value corresponding to an LAI of 2.5, the data from the four-year study were screened and the NRI was reassessed. The initial screening removed all bare-soil and many of the measurements collected before the V12 growth stage. In this example, 94.0% of the data associated with LAI below 2.5 was correctly filtered out, and 95.4% of the data collected over areas with LAI at or above 2.5 correctly passed the filter. This method of filtering was shown to be highly effective in screening low ground-cover and bare-soil areas before classification. Future incorporation of this technique into on-the-go filtering algorithms could improve crop parameter classification from remotely sensed measurements in early-season and mixed vegetative cover situations.