Normalization of uncalibrated late-season digital aerial imagery for evaluating corn nitrogen status

Using uncalibrated digital aerial imagery (DAI) for diagnosing in-season nitrogen (N) status of corn (Zea mays L.) is challenging because of the dynamic nature of corn growth and the difficulty of obtaining timely imagery. Late-season DAI is more accurate for identifying areas deficient in N than early-season imagery. Even so, the quantitative use of the imagery across many fields is still limited because DAI is often not radiometrically calibrated. This study tested whether spectral characteristics of corn canopy derived from normalized uncalibrated late-season DAI could predict final corn N status. Color and near-infrared (NIR) imagery was collected in late August or early September across Iowa from 683 corn fields in 2006, 824 in 2007, and 828 fields in 2007. Four sampling areas (one within a target-deficient area) were selected within each field for conducting the end-of-season corn stalk nitrate test (CSNT). Each image was enhanced to increase the dynamic range within each field and to normalize reflectance values across all fields within a year. The reflectance values of individual bands and three vegetation indices were used to predict corn N status expressed as Deficient and Sufficient (a combination of marginal, optimal, and excessive CSNT categories) using a binary logistic regression (BLR). The green reflectance had the highest prediction rate, which was 70, 64, and 60% in 2006, 2007, and 2008, respectively. The results suggest that the normalized (enhanced) late-season uncalibrated DAI can be used to predict final corn N status in large-scale on-farm evaluation studies.