This study was carried out to analyze the spectral reflectance response of different nitrogen levels for corn crops. Four different nitrogen treatments of 0%, 80%, 100% and 120% BMP (best management practice) were studied. Principal component analysis-loading (PCA-loading) was used to identify the effective wavelengths. Partial least squares (PLS) and multiple linear regression (MLR) models were built to predict different nitrogen values. Vegetation indices (VIs) were calculated and then used to build more prediction models. Both full and selected wavelengths-based models showed similar prediction trends. The overall PLS model obtained the coefficient of determination (R2) of 0.6535 with a root mean square error (RMSE) of 0.2681 in the prediction set. The selected wavelengths for overall MLR model obtained the R2 of 0.6735 and RMSE of 0.3457 in the prediction set. The results showed that the wavelengths in visible and near infrared region (350-1000 nm) performed better than the two either spectral regions (1001-1350/1425-1800 nm and 2000-2400 nm). For each data set, the wavelengths around 555 nm and 730 nm were identified to be the most important to predict nitrogen rates. The vogelmann red edge index 2 (VOG 2) performed the best among all VIs. It demonstrated that spectral reflectance has the potential to be used for analyzing nitrogen response in corn.
|Original language||English (US)|
|Number of pages||9|
|Journal||International Journal of Agricultural and Biological Engineering|
|State||Published - Jan 1 2018|
- Effective wavelengths
- Principal component analysis-loading (PCA-loading)
- Vegetation indices (VIs)