What can we learn from a multi-season-stage-variety potato (Solanum tuberosum L.) study using aerial hyperspectral imagery?

Proper monitoring of plant nitrogen (N) status and yield forecasting is essential to achieving a healthy crop and to maximizing profitability, especially in N-demanding crops such as potato. The most common method of monitoring potato N status (nitrate-N analysis of petioles) by the potato farmers in Wisconsin is time-consuming, destructive, and is impractical to sufficiently characterize spatial-temporal variability. This study utilized narrow-band hyperspectral imagery (including the visible and near-infrared (VNIR) and shortwave infrared (SWIR) spectral regions) collected over two growing seasons from two potato varieties (Russet Burbank and Soraya) grown under varied N treatments to develop robust partial least squares regression (PLSR) models for predicting potato in-season and at-harvest traits related to N. The results indicate that some traits such as leaf total N content, within-season tuber yield, and the marketable yield and quality at harvest could be well predicted for both varieties (R 2 up to 0.78). The best spectral regions for those predictions varied depending on the growth stages of the plants, with VNIR predominating during early and mid-tuber, and SWIR during late tuber bulking. Our research suggests that the narrow-band hyperspectral imaging technique could be utilized to develop robust models to assist and potentially improve crop N fertilization decision-making, which will eventually result in higher input use efficiency of the cropping systems and better environmental stewardship for the farmers.