Advancing precision agronomy for minimizing production risk

Abstract

Farming in Northwestern Ontario faces unique challenges, including a shorter growing season, severe weather conditions, and limited infrastructure and support services. Despite these obstacles, the region holds great potential for expanding agricultural production, particularly for crops like soybeans. Soybean, a crop of significant economic and nutritional value, is susceptible to pests, diseases, and environmental stresses that reduce productivity. Effective health monitoring is crucial to optimize yields and quality. This study explored the use of low-cost proximal field cameras and remote sensing techniques for monitoring soybean leaf chlorophyll. A Mapir Survey3W camera was selected to capture high spatial resolution images in the green, red, and near-infrared regions of the electromagnetic spectrum. The optimal camera setup was investigated by comparing vertical (90º) and oblique (45º) orientation angles and automating image capture using a Raspberry Pi 4 Model B powered by a solar panel system. The vertical camera showed higher spectral reflectance values, while no significant difference was detected for vegetation indices. Once a series of images were captured using the identified optimal camera configurations, the images were preprocessed to obtain spectral reflectance values. Vegetation indices, such as the Green Normalized Difference Vegetation Index (GNDVI), were calculated from the captured images over the growing season. For calibration and validation purposes, at each field visit (within 7-10 days time), soybean leaf chlorophyll content (LCC) was measured using Apogee Instruments MC-100 Chlorophyll Meter. The correlation between GNDVI and LCC was established over time using the inverse function of piecewise linear regressions. The robustness of the regression models was measured by a Kolmogorov–Smirnov statistical comparison test between the predicted LCC over time and the field-measured LCC. The results were statistically not significant, indicating the similarity between the two data sets. Finally, a user-friendly prototype software application was built to make the proposed model accessible to the public. This study provided valuable insights into the optimal setup of field cameras and the use of low-cost remote sensing techniques for soybean leaf chlorophyll monitoring. The proposed methodologies and analyses contribute to the remote sensing techniques in agriculture using affordable sensors, supporting sustainable agriculture practices, and minimizing production risks in soybean cultivation.