Application of geographically weighted regression to improve grain yield prediction from unmanned aerial system imagery

Abstract

Phenological data are important ratings of the in-season growth of crops, though this assessment is generally limited by the availability of spatial and temporal data during the growth cycle. Unmanned aerial systems (UAS) have the potential to provide high spatial and temporal resolution for phenotyping tens of thousands of small field plots without requiring substantial investments in time, cost, and labor. The objective of this research was to determine whether an accurate remote sensing-based method could be developed to estimate grain yield using aerial imagery in small-plot wheat (Triticum aestivum L.) yield evaluation trials. The UAS consisted of a modified consumer-grade camera mounted on a low-cost unmanned aerial vehicle and was deployed multiple times throughout the growing season in yield trials of advanced breeding lines with irrigated and drought-stressed environments at the International Maize and Wheat Improvement Center in Ciudad Obregon, Sonora, Mexico. We assessed data quality and evaluated the potential to predict grain yield on a plot level by examining the relationships between information derived from UAS imagery and the grain yield. Using geographically weighted (GW) models, we predicted grain yield for both environments. The relationship between measured phenotypic traits derived from imagery and grain yield was highly correlated (r = 0.74 and r = 0.46 (p < 0.001) for drought and irrigated environments, respectively). Residuals from GW models were lower and less spatially dependent than methods using principal component regression, suggesting the superiority of spatially corrected models. These results show that vegetation indices collected from high-throughput UAS imagery can be used to predict grain and be used for selection decisions as well as enhance genomic selection models