Automated detection of phenological transitions for yellow flowering plants such as Brassica oilseeds. Issue 1 (26th November 2020)

• Record Type: Journal Article
• Authors: Sulik, John J.; Long, Dan S.
• Is Part Of: Agrosystems, geosciences & environment. Volume 3:Issue 1(2020)
• Page Start: n/a
• Page End: n/a
• Publication Date: 2020-11-26
• Subjects: Agriculture -- Periodicals; Agriculture -- Environmental aspects -- Periodicals; Soil science -- Periodicals; Food science -- Periodicals; Food science; Agriculture; Agriculture -- Environmental aspects; Electronic journals; Periodicals; 630
• DOI: 10.1002/agg2.20125
• Languages: English
• ISSNs: 2639-6696

Abstract

Abstract: Monitoring crop phenology is crucial for making site‐specific management decisions for crop protection and nutrition. The prominent yellow bloom associated with canola ( Brassica napus L.) and similar yellow‐flowering plants can provide cues about spatial differences as well as timing of crop input requirements. The objective of this study was to remotely characterize the phenological development of Brassicaceae oilseeds such as canola and carinata ( B. carinata A. Braun) in terms of spectral‐temporal dynamics between vegetation density and yellow flower density. Temporal variation of spectral indices (normalized difference vegetation index [NDVI], normalized difference yellowness index [NDYI], and visible atmospherically resistant index [VARI]) were measured in small plots over the growing season in relation to changes in vegetation density and flower density in winter canola and spring carinata. Phenological change between vegetative and reproductive development could be automatically detected using the difference in the change of the sign of ΔIndex values between VARI and NDYI. An overall accuracy of 85% was obtained when testing the algorithm with Landsat 8 data of canola fields near Olds, AB, Canada. The contrasting behavior between reproductive and vegetation indices across flowering transitions was confirmed for three independent datasets across a range of genetic variation in Brassica oilseeds as well as geographic variation in soil types and management practices. A bivariate time series analysis procedure was developed for automatically estimating flowering transitions based on predictable, relative differences between vegetative and reproductive indices. Researchers and land managers can exploit optimal phenology windows to improve site‐specific models and disease risk assessments.