The Use of Vegetation Indices in Comparison to Traditional Methods for Assessing Overwintering of Grain Crops in the Breeding Process

Abstract

A breeder’s job is a creative process that requires the routine collection and painstaking personal analysis of many types of data. This information is often subjective and cannot be statistically processed, relying only on the intuition of a scientist. One of these criteria for evaluating the source material is the visual assessment of overwintering varieties and breeding lines of winter crops. Overwintering expressed in points is approximate, and it is impossible to reliably compare breeding lines using it, especially when the points are approximately equal. For the first time in Russia, in the fields of FSBSI “Federal Scientific Center of Legumes and Groat Crops” (FSBSI FSC LGC) located in the Oryol region, elements of the developed method of remote sensing of small areas in the breeding process were tested based on “winter hardiness” of winter wheat, reliably expressed by vegetation indices. To assess breeding lines and crop varieties for winter hardiness, data were collected using unmanned vehicles at an altitude of 50 m and multispectral cameras with a resolution of 2 cm/pix. of Federal Scientific Agroengineering Center VIM. Previously, to determine the date of photography, an analysis of the weather conditions of the region for 20 years was carried out. It was revealed that regardless of the date of snow melting, the time of resumption of the spring vegetation of winter crops is located around the date of April 12. Accordingly, the third decade of April is recommended for optimal remote sensing of awakening crops in the region. Comparison of the data of vegetation indices (NDVI, NDRE, ClGreen) with the traditional results of visual scoring showed high reliability of the obtained digital data (correlation coefficient r > 0.7) for all three indices. It was noted that the NDVI index is more informative since it carries additional information not only about the preservation but also about the diversity of crops, i.e. on the presence of areas with fully or partially fallen out vegetation. The focal sparseness of crops was maximally characterized by the standard deviation of the NDVI index, increasing by 10…15% in the “dropped out” areas (correlation coefficient r =  −0.55) in comparison to the other two. An objective statistical assessment of digital information from small areas (plot area of 8 sq. m.) on 29 variants made it possible to additionally identify 6 cultivars (with a score of overwintering, comparable to the reference cultivar for the region) and compare the 18 variants that stood out among them. This would be impossible with traditional visual methods. In general, remote assessment of winter crops overwintering using vegetation indices based on comparison to an adapted reference variety allows you to quickly and without the involvement of a highly specialized and highly qualified specialist in the field of plant growing accumulate objective information from year to year on the overwintering of a characteristic collection of winter crops. Accumulation of such objective structured digital information for different years, tied to the data of the reference variety, will allow the formation of data arrays for further machine learning of specialized neural networks.