AJP Asia Intelligence Agency

AI Predicts Wheat Harvest Timing with One-Day Accuracy

The Rural Development Administration has developed artificial intelligence (AI) technology that predicts the timing of wheat heading with an average error of one day by utilizing genetic information of domestic wheat. This advancement allows for the prediction of heading time based solely on genetic information, moving away from the traditional reliance on actual cultivation and surveys.

On November 17, the Rural Development Administration announced the establishment of a reference genome for the representative domestic wheat variety, Geumgang wheat, and the development of a model for predicting heading traits based on this genome.

Wheat is the second most consumed grain in the country after rice, but most of the demand is met through imports. To increase self-sufficiency in wheat, the Rural Development Administration has been promoting the expansion of double cropping of rice and wheat and the development of early-harvesting varieties.

To cultivate wheat and rice on the same farmland, wheat must be harvested early to align with the summer monsoon and rice planting schedules. Therefore, developing varieties that head early is crucial.

Previously, to select fast-heading wheat varieties, researchers had to cultivate wheat directly in test fields and investigate the heading times individually. This process required significant time and manpower.

To overcome these limitations, the Rural Development Administration conducted research analyzing the genetic information of domestic wheat. They first decoded the genome of Geumgang wheat to establish a reference genome.

Until now, the standard genome of the Chinese variety Chinese Spring, established by the International Wheat Genome Sequencing Consortium (IWGSC), had been used in domestic wheat research. This limited the precise analysis of the unique genetic traits of domestic wheat and differences between varieties.

This research has laid the groundwork for comparing and analyzing genomes based on Geumgang wheat. The related paper was approved for publication in the international journal Scientific Data last October.

The research team utilized 614 core wheat accessions established by the National Institute of Crop Science in 2018 and heading time survey data accumulated over eight years. By analyzing the genetic information of the core wheat accessions using supercomputers based on the Geumgang wheat reference genome, they identified approximately 600 million genetic variations. They subsequently selected 10,725 candidate genetic variations associated with heading time.

The research team trained the selected genetic variations and actual heading time data in AI to develop a model that predicts heading time based solely on genetic information. The prediction results showed an average difference of just one day from the actual heading date.

The Rural Development Administration has developed a digital breeding platform that provides the Geumgang wheat reference genome, genotypes, heading time data, and prediction models for use in breeding practices. The platform is set to be made public in 2027.

In the future, the prediction targets will expand beyond heading time to include resistance to Fusarium head blight and cold tolerance, and the range of applicable crops will also increase. The plan is to broaden the utilization scope to private seed companies through platform enhancement.

Ok Hyun-chung, head of the Agricultural Bio-Resources Division at the Rural Development Administration, stated, "This research is a collaborative achievement that combines the wheat resources of the National Institute of Crop Science, eight years of survey data, and the genomic analysis and AI technology from the National Agricultural Science Center. We will continue to lead the digital breeding transition for various crops based on such collaborations."




* This article has been translated by AI.