Bangyou Zheng1, Dr Pengcheng Hu2, Prof Scott C. Chapman3
1CSIRO, St Lucia, Australia, 2CSIRO, Canberra, Australia, 3University of Queensland, St Lucia, Australia
Biography:
Dr Bangyou Zheng is a data and experimental scientist at CSIRO in Brisbane, Australia. He received his PhD degree in agriculture science at China Agricultural University. His research focuses on crop physiology, crop genotype to phenotype prediction, crop modelling, climate adaptation, high throughput phenotyping, big data management, processing and visualization. He is familiar with R program language to solve questions in multiple disciplines.
Abstract:
The crop yield is strongly determined by interaction of genotype and environment, which are captured in the multiple environmental trials (METs) through statistics, machine learning and crop growth model. It is still a challenge to utilise dataset in the multiple scales of METs (e.g. remote sensing and field observations). Crop growth model provides the opportunity to integrate temporal and spatial datasets. The Wheat model in APSIM Next Generation was used to characterise environmental and genotypic variations in National Variety Trials from 2015 to 2021. The parameters related to wheat phenology were calibrated using observed Zadoks Score from flag leaf (Z39) to flowering (Z59). A virtual cultivar was selected to match the average ground coverage of all cultivars (simulated) and represent the average status in the trial level. Environmental variations were calibrated through matching the normalised difference vegetation index (NDVI) during growth season and average yields in the trial level (i.e. initial soil water and nitrogen) with three strategies (i.e. NDVI only, yield only, and NDVI and yield). Finally genotypic variations were calibrated for each cultivar across multiple trials (e.g. potential grain and leaf size). The results indicated the good agreement between observed and predicted yields for all cultivars among multiple trials (N=8589, R2 = 0.90, RMSE = 0.42). The calibrated environmental and genotypic variations in the NVT capture the major impact factors and can be used to represent the trial conditions for environmental characteristics in the further statistics analysis. The workflow also provides a solution to integrate remote sensing and field observation with crop growth models.