Spatial sampling of multi-environment trials data for station layout of maize variety

Classification methodology is widely used by plant breeders to group environments on the results of regional evaluation trials to assist selection among genotypes. To be effective, Multi-environment trial (MET) is a key point for breeders to test productivity and adaptability of crop varieties. With the development of science and technology in breeding, for METs at any stage of a breeding program, the breeder is dealing with a finite sample of environments that sample a relatively small subset of potential geographical and crop management conditions in a limited number of years. The general term ‘enviro typing’, used as a complement to the more familiar terms genotyping and phenotyping, is classifying the environments (location-year-management combinations) within a MET to assess how they represent target population of environments. However, the planting environment is complex and varied. Therefore, on one hand, the challenges associated with conducting METs in a finite number of years to predict genotype-trait performance in the target population-environment are crucial. On the other hand, it is also very important to ensure the accuracy of MET and decrease the loss of seed promotion to have a good understanding of the type of the planting environment and make reasonable decision to set the station of METs. In the paper, we mainly study on a new spatial sampling method of multi-environment trial layout and take maize in Jilin province in China for example. According to characteristics of METs, the algorithm combined with spatially stratified sampling and spatial balanced random sampling, which set a reasonable scheme of spatial distribution of MET station for maize. The main steps are as follows: Firstly, we build an index system for the classification of maize environment with sufficiently detailed data such as weather, soil, DEM, road network, cultivated area, actively accumulated temperature, precipitation, the sunshine duration during the whole growing period and gradient. Secondly, we develop methods for identifying the optimal numbers of classes for the planting environment and use spatially stratified sampling model to calculate the minimum number of stations to meet the requirements. Finally, the method was applied to the spatial distribution of MET station for maize in Jilin province, and compute the optimal number of clusters for planting environment and METs stations number, which are 5 and 25, respectively. The result is roughly consistent with the existing MET station for maize with the sample precision of 94.5%, and have a certain reference significance for the layout of MET stations for maize in Jilin province.