Quantifying the relationships of soil properties and crop growth with yield in a NPK fertilizer application maize field

Abstract

Spatial variability of soil properties, crop performance, fertilization management, and tillage methods play an important role in yield prediction and improvement. However, the farm nutrient management in most regions of the world is still based on conventional and uniform management strategies, without take into consideration the spatial variability of the above factors. Therefore, the objectives of this study were to delineate potential management zones (MZs), with the considering of spatial variability of soil properties, crop performance, fertilization management, tillage methods, and investigate their relationships with yield, and their contributions to yield in each management zone. A 1.60 ha maize field with fertilizer and tillage management, in Lishu County, Jilin Province, China, was selected as the study area and soil samples with soil organic matter, total nitrogen and soil pH measured values, crop samples with crop height, leaf nitrogen content, above ground biomass at V6, silking, and maturity growth stages, and yield measured values were collected in 2019 and 2020. Firstly, the rectangular MZs were defined to divide the study area into relatively homogeneous areas based on the spatial variability of soil properties, crop performance, and yield of 2019 and 2020. Secondly, the correlations of soil properties and crop performance with yield were obtained across MZs and growth stages in 2019 and 2020. Lastly, the yield prediction models were established based on the above four types of factors by using random forest (RF) model and the contributions of main factors were obtained quantitatively across different MZs and years. The results showed that (1) the small maize field was divided into three rectangular MZs and most of the soil and crop attributes presented homogeneous characteristics in each zone; (2) correlations of soil properties and crop performance with yield varied across different MZs and growth stages, and even different years; (3) RF model obtained good performance across MZs and years, with R2 = 0.86 and RMSE = 0.77 kg/ha in 2019 and R2 = 0.89 and RMSE = 1.04 kg/ha in 2020 both for the whole field. The contribution of main factors for yield varied across both MZs and years, while some characteristics of the main factors’ contributions remain unchanged. This study would open the opportunity to farmers and agricultural experts for site-specific nutrient management in the maize field.