Variable-rate in corn sowing for maximizing grain yield

Abstract

Sowing density is one of the most influential factors affecting corn yield. Here, we tested the hypothesis that, according to soil attributes, maximum corn productivity can be attained by varying the seed population. Specifically, our objectives were to identify the soil attributes that affect grain yield, in order to generate a model to define the optimum sowing rate as a function of the attributes identified, and determine which vegetative growth indices can be used to predict yield most accurately. The experiment was conducted in Chapadão do Céu-GO in 2018 and 2019 at two different locations. Corn was sown as the second crop after the soybean harvest. The hybrids used were AG 8700 PRO3 and FS 401 PW, which have similar characteristics and an average 135-day cropping cycle. Tested sowing rates were 50, 55, 60, and 65 thousand seeds ha^-1. Soil attributes evaluated included pH, calcium, magnesium, phosphorus, potassium, organic matter, clay content, cation exchange capacity, and base saturation. Additionally, we measured the correlation between the different vegetative growth indices and yield. Linear correlations were obtained through Pearson's correlation network, followed by path analysis for the selection of cause and effect variables, which formed the decision trees to estimate yield and seeding density. Magnesium and apparent electrical conductivity (EC_a) were the most important soil attributes for determining sowing density. Thus, the plant population should be 56,000 plants ha^-1 to attain maximum yield at EC_a values > 7.44 mS m^-1. In addition, the plant population should be 64,800 plants ha^-1 at values < 7.44 mS m^-1 when magnesium levels are greater than 0.13 g kg^-1, and 57,210 plants ha^-1 when magnesium content is lower. Trial validation showed that the decision tree effectively predicted optimum plant population under the local experimental conditions, where yield did not significantly differ among populations.

Figure 1

Experimental sites in the Alto Formoso (A) and Porto Seguro (B, C) farms. The software used to create this Figure was ArcGIS Desktop (v10.5, https://desktop.arcgis.com/en/system-requirements/10.5/).

Figure 2

Experimental plots and planned seeding rates for sites A and B in the 2017/2018 cropping season and sites C and D in the 2018/2019 cropping season (used to validate the results). The software used to create this Figure was ArcGIS Desktop (v10.5, https://desktop.arcgis.com/en/system-requirements/10.5/).

Figure 3

Mean rainfall (mm), maximum, minimum, and mean temperature (°C) recorded at 10-day intervals during the corn cultivation in 2018 (A) and 2019 (B). Chapadão do Céu. The software used to create this figure was Microsoft Excel (v1804, https://www.microsoft.com).

Figure 4

Final plant stands in sites A and B used for path analysis and tree decision and in area C used to validate results. The software used to create this Figure was ArcGIS Desktop (v10.5, https://desktop.arcgis.com/en/system-requirements/10.5/).

Figure 5

Pearson correlation network between soil attributes pH, calcium (Ca), magnesium (Mg), cation exchange capacity (CEC), potassium (K), phosphorus (P), organic matter (OM), clay, apparent electrical conductivity (EC_a), and grain yield (GY). The package used of R to create this figure was qgraph (v1.6.9, https://cran.r-project.org/web/packages/qgraph/index.html).

Figure 6

Path analysis of soil variables as a function of grain yield. Softwares used to create this figure were Genes (v1999.2019.44, http://arquivo.ufv.br/dbg/genes/gdown1.htm) and Corel draw (vX7, https://www.corel.com/).

Figure 7

Decision tree generated using apparent electrical conductivity (EC_a) and magnesium (Mg) as the variables selected by the trial analysis. The software used to create this figure was Genes (v1999.2019.44, http://arquivo.ufv.br/dbg/genes/gdown1.htm).

Figure 8

Maps of EC_a (A, D) and Mg (B, E) used in the decision tree and grain yield (GY, C, F) for sites B and C using VRS, respectively. The software used to create this figure was ArcGIS Desktop (v10.5, https://desktop.arcgis.com/en/system-requirements/10.5/).