Abstract
Spatial variation is a major source of error in agricultural field experiments affecting genotype performance prediction. Implementing statistical models that account for spatial effects can improve the prediction of genotype performance. This study evaluated the impact of the P-spline spatial correction method on the estimation of genetic parameters and AIC values in two distinct crops, wheat and cassava, using four models: Block, Block + Spatial, Block + Marker, and Block + Marker + Spatial. Analyses were performed on data from 115 and 68 trials obtained from the T3/WheatCAP and Cassavabase databases, respectively. As assessed using Cullis heritability estimates and AIC values, the results demonstrated that correcting for spatial variation improved analyses of grain yield, test weight, plant height, powdery mildew, stripe rust, and bacterial streak disease in wheat. Similar improvements were observed in cassava for dry matter content, dry yield, and plant height. However, no improvement was observed for cassava mosaic disease or bacterial blight. These results were consistent whether or not marker effects were fitted in the models. This study demonstrates that incorporating spatial correction into statistical analyses substantially improves the precision of variety evaluation. By accounting for field heterogeneity, spatial modeling complements experimental design and enhances the accuracy of treatment comparisons. Therefore, integrating robust experimental designs with appropriate spatial analyses is essential for achieving optimal precision and reliability in field trial evaluations.