PRINCIPAL COMPONENT ANALYSIS FOR PREDICTING CORN BIOMASS AND GRAIN YIELDS

Abstract

Soil variability, a rule rather than an exception, necessitates site-specific management for optimizing the efficient use of inputs. Biomass and grain yields depend on complex interactions among spatially variable physical and chemical properties of soil. The objectives of this study were to determine (i) the variability in soil properties using coefficient of variation (CV) and range as indicators and (ii) the relationships between principal components (PCs) derived from measured soil attributes and corn (Zea mays)biomass and grain yields. Soil physical and chemical properties were determined for four treatments: no-till corn without manure (Field 1), no-till corn with manure (Field 2), no-till corn-soybean (Glycine max) rotation (Field 3), and conventional tillage corn (Field 4). The CV for saturated hydraulic conductivity (Ks) (0.59 to 1.43) and infiltration rate at 5 min (i5) (0.38 to 0.86) was classified as most variable; for 3-h infiltration (ic) (0.30 to 0.97) and cumulative infiltration (I) (0.27 to 0.76) as moderate to most variable; for SOC (0.21 to 0.02) and MWD (0.12 to 0.23) as moderate to least variable; and the CV for WSA (0.03 to 0.12), pH (0.1 to 0.2), and soil bulk density (ρb) (0.05 to 0.11) as least variable. The highest CV overall was obtained for soil properties in the conventional till treatment (Field 4), although mean values of Ks, i5, and I for this field were the smallest. The biomass and grain yields were also the lowest for Field 4. The principal component (PC) analysis showed that four PCs with eigenvalues greater than one explained more than 78% of the variability in soil physical and chemical properties. The stepwise regression analysis showed that the grain yield (Yg) was linearly correlated with PCs as Yg = 8.77 + 0.81 PC1 (r2 = 0.19; P< 0.02) and biomass yield (Yb) as Yb = 4.85 + 0.41 PC1 + 0.32 PC2 (r2= 0.36; P < 0.003). The coefficients on PC1 were positive and were the highest for both Yg and Yb. Therefore, PC1 was the most dominant yield determinant.