Principal components approach for the linearization of yield prediction and tilth index equations

After 11 consecutive years of sustained farm input management treatments, a selection of soil physical properties was analyzed using a multivariate scheme to explore the impact that management systems (conventional, low input, and organic) have had on the physical state of Yolo loam soil. The crop in question was dry beans, and the time frame was the beginning of the flowering stage (46 days after planting). Principal Component Analysis was used to reduce the number of variables required to estimate the Tilth Index. The above strategy resulted in the suppression of multicollinearity, which in turn allowed for a more reliable estimation of variable coefficients once they were transformed back to the original coordinate system. Based on empirical considerations in recent reports of soil tilth definition and quantification, alternative linear expressions of the soil tilth index were developed using the coefficients of a principal component analysis. Even though this brought apparent improvements in the alternative expressions of soil tilth index, such as values having a wider range while remaining normalized from zero to unity, correlation with yield did not improve significantly.