Identification of Relationships between Cotton Yield, Quality, and Soil Properties

Abstract

most common ways to describe the relationship between two or more variables. Correlation coefficients are a mea-Intercorrelation among soil properties can result in multicollinearsure of the degree of linear association between any two ity problems regarding relationships between soil properties and crop yield. The objective of this study was to compare statistical methods variables when other variables are fixed. The regression of examining relationships between cotton (Gossypium hirsutum L.) model can indicate the extent by which the dependent yield, quality, and soil properties. Soil and plant samples were collected variable can be predicted by the independent variables; from 1-ha grids on an irrigated production cotton field in Texas from the strength of prediction is expressed as r 2 , known as 1998 through 2000. Ordinary least square regression (OLS), partial variation in dependent variable explained by indepenleast square regression (PLS), and principal component regression dent variables or coefficient of determination. The inter-(PCR) were compared as methods for quantifying relationships becorrelation between soil properties does not violate the tween cotton yield or quality and soil properties. The PLS method assumptions of the classical regression modeling, but can eliminated multicollinearity problems and resulted in the coefficient cause multicollinearity problems that may result in large estimations with meaningful signs compared with their associations variances of parameter estimates, unexplainable signs to cotton yield and fiber quality. Furthermore, loadings from linear combinations of variables in PLS allowed identifying soil properties of explanatory variables, and rejection of regression cothat had the greatest influence on yield. While PCR identified the prin-efficient estimates that are significant (Morzuch and cipal components that maximized the variance of independent vari- Ruark, 1991; Neter et al., 1996; Fekedulegn et al., 2002). ables, it did not improve the modeling of crop-soil relationships. AmongThe general methods to solve multicollinearity probthe selected soil and landscape properties, sand and clay content, exlems between independent variables include removing changeable Ca 2؉ and Mg 2؉ , NO 3 Ϫ , Olsen-P, pH, relative elevation, and less important variables; combining variables via princislope were important factors affecting lint yield and fiber quality.pal component analysis (PCA) or factor analysis; andHigher lint yields were usually accompanied by higher fiber quality.using ridge regression, partial least squares (PLS), or Magnitudes of influence of different soil properties on yield and qual-Bayesian methods (Neter et al., 1996). Each method has ity, however, varied among the 3 yr, suggesting that long-term studies its own limitations. For example, variable removal can are needed to establish robust relationships for site-specific management.focus on the influential variables, but could lose some information due to reduced number of independent variables (Morzuch and Rua...

Influence of soil properties on crop yield: a multivariate statistical approach

“…To solve multicollinearity problems between independent variables, several authors applied partial least squares regression (PLS) (Corwin et al, 2003;Ping et al, 2004). Loadings from linear combinations of variables in PLS allowed identifying soil properties that have the greatest influence on yields.…”

Sunflower yield related to multi-temporal aerial photography, land elevation and weed infestation

“…3. The stepwise regression (SR) and principal component analysis (PCA) of the multispectral dataset from field 1 for the three studied dates were also performed to reduce multi-collinearity between independent variables (Ping et al 2004) and to determine the combined relationship of all the spectral variables on yield. The correlation matrix was used to calculate the eigenvalues and the eigenvectors for the PCA.…”

Yield Prediction of some Orchard Trees Based on Soil Characteristics under Rainfed Conditions

“…The good relationship among the studied soil characteristics was extracted in the form of the following equation that produced a R 2 of 0.892 and RMSE of 0.093, Fig (4). Olive yield = 1.48571 + 0.00314*Depth -0.00914*Gravels -0.00918*Sand + 0.03213*Silt + 0.00729*Clay + 0.01542*Salinity -0.04413*pH -0.00863* CaCO 3 Whereas the output from the linear combination of variables in PLSRM, stated by (Ping et al, 2004), allows identifying soil properties that have the greatest influence on yield. Accordingly, the current study indicated that most influencing soil properties could be arranged in the order of; gravel, depth, silt, sand, carbonate, and clay, while salinity and pH are not having significant influences.…”

“…However, neglecting some variables could contribute to missing important information. Several authors have applied partial least squares regression (PLS) to overcome the multicollinearity problems between independent variables (Corwin et al, 2003); (Ping et al, 2004). Running linear combinations of variables in PLS allowed to identify soil properties that own the greatest influence on yields.…”