Study and application of partial least squares regression on relationship between soil nutrient and fruit quality

Astract:【Objective】The objective of the study is to elucidate the duplicate collinearity diagnosis methodfor soil nutrient and to provide basis for establishment of multiple regression model of relationship between independent variable and dependent variable in fruit trees. Soil is the foundation of growth for ap-ple,and the types,physical properties,nutrient content and proportion relationship of soil directly affectthe distribution and nutrient uptake for apple root,thus affecting aboveground vegetative growth,flowerbud formation,yield and quality index. Nitrogen,phosphorus and potassium are often thought to be themacroelements for apple growth,with calcium,iron,zinc,boron,magnesium to be microelements. Thevarious elements and their interaction directly affect the growth of tree body and production quality.【Methods】An investigation of soil nutrients,‘Fuji'fruit quality was conducted at 153 commercial appleorchards in the production regions of 51 counties in China from 2011 to 2012. Three to five sampling plotswere randomly selected in August,September,for every orchard,with three trees selected from everysampling plot. The soil was collected in the outer canopy with earth borer respectively and was mixed ho-mogeneously. The analysis of soil nutrients were performed after indoor drying for the collected soil sam-ples. The soil organic matter were measured by the high temperature heated potassium dichromate oxida-tion volumetric method,the soil total nitrogen were measured by the Kjeldahl determination,the alkalihydrolyzale nitrogen were measured by the alkali solution diffusion method,the available phosphate weremeasured by the bicarbonate of soda leaching-molybdenum antimony colorimetric method,the availablepotassium were measured by the ammonium acetate extraction-flame photometer method,the effectiveiron,zinc were measured by the DTPA extraction-atomic absorption spectrophotometry,the availablecalcium were measured by the sodium acetate extraction-atomic absorption method,the available boronwere measured by the boiling water extraction-methylene amine-H acid colorimetric method. At the endof October to early November of each year,10 fruits were taken in the tree canopy of the southeast andsouthwest,with 90 or so fruits for each orchard. GY-1 fruit hardness tester was applied to measure thefruit hardness,PR-100 digital refractometer was applied to measure the soluble solid content,with 0.1mol·L^-1 Na OH was used to measure the titratable acid content.【Results】We surveyed 153 orchards in ap-ple production areas and calculated the eigenvalue,conditional number,and correlation coefficient ofsoil nutrients. The results showed the conditional number between soil nutrients was 61.31,the variancedecomposition proportion of soil p H and intercept was 0.96 and 0.95 respectively,the correlation coeffi-cient of soil p H and available calcium,iron was 0.82 and-0.824 respectively,which demonstrated thatthere was multicollinearity between various soil nutrient contents. Using soil organic matter,total nitro-gen,available potassium,available nitrogen,available phosphorus,available calcium,available zinc,available iron,available boron and soil pH as independent variable,fruit weight,fruit shape index,fruithardness,soluble solids content,titratable acid,solid acid ratio and fruit coloring area as the dependentvariable,we could carry on the model effect and weight analysis about the impact of soil nutrient contentson fruit quality and factors screening on major soil nutrients influencing the different fruit quality. The soilorganic matter,total nitrogen,available potassium,available nitrogen,available phosphorus,availablezinc,available iron and available boron showed positive correlation effect on fruit quality and weight,while the available calcium and soil p H showed negative correlation effect on fruit quality and weight,which illuminated that the adjustment of soil p H and reduction of available calcium content showed nega-tive load and weight on the quality of‘Fuji'apple in soil management of apple orchard in China. As forthe load and weight analysis about the factors of soil nutrients to fruit quality,the available iron and avail-able phosphorus showed great effect on fruit quality,followed by the available nitrogen and availablezinc. In order to illuminate the significance of the various factors about soil nutrients to fruit quality,thevariable importance for projection method was adopted to measure the VIP values about the impact of soilorganic matter,total nitrogen,available potassium,available nitrogen,available phosphorus,availablecalcium,available zinc,available iron,available boron and soil p H to fruit weight,fruit shape index,fruit hardness,soluble solids content,titratable acid,solid acid ratio and fruit coloring area. In general,the soil organic matter,total nitrogen,available potassium,available nitrogen,available phosphorus,available calcium,available zinc,available iron,available boron and soil p H had positive effects on fruitquality,while available calcium and soil p H were negative. The order from large to small in soil nutrientfactors affecting fruit qualities were obtained that fruit mass was mainly affected by the soil organic mat-ter,available nitrogen,available phosphorus,available calcium,available iron and available boron,while soluble solids were affected by soil organic matter,available phosphorus,available iron,availableboron and soil pH.【Conclusion】An analysis method and theoretical basis were provided for study andanalysis of relationships between dependent variables and multi-objective optimization,soil nutrient effect on fruit quality in fruit tree. The order of the impact about soil nutrient factors on fruit quality was clar-ified quantitatively and further studies were needed for the regression equations and optimum proposalsabout the impact of soil nutrients on the fruit quality.