Assessment of paddy soil productivity by using the multiple linear mixed-effect regression model

Soil productivity makes an important contribution to rice yield. Selection of the key factors as well as model assessment for soil productivity can provide better guidelines for agricultural activity. Six 50-cm depth undisturbed typical paddy soil profiles from the high, moderate and low-yield fields, respectively, were collected in two locations (Dao county and Li county) in Hunan province and then stratified to measure soil physical and chemical properties. The Pearson’s correlation analysis was carried out to determine the significant factors, and soil productivity predictive models were developed using the multiple linear regression and the mixed-effect linear regression. Our results showed that the importance of soil properties was in the following order: cation exchange capacity (CEC) = Ca2+ > K+ > available K (AVK) > pH > clay particle content (CLAY); The mixed-effect regression model derived using the sample data by layers of 0-20 cm, 20-30 cm, 30-50 cm performed the best; it contained nine fixed effects (AVK, AVP, total P, Ca2+, K+, CEC, pH, CLAY and sand particle content) and four random effects (K10, SOC, Na+ and BD); and it can explain 85% of variability of soil productivity, in comparison with the determination of 73% produced by the ordinary multiple linear regression model.