Impact of Soil Factors on the Yield and Agronomic Traits of Hemerocallis citrina Baroni in the Agro-Pastoral Ecotone of Northern China

The ecologically fragile agro-pastoral ecotone in northern China is characterized by relatively poor arable land quality. Yunzhou District in Datong City, which is situated within this transitional zone, boasts over 600 years of Hemerocallis citrina Baroni cultivation. Exploring the effects of soil physicochemical properties on daylily yield and related agronomic traits is essential for enhancing the ecological and economic value of dominant crops in ecologically fragile areas. Therefore, in this study, we focused on the daylily, a characteristic cash crop that is grown in the agro-pastoral ecotone in Yunzhou District. Physicochemical property measurement and yield estimation were performed using soil samples collected from 37 sites, with Spearman’s correlation analysis, one-way analysis of variance with multiple comparisons, path analysis, and stepwise regression analysis used to analyze the generated data. The results showed the following: (1) The pathway analysis of daylily yield with each agronomic trait showed that the BN and PH directly affected the yield of daylily with direct pathway coefficients of 0.844 and 0.7, respectively, whereas the SN indirectly affected the yield of daylily through the BD and PH, with indirect pathway coefficients of 0.827 and 0.566, respectively. (2) A total of four principal components were extracted for the soil factors, of which SMC, ST and BD had large loadings on PC1; OM, TN and pH had large loadings on PC2; AK had large loadings on PC3; and AP had large loadings on PC4. (3) From the principal component regression and stepwise regression, it can be seen that SMC is the most critical factor affecting the yield of daylily, as well as the related agronomic traits, and the results also show that yield prediction was affected by OM, ST, and AK, while BN was influenced by OM and ST, and SN and PH were influenced by AP. Comparing the goodness of fit and significance of the two models, it can be concluded that the stepwise regression model is the optimal model for this study.