Supplementary file 1_Soil geochemical associations of micronutrients in arid agroecosystems: insights from machine learning and structural equation modeling.docx

Introduction

Micronutrient availability in arid agricultural soils is highly variable and difficult to predict due to the complex interactions among soil physical and chemical properties. This study explored the geochemical associations of iron (Fe), zinc (Zn), manganese (Mn), and copper (Cu) across a large soil dataset from North Africa.

Methods

The study used a combination of statistical and modeling approaches, including principal component analysis (PCA), machine learning (XGBoost), and structural equation modeling (SEM).

Results

The soils were predominantly neutral to moderately alkaline (mean pH = 7.68) and showed considerable variation in soil organic carbon (SOC), cation exchange capacity (CEC), and texture. Clear patterns emerged across analyses. Texture played a central role: clay-rich soils were associated with higher SOC and CEC, while sandy soils showed reduced nutrient retention. Micronutrients followed distinct relationships: Fe was moderately associated with Zn, whereas Mn and Cu were strongly coupled and closely linked to CEC and clay content. The machine learning model identified Fe as the strongest predictor of Zn variability, followed by Mn and sand content. Structural equation modeling suggested negative associations of pH with Fe, Mn, and Cu, while SOC showed positive associations with Fe and Zn but negative associations with Mn and Cu. CEC was strongly associated with Mn and Cu, highlighting the importance of soil charge properties.

Conclusion

The results point to the presence of two contrasting micronutrient systems: one associated with Fe and Zn, and another dominated by Mn and Cu. These findings suggest that micronutrient variability in arid soils is unlikely to reflect a single factor, but instead reflects the combined associations with soil texture, chemical properties, and geochemical interactions. This integrated approach provides a clearer understanding of micronutrient behaviour in dryland systems and offers a useful framework for improving soil fertility management in arid agricultural landscapes.