Isfahan University of Technology Dryland Soil Research (DLSR) 3115-9486 2 1 2025 03 01 Prediction of soil potassium forms using physicochemical properties and exchangeable potassium: II. Influence of soil properties on potassium distribution 81 88 3735 10.47176/dlsr.02.01.1043 EN Manoochehr Gholipoor Deparment of Soil Science, College of Agriculture, Shahrood University of Technology, 3619995161, Shahrood, Iran 0000-0001-9229-4744 Journal Article 2025 07 30 A quantitative understanding of the factors regulating potassium (K) distribution in soils is essential for optimizing nutrient management in agricultural systems, as K availability directly influences crop productivity and soil health. To systematically evaluate the drivers of K dynamics, a suite of artificial neural networks, coupled with Deep Learning Important FeaTures (DeepLIFT) attribution analysis, was employed to quantitatively assess the relative influence of key soil properties, including clay, silt, and sand content, pH, organic carbon (OC), cation exchange capacity (CEC), and electrical conductivity (EC), on four distinct K pools: water-soluble K, non-exchangeable K, fixed K, and total K. Additionally, the study investigated the association of soil initial water-soluble K and fertilizer K application rate with the fixation of K fertilizer to elucidate the relationship between them. Key findings revealed divergent drivers across K fractions, highlighting the complexity of K dynamics in soils. CEC emerged as the dominant factor influencing water-soluble K variability (+22.43%), underscoring its role in regulating K mobility. Clay content exhibited contrasting effects, positively influencing non-exchangeable K (+13.42%), total K (+20.59%), and fixed K (+13.81%), while negatively impacting water-soluble K (-14.38%). EC was the primary determinant of non-exchangeable K (+34.27%), suggesting salinity’s role in K retention. In contrast, pH showed a strong association with fixed K (+26.58%), reflecting its influence on interlayer trapping within 2:1 clay minerals. To bridge predictive modeling and practical applications, a genetic algorithm was integrated into an open-source, user-friendly Excel-based tool. This tool enables farmers and agronomists to optimize soil conditions for maximizing the sum of water-soluble and exchangeable K (plant-available K), thereby supporting precision nutrient management. By elucidating soil-specific K dynamics and providing actionable insights, this research advances sustainable K stewardship. The tool is accessible for download at: https://drive.shahroodut.ac.ir/index.php/s/fayE0zUH16TQe2M https://dlsr.iut.ac.ir/article_3735_dc0c398086fee58f9d64e1e47aa4e586.pdf