نوع مقاله : مقاله پژوهشی
نویسندگان
1 استادیار پژوهش، بخش تحقیقات خاک و آب مرکز تحقیقات کشاورزی و آموزش منابع طبیعی استان کرمانشاه، سازمان تحقیقات، آموزش و ترویج کشاورزی،
2 استادیار پژوهش، بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی استان کرمانشاه، سازمان تحقیقات، آموزش و ترویج
3 موسسه تحقیقات خاک و آب، سازمان تحقیقات، اموزش و ترویج کشاورزی، کرج، ایران
4 پژوهشگر، بخش تحقیقات خاک و آب، مرکز تحقیقات و آموزش کشاورزی و منابع طبیعی استان کرمانشاه، سازمان تحقیقات، آموزش و ترویج کشاورزی، کرمانشاه،
چکیده
کلیدواژهها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Background and Objectives: Soil fertility is the capacity of soil to supply essential nutrients for plant growth and is a key factor in food security and sustainable agriculture. Low organic carbon is one of the most important constraints on soil fertility in Iran, with about 62% of soils containing less than 1% organic carbon. In addition, phosphorus and potassium deficiencies have been reported in 71.8% and 21% of agricultural soils, respectively. Preparing accurate nutrient maps using geostatistical methods, particularly kriging, can help identify fertility constraints and improve agricultural management. However, the risk component has received less attention in many studies. In risk zoning, the probability that a soil property falls above or below a specified threshold is determined. Therefore, the objective of this study was to produce risk maps of organic carbon, available phosphorus, and exchangeable potassium across approximately 336,000 hectares of the plains of Kermanshah Province using both parametric and nonparametric methods, including indicator kriging.
Materials and Methods: The study area is located in Kermanshah Province, between 46°22′ and 47°41′ E longitude and 33°48′ and 34°47′ N latitude. In this study, 530 surface soil samples were randomly collected from a depth of 0–30 cm, and soil organic carbon, available phosphorus, and exchangeable potassium were measured in the laboratory. Ordinary kriging was used to predict these properties, while universal kriging was applied when a spatial trend was detected, and different variogram models were fitted. Risk zoning was performed using both parametric and nonparametric approaches, including indicator kriging. In the nonparametric approach, continuous data were transformed into binary values based on the defined threshold. Finally, the performance of the two methods was evaluated using cross-validation and ROC curves.
Results: After removing outliers from the initial 530 samples, geostatistical analysis was performed on 525 soil organic carbon samples, 475 phosphorus samples, and 488 potassium samples. The mean soil organic carbon content was 1.35% (range: 0.1–3.61), the mean available phosphorus content was 8.6 mg kg⁻¹ (below the critical threshold of 15 mg kg⁻¹), and exchangeable potassium ranged from 158 to 878 mg kg⁻¹. The coefficient of variation was moderate for soil organic carbon and potassium, but high for phosphorus. Analysis of the spatial trend showed that about 17% of the variation in potassium could be explained by spatial trend; therefore, ordinary kriging was used to predict soil organic carbon and phosphorus, whereas universal kriging was applied for potassium. In risk zoning, the data were transformed into binary values based on the critical thresholds of 1% for soil organic carbon, 15 mg kg⁻¹ for phosphorus, and 330 mg kg⁻¹ for potassium. The spherical model provided the best fit for phosphorus, while the Matérn model performed best for soil organic carbon and potassium. ROC curve analysis showed that the parametric method outperformed the nonparametric method for soil organic carbon and phosphorus (AUC ≈ 0.64 and 0.59, respectively), whereas for potassium the two methods performed similarly, with an AUC of about 0.73.
Conclusion: The results of this study showed that the parametric method was more efficient than the non-parametric method for risk zoning of soil organic carbon and available phosphorus. However, for exchangeable potassium, both methods showed similar performance. The observed discrepancy in model performance may be attributed to limited sample size, the spatial distribution of samples across the study area, and local agricultural management practices. Overall, the generated risk maps provide a reliable framework for implementing integrated soil fertility management strategies and site-specific fertilizer recommendations.
کلیدواژهها [English]
)