Related Experiment Video
Updated: Jul 12, 2026
09:44
Use of Principal Components for Scaling Up Topographic Models to Map Soil Redistribution and Soil Organic Carbon
Published on: October 16, 2018
Machine learning-based estimation and spatial mapping of soil phosphorus and potassium using sentinel-2 and
Soraya Bandak1, Abdolhossein Boali2, Chooghi Bairam Komaki2
1Department of Water and Soil Sciences, Gorgan University of Agricultural Sciences and Natural Resources, Gorgan, Iran. Soraya.Bandak@gmail.com.
Scientific Reports
|July 9, 2026
Summary
This study shows that combining Sentinel-2 satellite data with Random Forest machine learning accurately maps soil phosphorus and potassium. This approach supports precision agriculture and sustainable land management in semi-arid regions.
Area of Science:
- Soil Science
- Remote Sensing
- Machine Learning
- Geospatial Analysis
Background:
- Accurate soil nutrient mapping is crucial for effective agricultural management.
- Traditional soil sampling is labor-intensive and provides limited spatial coverage.
- Remote sensing and machine learning offer promising alternatives for large-scale soil analysis.
Purpose of the Study:
- To evaluate the integration of remote sensing data with machine learning for estimating soil available phosphorus (Pav) and exchangeable potassium (Kex).
- To develop high-resolution digital soil nutrient maps for the Gonbad Kavous region.
- To compare the performance of different machine learning algorithms in predicting soil nutrient spatial distribution.
Main Methods:
- Collected 394 soil samples for Pav and Kex analysis.
- Utilized Sentinel-2 satellite imagery and environmental covariates as predictor variables.
- Trained and validated four machine learning algorithms: Random Forest (RF), Support Vector Machine (SVM), Boosted Regression Trees (BRT), and Generalized Linear Model (GLM).
Main Results:
- Random Forest (RF) demonstrated the highest predictive accuracy for both Kex (R² = 0.79) and Pav (R² = 0.83).
- Support Vector Machine (SVM) also showed good performance in capturing spatial variability.
- The study successfully generated high-resolution spatial distribution maps of Pav and Kex.
Conclusions:
- Combining Sentinel-2 spectral information with machine learning, particularly RF, is highly effective for digital soil mapping.
- The generated maps provide valuable data for soil fertility assessment and precision agriculture.
- This integrated approach supports sustainable nutrient and land management in semi-arid agricultural areas.
