Remote monitoring of soil physicochemical properties via UAV in precision agriculture: Systematic review and comparative analysis

The monitoring of soil physicochemical properties represents a fundamental requirement for precision agriculture, enabling site-specific resource management and the optimization of crop productivity. This systematic review analyses 21 scientific studies published between 2020 and 2025 on the application of UAV (Unmanned Aerial Vehicle) technologies for the remote monitoring of soil properties in agricultural contexts. The articles were selected through the PRISMA protocol from international scientific databases, focusing on applications related to soil moisture, fertility and nitrogen content, salinization, and organic matter. The comparative analysis highlights the clear predominance of indirect approaches based on vegetative response, with differentiated indices for each target parameter: NDRE proves superior to NDVI for leaf nitrogen, while the thermal indices TDDI and TVDI reach R²=0.966 for water content. Random Forest and ensemble methods demonstrate systematic superiority over SVM and linear models, with accuracy gains ranging from 3% to 15%. UAV-Satellite integration produces gains in R² of up to 0.384, combining the high spatial resolution of UAV with regional satellite coverage. Spectral mixing issues in row crops require specific unmixing or threshold optimization techniques, while acquisition timing proves critical for maximising correlations with agronomic parameters. Accuracy varies with soil depth, being optimal for surface layers (0–10 cm). Future directions include the development of sensors with greater penetration depth, integration with emerging technologies (such as LiDAR and hyperspectral sensing) and deep learning techniques, as well as the standardisation of multi-temporal protocols for the quantitative assessment of management practices over the long term.

Knowing what is happening beneath our feet is essential for modern farming. Healthy soil means better crops, less waste, and fewer chemicals. But measuring soil properties across large fields has traditionally been slow, expensive, and labour-intensive. This review analysed 21 scientific studies published between 2020 and 2025 to understand how agricultural drones — small, unmanned aircraft equipped with sensors and cameras — can help monitor key soil properties such as moisture, nitrogen, salinity, and organic matter, without touching the ground. The main finding is that drones do not usually "see" the soil directly. Instead, they read the health signals of the plants growing above it. A stressed or nutrient-deficient crop reflects light differently, and drone sensors can detect those differences with great precision. Specialised vegetation indices — essentially mathematical combinations of different light wavelengths — allow scientists to estimate what is happening in the soil from above. When drone data is combined with satellite imagery, accuracy improves even further, offering both fine detail at field level and broader regional coverage. Among the analytical methods tested, machine learning algorithms — particularly Random Forest — consistently outperformed traditional statistical models, improving prediction accuracy by up to 15%. One important limitation is depth: drone-based remote sensing works best for the topmost 10 cm of soil. Deeper layers remain harder to assess remotely. Looking ahead, researchers point to hyperspectral sensors, LiDAR, and artificial intelligence as the next frontier — tools that could make routine, large-scale soil monitoring faster, cheaper, and more accessible for farmers everywhere.