Using Reflectance Spectroscopy and Cloud Computing in Analyzing Soil

Recently, a team of researchers from the University of São Paulo (USP), in collaboration with the Federal University of Amazonas (UFAM), explored a new way to improve soil analysis. Their study, which was published in the Journal of Environmental Management, shows how advanced spectroscopic methods, including reflectance spectroscopy, can provide faster, cheaper, and more sustainable alternatives to traditional laboratory-based soil testing (1). The researchers tested their method in the Amazon River Basin (ARB), which is one of the most challenging ecosystems to study.

The ARB is a large ecosystem that spans a sizable portion of South America. Covering more than 7 million square kilometers and nine countries (Bolivia, Brazil, Colombia, Ecuador, Guyana, Peru, Suriname, and Venezuela), the ARB lies east of the Andes Mountain range (2). Most of the ARB is tropical rainforest and contains more than 56% of all broad-leaf forests on Earth (2). What makes the ARB a fascinating area of the globe for researchers is its biodiversity of plant and animal life. It is estimated that more than 30,000 plant species, 60 reptile species, 2,000 fish species, and 35 mammal families call the ARB home (2).

Reflection of a sunset by a lagoon inside the Amazon Rainforest Basin. The Amazon river basin comprises the countries of Brazil, Bolivia, Colombia, Ecuador, Guyana, Suriname, Peru and Venezuela. | Image Credit: © SL-Photography - stock.adobe.com

Apart from studying biology in the ARB, researchers are also interested in learning more about the soil composition in the ARB. Traditional wet laboratory methods are limited by scarce facilities, high costs, and the intensive labor required to process large data sets (1).

As part of their experimental procedure, the researchers collected 211 soil samples from the ARB and subjected them to visible, near-infrared (NIR), and short-wave infrared (Vis-NIR-SWIR) spectroscopy in the laboratory. These spectra were then used to predict key soil attributes, including soil organic carbon (SOC) and particle size distribution (sand, silt, and clay) (1). The researchers tested two computational approaches. The first one was a cloud-based method using the Brazilian Soil Spectral Service (BraSpecS) platform. The second was an offline approach using the R programming language (1). Both relied on the Cubist machine learning (ML) algorithm to model soil properties.

To evaluate the performance of both methods, the researchers considered variables such as the coefficient of determination (R²), mean absolute error (MAE), and root mean squared error (RMSE). The results showed that both approaches provided reliable predictions, but the offline model consistently outperformed the cloud-based platform in terms of precision (1). For example, in predicting clay content, the offline model achieved an R² of 0.85 compared to 0.70 for the online model (1). Similarly, for SOC prediction, the offline method delivered an R² of 0.81 versus 0.72 for the cloud-based approach (1).

When comparing the two methods, the researchers concluded that cloud computing had several key advantages compared to the other method. Unlike offline models, the cloud-based approach does not require locally observed training data, making it particularly practical for real-time, large-scale applications in remote environments (1). This flexibility is critical for regions like the Amazon, where accessibility and resource limitations make conventional methods impractical (1).

Another key finding of the study was the strong sensitivity of NIR and SWIR spectral regions (350–2500 nm) to key soil properties, such as clay minerals, iron oxides, and soil organic matter composition. This spectral responsiveness enhances the potential for precise, non-destructive soil characterization (1).

Accurate soil mapping is essential for monitoring carbon stocks, managing agricultural practices, and supporting conservation policies in the Amazon, a region central to global climate regulation (1). The researchers argue that integrating soil spectroscopy with cloud computing can help inform sustainable development policies, optimize land use, and improve environmental management strategies (1).

Ultimately, the synergistic use of spectroscopy and cloud platforms offers not only technical efficiency but also a sustainable solution to managing one of Earth’s most critical natural resources. As the Amazon faces increasing environmental pressures, such innovations could prove pivotal in balancing development with conservation (1).

References

1. Novais, J. J. M.; Melo, B. M. D.; Neves Jr., A. F.; et al. Online Analysis of Amazon's Soils through Reflectance Spectroscopy and Cloud Computing can Support Policies and the Sustainable Development. J. Environ. Manag. 2025, 375, 124155. DOI: 10.1016/j.j.envman.2025.124155
2. Organization of American States, Amazon River Basin. OAS.org. Available at: https://www.oas.org/dsd/events/english/documents/osde_8amazon.pdf (accessed 2025-09-18).