The Brazilian Soil Spectral Library (VIS-NIR-SWIR-MIR) Database: Open Access

Abstract:

NEW VERSION V.002 (Some Lat Long Coordinates added).

Soil spectroscopy has emerged as a solution to the limitations associated with traditional soil surveying and analysis methods, addressing the challenges of time and financial resources. Analyzing the soil's spectral reflectance enables to observe the soil composition and simultaneously evaluate several attributes because the matter, when exposed to electromagnetic energy, leaves a "spectral signature" that makes such evaluations possible. The Soil Spectral Library (SSL) consolidates soil spectral patterns from a specific location, facilitating accurate modeling and reducing time, cost, chemical products, and waste in surveying and mapping processes. Therefore, an open access SSL benefits society by providing a fine collection of free data for multiple applications for both research and commercial use.

BSSL Description and Usefulness

The Brazilian Soil Spectral Library (BSSL), available at https://bibliotecaespectral.wixsite.com/english, is a comprehensive repository of soil spectral data. Coordinated by JAM Demattê and managed by the GeoCiS research group, the BSSL was initiated in 1995 and published by Demattê and collaborators in 2019. This initiative stands out due to its coverage of diverse soil types, given Brazil's significance in the agricultural and environmental domains and its status as the fifth largest territory in the world (IBGE, 2023). In addition, a Middle Infrared (MIR) dataset has been published (Mendes et al., 2022), part of which is included in this repository. The database covers 16,084 sites and includes harmonized physicochemical and spectral (Vis-NIR-SWIR and MIR range) soil data from various sources at 0-20 cm depth. All soil samples have Vis-NIR-SWIR data, but not all have MIR data.

The BSSL provides open and free access to curated data for the scientific community and interested individuals. Unrestricted access to the BSSL supports researchers in validating their results by comparing measured data with predicted values. This initiative also facilitates the development of new models and the improvement of existing ones. Moreover, users can employ the library to test new models and extract information about previously unknown soil properties. With its extensive coverage of tropical soil classes, the BSSL is considered one of the most significant soil spectral libraries worldwide, with 42 institutions and 61 researchers participating. However, 47 collaborators from 29 institutions have authorized the data opening. Other researchers can also provide their data upon request through the coordinator of this initiative.

The data from the BSSL project can also help wet labs to improve their analytical capabilities, contributing to developing hybrid wet soil laboratory techniques and digital soil maps while informing decision-makers in formulating conservation and land use policies. The soil's capacity for different land uses promotes soil health and sustainability.

Coverage

The BSSL data covers all regions of Brazil, including 26 states and the Federal District. It is in a .xlsx format and has a total size of 305 Mb. The table is structured in sheets with rows for observations, and columns, representing various soil attributes in the surface layer, from 0 to 20 cm depth. The database includes environmental and physicochemical properties (22 columns and 16,084 rows), Vis-NIR-SWIR spectral bands (2151 columns and 16,084 rows), and MIR channels (681 columns and 1783 rows). An ID unique column can merge the sheet for each attribute or spectral range.

Accessing original data source

Using these data requires their reference in any situation under copyright infringement penalty. Three mechanisms are available for users to reach the original and complete data contributors:

a) Refer to sheet two for name and code-based searches;

b) Visit the website https://bibliotecaespectral.wixsite.com/english/lista-de-cedentes or locate the contributors' list by Brazilian state;

c) Visit the website of the Brazilian Soil Spectral Service – Braspecs http://www.besbbr.com.br/, an online platform for soil analysis that uses part of the current SSL (Demattê et al., 2022) - It was developed and managed by GeoCiS. There, owners from all over the country can be found.

Proceeding to data analysis

We registered and organized the samples at the ESALQ/USP Soil Laboratory. Some samples arrived without preliminary data analyses, so we analyzed them for soil organic matter (SOM), granulometry, cation exchange capacity (CEC), pH in water, and the presence of Ca, Mg, and Na, following the recommendations of Donagemma et al. (2011).

The GeoCiS research group performed spectral analyses following the procedures described by Bellinaso et al. (2010). Demattê et al. (2019) provide detailed methods for sampling, preparation, and soil analyses, including reflectance spectroscopy. Latitude and longitude data can be requested directly from the data owner. In summary, the following steps are involved in data acquisition.

a) We subjected the soil samples to a preliminary treatment, which involved drying them in an oven at 45°C for 48 hours, grinding them, and sieving them through a 2mm mesh;

b) We placed the samples in Petri dishes with a diameter of 9 cm and a height of 1.5 cm;

c) We homogenized and flattened the surface of the samples to reduce the shading caused by larger particles or foreign bodies, making them ready for spectral readings;

d) The spectral analyses took place in a darkened room to avoid interference from natural light. We used a computer to record the electromagnetic pulses through an optical fiber connected to the sensor, capturing the spectral response of the soil sample;

e) We obtained reflectance data in the Visible-Near Infrared-Shortwave Infrared (Vis-NIR-SWIR) range using a FieldSpec 3 spectroradiometer (Analytical Spectral Devices, ASD, Boulder, CO), which operates in the spectral range from 350 to 2500 nm;

f) The sensor had a spectral resolution of 3 nm from 350-700 nm and 10 nm from 700-2500 nm, automatically interpolated to 1 nm spectral resolution in the output data, resulting in 2151 channels (or bands); and

g) We positioned the lamps at 90° from each other and 35 cm away from the sample, with a zenith angle of 30°.

The sensor captured the light reflected through the fiber optic cable, which was positioned 8 cm from the sample's surface.

We used two 50W halogen lamps as the power source for the artificial light. It's important to note that we took three readings for each sample at different positions by rotating the Petri dish by 90°.

Each reading represents the average of 100 scans taken by the sensor. From these three readings, we calculated the final spectrum of the samples. Notably, the laboratory's equipment and procedures for soil sample spectral analyses followed the ASD's recommendations, particularly about sensor calibration using a white spectralon plate as a 100% reflectance standard.

For the analysis in the Middle Infrared (MIR) spectral region, we followed the procedures outlined by Mendes et al. (2022). We milled the soil fraction smaller than 2 mm, sieved it to 0.149 mm, and scanned it using a Fourier Transform Infrared (FT-IR) alpha spectroradiometer (Bruker Optics Corporation, Billerica, MA 01821, USA) equipped with a DRIFT accessory.

The spectroradiometer measured the diffuse reflectance using Fourier transformation in the spectral range from 4000 cm^-1 to 600 cm^-1, with a resolution of 2 cm^-1. We conducted these measurements in the Geotechnology Laboratory of the Department of Soil Science at Esalq-USP. We took the average of 32 successive readings to obtain a soil spectrum. Sensor calibration took place before each spectral acquisition of the sample set by standardizing it against the maximum reflectance of a gold plate.

Dataset characterization

The database, named BSSL_DB_Key_Soils, has five sheets containing the key soil attributes, Vis-NIR-SWIR and MIR datasets, descriptions of the contributors and the proximal sensing methods used for spectral soil analysis. The sheets can be linked by "ID_Unique" columns, which bring the corresponding rows according to the data type. Some cells are empty because collaborators have already provided data in this way. However, we have decided to keep them in the database because they have other soil key attributes. Every Column in the data sheets is described as follows:

Sheet 1.       BSSL_Soil_Attributes_Dataset

Column 1.    ID_unique: Sequential code assigned to every record;

Column 2.    Owner code: Acronym assigned to each contributor who allowed access to their proprietary data;

Column 3.    Vis_NIR_SWIR_availability: availability of spectral data in visible, near-infrared, and shortwave infrared ranges;

Column 4.    MIR_availability: availability of spectral data in the middle infrared range;

Column 5.    Sampling: type of soil sampling;

Column 6.    Depth_cm: soil surface layer depth in centimeters;  

Column 7.    Lat: Latitude;  

Column 8.    Lat: Longitude;  

Column 9.    Region: Brazilian geographical region of samples' source;

Column 10.    Municipality: Brazilian municipality of samples' source;

Column 11.   State: Brazilian Federation Unit of samples' source;

Column 12. Vegetation: type of vegetal covering;

Column 13. Biome: groupings of ecosystems that share similar characteristics and span different regions;

Column 14. Geology: type of rock matter from local soil sampling;

Column 15. Sand_gkg: Content of the soil fraction with grain size between 2 and 0.053 mm, expressed in grams per kilogram;

Column 16. Clay_gkg: Content of soil fraction with grain size smaller than 0.002 mm, expressed in grams per kilogram;

Column 17. SOM_gkg: Soil organic matter content, expressed in grams per kilogram;

Column 18. pH_H2O: Soil hydrogen ion potential measured in water;

Column 19. Ca_mmolkg: Exchangeable calcium content in the soil, expressed in millimoles per kilogram;

Column 20. Mg_mmolkg: Exchangeable magnesium content in the soil, expressed in millimoles per kilogram;

Column 21. Na_mmolkg: Exchangeable sodium content in the soil, expressed in millimoles per kilogram; and

Column 22. CEC_Ph7_mmolkg: Cation exchange capacity of the soil at neutral pH, expressed in millimoles per kilogram.

Sheet 2.     BSSL_Vis_NIR_SWIR_Dataset

Column 1. ID_Unique: Sequential code assigned to every record;

Column 2. Owner code: Acronym assigned to each contributor who allowed access to their proprietary data; and

Column 3 – 2153. 350 – 2500: Reflectance in 2151 spectral bands in nanometers from visible and near-infrared to shortwave infrared range (350 – 2500 nm).

Sheet 3.    BSSL_MIR_Dataset

Column 1. ID_Unique: Sequential code assigned to every record;

Column 2. Owner_code: Acronym assigned to each contributor who allowed access to their proprietary data; and

Column 3 – 683. 4000 – 600: Reflectance in 681 spectral bands in centimeters in the middle infrared range (4000 – 600 cm^-1).

Sheet 4.       Contributors

Column 1. Owner_code: Acronym assigned to each contributor who allowed access to their proprietary data, which identifies and links it to datasets;

Column 2. Owner: Name of the collaborator who agreed to the availability of the data;

Column 3. E-mail: Contact the e-mail of the owner for more information or a data request;

Column 4. Institution: Contributor's affiliation;

Column 5. Samples NIR: Number of Vis-NIR-SWIR samples sent to the BSSL collection;

Column 6. Samples MIR: Number of MIR samples sent to the BSSL collection;

Sheet 5.       Metadata

Column 1. Material and Methods: Description of procedures performed for soil data analyses

Expectation and Social Relevance

These data can impact various disciplines such as soil surveying, soil attribute mapping, soil analysis, soil mineralogy, soil management zones, precision agriculture, development of new datasets and scientific groups, and others. We expect this contribution to be valuable and useful to the soil research community in promoting this non-renewable natural resource's conservation and sustainable use.