Dados do Trabalho

Título

BRAZILIAN SOIL SPECTRAL SERVICE (BRASPECS): AN ONLINE SOIL ANALYSIS SYSTEM SUPPORTING SOIL HEALTH MONITORING

Resumo

Investigate the condition of soils is essential to keep them healthy and environmentally functional. However, soil mapping over large areas such as the Brazilian territory, the fifth largest country in the world, is challenging because it is labor-intensive, time-consuming, costly and wasteful. Technologies such as the Brazilian Soil Spectral Service (BraSpecS), an online system based on soil spectroscopy (Vis-NIR-SWIR range), help in this task by providing a time-efficient soil survey with low environmental impact. Therefore, we aimed to evaluate the performance of BraSpecS in modeling data from distinct regions of Brazil for soil health monitoring purposes. Thus, we invited seven users from north to south to submit their spectra to this service and return the predicted soil attribute analysis results (carbon, clay and ferric oxide content) to compare with traditional measured data using the cubist model. We expected the models to work similarly for every user, regardless of the sample's origin. The mean R2 of 0.62 for carbon, 0.69 for clay and 0.58 for ferric oxide content and the metrics, RMSE and RPQI, for clay, carbon and ferric oxide were around 40.65 and 4.92, 5.71 and 1.91, 1.55 and 18.91, respectively. The results were consistent and they can assist soil scientists with soil survey and mapping because they are in line with the field, considering the empirical knowledge. This technique has the advantage of rapid results through an online service, using data from a sensor in the laboratory or directly obtained in the field. Its main limitation is the high cost of the spectrometer, which is out of reach for most users. The system must therefore be able to use data from different types of instruments, including multispectral sensors, according to their wavelength. The model's improvement and the insertion of other sensors data are alternatives to support large-scale soil health monitoring.

Palavras-chave

reflectance spectroscopy, proximal sensing, pedometrics, soil mapping, soil health.

Instituição financiadora

The State of São Paulo Research Foundation (FAPESP) 2021/05129-8.

Agradecimentos

The authors thank GeoCis Research Group and BraSpecS members for collaboration.