ABSTRACT
Surface sediment concentration (SSC) is linked to several problems related to water quality and its monitoring is costly because of the required fieldwork and laboratory analyses. Thus, sediment measurements are often sporadic, punctual, and performed during a short period. Orbital remote sensing allows the monitoring of SSC along the river channel permitting continuous and spatial information. This work had two objectives: (1) to model the surface concentration of sediments in the main channel of the Doce river using data from Multispectral Instrument (MSI)/Sentinel 2 and Operational Land Imager (OLI)/Landsat 8 satellite sensors; and (2) to compare different linear modeling approaches to select the best variables for SSC monitoring. For comparison with actual field data, we used mean SSC measurements in 14 sediment gauge stations from 2013 to 2020. Reflectance data of the MSI/Sentinel 2 and OLI/Landsat 8 satellites bands and spectral indices related to the monitoring of water resources were used as explanatory variables. Simple and multiple linear regression models (SLR and MLR), least absolute shrinkage and selection operator (LASSO), and Elastic Net regression were used to predict the SSC. The near-infrared band images from both MSI/Sentinel 2 and OLI/Landsat 8 satellites showed a strong linear relationship with the SSC. Multiple linear regression, LASSO and Elastic Net regressions showed good performance for SSC prediction. Sediment flow maps show an SSC reduction in the Doce river channel in recent years, indicating that part of the material from the Fundão tailings dam rupture may have been transported through sediment resuspension and transport processes.
Subject(s)
Environmental Monitoring , Rivers , Environmental Monitoring/methods , Remote Sensing Technology , Water QualityABSTRACT
RESUMO Este estudo avaliou o potencial poluidor da bacia de contribuição do reservatório de Funil (BCRF), localizado na bacia hidrográfica do rio Paraíba do Sul, considerando a geração da carga de nutrientes, nitrogênio (N) e fósforo (P), por fontes pontuais e difusas, a partir de uma modelagem distribuída utilizando Sistema de Informação Geográfica (SIG). As cargas e concentrações médias anuais desses nutrientes foram geradas a partir do acoplamento de equações empíricas, em SIG, considerando informações espaciais de uso e cobertura do solo, população residente na bacia e vazão média anual de longo período, obtida por equações do tipo chuva vazão. Os resultados indicaram que 80% da carga total de nitrogênio foram provenientes de fontes pontuais e 20% de fontes difusas, enquanto que, da carga total de fósforo, 89,1% foram originadas de fontes pontuais e 10,9% de fontes difusas. As concentrações de nutrientes estimadas pelo modelo empírico apresentaram bons ajustes em relação aos valores observados de fósforo e de nitrogênio no rio Paraíba do Sul, com R²=0,96 (p<0,01) e R²=0,70 (p<0,01), respectivamente. Dessa forma, o modelo foi capaz de detectar, de forma significativa, a tendência das variações nas concentrações de nutrientes ao longo de diferentes trechos da BCRF.
ABSTRACT This study evaluated the potential polluter of the Funil Reservoir Contribution Basin (BCRF), located in the Paraíba do Sul River basin, considering the generation of nutrient loading, nitrogen (N) and phosphorus (P), due to point and diffuse sources from a distributed modeling using Geographic Information System (GIS). Loads and annual average concentrations of these nutrients were generated from the coupling of empirical equations, in GIS, considering spatial information such as land use/cover, population living in the basin and long period average annual flow obtained by equations of rainfall runoff. The results indicated that 80% of the total load of nitrogen was generated from point sources and 20% from diffuse sources, while 89.1% were originated from point sources and 10.9% from diffuse sources for the corresponding total load of phosphorus. The model estimated adequately the concentration when compared to the observed values and it was able to detect the trend of changes in nutrient concentrations along different sections of the BCRF highlighting significant correlations between the observed and simulated concentrations of phosphorus and nitrogen with R²=0.96 (p<0.01) and R²=0.70 (p<0.01), respectively.
ABSTRACT
No presente trabalho, desenvolveu-se equação para predição do coeficiente de dispersão longitudinal (D L) em rios de médio porte, com vazões entre 16,20 e 98 m³ s-1, a partir de ensaios experimentais utilizando traçadores fluorescentes. A equação foi deduzida utilizando-se análise dimensional e ajustada aos dados de campo pela técnica de regressão linear múltipla. Em seguida, realizou-se a validação da equação testando-a em outra base de dados, diferente daquela para a qual foi desenvolvida, com vazões entre 19,57 a 48,54 m³s-1. Adicionalmente, foram comparados os desempenhos de outras quatro equações empíricas de predição do D L, utilizando-se três métodos comparativos. A equação desenvolvida apresentou ajuste estatístico adequado e bom resultado, com melhor desempenho geral quando comparada a outras equações propostas na literatura.
In this paper, an equation has been developed to estimate the longitudinal dispersion coefficient (D L) in medium-sized rivers, with flow between 16.20 and 98 m³s-1, from experimental tests using fluorescent tracers. The equation was deduced by dimensional analysis and adjusted to the field data by the multiple linear regression technique. Thereafter, the validation of the equation was performed by testing it in another database, which is different from the one it had been developed for, with flow between 19.57 and 48.54 m³s-1. Additionally, the performance of four other empirical equations was compared for D L prediction using three comparative methods. The developed equation showed appropriate statistical adjustment and good result, with better overall performance when compared with other equations proposed in literature.
