Monitoring water quality using proximal remote sensing technology.

Accurate, high spatial and temporal resolution water quality monitoring in inland waters is vital for environmental management. However, water quality monitoring in inland waters by satellite remote sensing remains challenging due to low signal-to-noise ratios (SNRs) and instrumental resolution limitations. We propose the concept of proximal remote sensing for monitoring water quality. The proximal hyperspectral imager, developed by Nanjing Institute of Geography and Limnology, Chinese Academy of Sciences (CAS) and Hikvision Digital Technology, Ltd., is a high spatial, temporal and spectral resolution (1 nm) sensor for continuous observation, allowing for effective and practical long-term monitoring of inland water quality. In this study, machine learning and empirical algorithms were developed and validated using in situ total nitrogen (TN), total phosphorus (TP), chemical oxygen demand (COD) concentrations and spectral reflectance from Lake Taihu (N = 171), the Liangxi River (N = 94) and the Fuchunjiang Reservoir (N = 109) covering different water quality. Our dataset includes a large range for three key water quality parameters of TN from 0.93 to 6.46 mg/L, TP from 0.04 to 0.62 mg/L, and COD from 1.32 to 15.41 mg/L. Overall, the back-propagation (BP) neural network model had an accuracy of over 80% for TN (R2 = 0.84, RMSE = 0.33 mg/L, and MRE = 11.4%) and over 90% for TP (R2 = 0.93, RMSE = 0.02 mg/L, and MRE = 12.4%) and COD (R2 = 0.91, RMSE = 0.66 mg/L, and MRE = 9.3%). Our results show that proximal remote sensing combined with machine learning algorithms has great potential for monitoring water quality in inland waters.

Empirical Formulas for Estimating Backscattering and Absorption Coefficients in Complex Waters from Remote-Sensing Reflectance Spectra and Examples of Their Application.

Many standard methods used for the remote sensing of ocean colour have been developed, though mainly for clean, open ocean waters. This means that they may not always be effective in complex waters potentially containing high concentrations of optically significant constituents. This paper presents new empirical formulas for estimating selected inherent optical properties of water from remote-sensing reflectance spectra Rrs(λ), derived, among other things, for waters with high concentrations of dissolved and suspended substances. These formulas include one for estimating the backscattering coefficient bb(620) directly from the magnitude of Rrs in the red part of the spectrum, and another for estimating the absorption coefficient a(440) from the hue angle α. The latter quantity represents the water's colour as it might be perceived by the human eye (trichromatic colour vision); it is easily calculated from the shape of the Rrs spectrum. These new formulas are based on a combined dataset. Most of the data were obtained in the specific, optically complex environment of the Baltic Sea. Additional data, taken from the NASA bio-Optical Marine Algorithm Dataset (NOMAD) and representing various regions of the global oceans, were used to widen the potential applicability of the new formulas. We indicate the reasons why these simple empirical relationships can be derived and compare them with the results of straightforward modelling; possible applications are also described. We present, among other things, an example of a simple semi-analytical algorithm using both new empirical formulas. This algorithm is a modified version of the well-known quasi-analytical algorithm (QAA), and it can improve the results obtained in optically complex waters. This algorithm allows one to estimate the full spectra of the backscattering and absorption coefficients, without the need for any additional a priori assumptions regarding the spectral shape of absorption by dissolved and suspended seawater constituents.

Mapeo de la dinámica regional de la transparencia en aguas continentales usando productos de reflectancia MOD09GA / Mapeamento da dinâmica regional de transparência nas águas interiores utilizando produtos de reflectância MOD09GA / Mapping of regional transparency dynamics on inland waters using reflectance products MOD09GA

Resumen La transparencia del agua es un indicador de la calidad de este recurso, y ha sido demostrado que es posible estimar en imágenes generadas por plataformas orbitales la profundidad medida con el Disco de Secchi, con el uso de algoritmos empíricos. Los modelos empíricos correlacionan la intensidad del píxel capturado por la imagen en la banda centrada en los 645nm y el Disco de Secchi. Este trabajo investigativo se desarrolló en el embalse de Porto Primavera, localizado en la frontera que divide los estados de São Paulo, Matto Grosso do Sul e Paraná. Para calibrar y validar el modelo de correlación (reflectancia de superficie Vs Mediciones del Disco de Secchi) se realizaron dos campañas de campo (octubre de 2014 y abril de 2015). El modelo de calibración encontrado muestra que existe un coeficiente de determinación que puede ser considerado como adecuado para la estimación del DS (R2 = 0.8) y los errores asociados con las observaciones también son bajos (RMSE = 0,47m). La dinámica espacial del Disco de Secchi muestra que el embalse de Porto Primavera presenta los menores valores para esta variable en la margen situada en el estado de Matto Grosso do Sul, y en la temporada de otoño del hemisferio austral esta condición puede extenderse hasta la parte norte del lago.

Abstract The water transparency is an indicator of the quality of this resource, and it has been demonstrated that it is possible to estimate the depth measured with the Secchi Disk in generated images by orbital platforms with the use of empirical algorithms. Empirical models correlate the intensity of the pixel captured by the image in the 645nm band and Secchi Disk measurements. This investigative work was carried out in the Porto Primavera reservoir located on the border that divides São Paulo, Matto Grosso do Sul and Paraná states. In order to calibrate and validate the correlation model (surface reflectance Vs Secchi Disk measurements) two field sampling campaigns were carried out (October 2014 and April 2015). The calibration model found shows that there is a coefficient of determination that can be considered adequate for the estimation of Secchi Disk (R2 = 0.8), the errors associated with these observations were also low (RMSE = 0.47m). The spatial dynamics of the Secchi Disk show that the Porto Primavera reservoir presents the lowest values for this variable in the margin located in the state of Matto Grosso do Sul and in the autumn season of the southern hemisphere this condition can extend to the northern part of the lake.

Resumo A transparência da água é um indicador da qualidade deste recurso, e foi demonstrado que é possível estimar a profundidade medida com o Disco Secchi em imagens geradas por plataformas orbitais, com o uso de algoritmos empíricos. Os modelos empíricos correlacionam a intensidade do pixel capturado pela imagem na banda centrada nos 645nm e o Disco Secchi. Este trabalho de pesquisa foi realizado no reservatório Porto Primavera, localizado na raia divisória dos estados de São Paulo, Matto Grosso do Sul e Paraná. Para calibrar e validar o modelo de correlação (reflectância superficial Vs Secchi Disk Measurements), foram realizados dois trabalhos de campo (outubro de 20l4 e abril de 20l5). O modelo de calibração encontrado mostra que existe um coeficiente de determinação que pode ser considerado adequado para a estimativa do DS (R2 = 0,8) e os erros associados às observações também são baixos (RMSE = 0,47 m). A dinâmica espacial do Disco de Secchi mostra que o reservatório do Porto Primavera apresenta os valores mais baixos desta variável na margem localizada no estado de Matto Grosso do Sul e, na época de outono do hemisfério sul, esta condição pode ser estendida à ao norte do lago.

Time-series MODIS image-based retrieval and distribution analysis of total suspended matter concentrations in Lake Taihu (China).

Although there has been considerable effort to use remotely sensed images to provide synoptic maps of total suspended matter (TSM), there are limited studies on universal TSM retrieval models. In this paper, we have developed a TSM retrieval model for Lake Taihu using TSM concentrations measured in situ and a time series of quasi-synchronous MODIS 250 m images from 2005. After simple geometric and atmospheric correction, we found a significant relationship (R = 0.8736, N = 166) between in situ measured TSM concentrations and MODIS band normalization difference of band 3 and band 1. From this, we retrieved TSM concentrations in eight regions of Lake Taihu in 2007 and analyzed the characteristic distribution and variation of TSM. Synoptic maps of model-estimated TSM of 2007 showed clear geographical and seasonal variations. TSM in Central Lake and Southern Lakeshore were consistently higher than in other regions, while TSM in East Taihu was generally the lowest among the regions throughout the year. Furthermore, a wide range of TSM concentrations appeared from winter to summer. TSM in winter could be several times that in summer.