Dinâmica de atributos limnológicos e uso de sensoriamento remoto para estimar as concentrações de clorofila-a em reservatório no Semiárido

Abstract

Continental aquatic systems have a direct function of meeting the high demand of water for the population and the most diverse uses in society. In the case of reservoirs in the semi-arid region, the quantitative and qualitative factors become relevant in any social sphere. In this context, the reduction of the volume of water stored in the reservoir of the semi-arid region deteriorates the quality of the water. Thus, when remote sensing is used from optical sensors as a management tool to remotely obtain concentrations of chlorophyll in water bodies, it is possible to reduce costs and create a vast database. In this way, the objective was to evaluate the qualitative status of the waters of a superficial artificial reservoir, to classify it according to its trophic state and to quantify the concentrations of chlorophyll a using remote sensing data. The study was carried out at the Pereira de Miranda Reservoir, Pentecoste, Ceará. The campaigns were conducted quarterly, between April 2015 and September 2016, in five sample points, geo-referenced. The limnological attributes analyzed were Chlorophyll-a, pH, EC, total nitrogen, total phosphorus, transparency, fixed, volatile and total suspended solids and temperature. In situ radiometric data collection was obtained by the hyperspectral sensor, FieldSpec® 3 Hi-Res ASD, with a 25º field of view, a spectral resolution of 1,4 nm and covering the spectral range from 350 to 2500 nm. For the limnological attributes, the multivariate statistical model, Principal Component Analysis (PCA), was applied with the objective of identifying the determinants of water quality. With in situ radiance data, the spectral responses of the OLCI / Sentinel-3 sensor bands and parameterization of the regression models for remote retrieval of chlorophyll-a concentrations were simulated. 75% of the data were used for the parameterization step, and then the models were validated using the other 25%. The limnological attributes chlorophyll a, total N and total P had concentrations above the recommended limit for human supply. The waters of the Pereira de Miranda reservoir are highly trophic, that is, eutrophic. The main component analysis technique allowed the selection of three components indicating the quality of surface water, explaining 82.34% of the total variance. Variations in water quality were defined by a group of suspended solids, a group of nutrients, soluble salts and an organic group. Two- and three-band models (8 and 11) and (8, 11 and 12) with respective wavelengths of (665, 708.75 and 761.25 nm) were developed. The two- and three-band models showed coefficients of determination (R²) of 0.71 and 0.82, respectively. The models, in the validation stage, presented high determination coefficients (R²), 0.91 and 0.85, strong correlation coefficients (r), 0.95 and 0.92, and NSE of 0.85 and 0.82, respectively, for the two- and three-band models. The good performances of the models using the simulated data show the feasibility of remotely estimating chlorophyll-a concentrations in surface semi-arid reservoirs from data from the OLCI / Sentianel-3 orbital sensor.