ESTIMATION OF RIVER WATER QUALITY USING DIFFENTIAL ULTRAVIOLET-VISIBLE SPECTRA BASED ON PARTIAL LEAST SQUARES REGRESSION

抄録

 The collection of river water quality data with high frequency and low cost is demanded for better environmental management of rivers, lakes and reservoirs. Although the feasibility of measuring nitrate, organics and suspended solids using in situ spectrometer has been demonstrated in previous researches, the calibration accuracy of phosphorous and some other dissolved nutrients is yet to be improved and the applicability to turbid water during floods has to be examined. For this purpose, we applied UV-Vis spectroscopy to rivers in the Kasumigaura watershed to estimate the nutrient concentrations. Water samples were collected from the study rivers and their water quality and spectra were analyzed in laboratory. Partial least squares (PLS) regression was used to develop models with which chemical oxygen demand, nitrate, nitrite, total nitrogen, total phosphorous and total suspended solids can be estimated from UV-Vis specta data. The difference method applied to raw spectra data was proved to eliminate the effect of spectra baseline and increase signal to noise ratio. Regression equations between concentrations of these water quality parameters and raw or differential spectra data were established and the results were assessed by the determination coefficients. Calibration result of phosphorous based on PLS regression was improved a lot by using differential spectra data. This study shows the potential to measure multiple components of water quality using both raw and differential UV-Vis spectra data based on PLS regression, and provides theoretical basis for high frequency and low cost water quality monitoring using in situ UV-Vis spectrometers.