Abstract: In order to meet the requirements of rapid, accurate and continuous online detection of the water acidity (pH), a method for rapid detection of the water acidity was proposed based on visible-near infrared spectrometry (Vis-NIRS) technique in combination with chemometric methods. The original Vis-NIRS data were collected from 60 water solution samples with varying acidities. These samples sets were split by Kennard-Stone (K-S) algorithm and sample set partitioning based on joint X-Y distance (SPXY) algorithm. The original spectral data were pre-processed by the methods including Savitzky-Golay (S-G) convolution smoothing, standard normal variate (SNV), first derivative (1D), second derivative (2D) and orthogonal signal correction (OSC). Successive projections algorithm (SPA) and competitive adaptive reweighted sampling (CARS) algorithm were used for characteristic wavelength screening. Different partial least squares (PLS) quantitative analysis models were established and compared to determine the effect of the best model. As shown by the results, the SPXY algorithm for sample set partition, SNV for spectral data pre-processing, and CARS for characteristic wavelength screening produced better performance for water acidity PLS quantitative analysis model. The prediction set had a coefficient of determination of 0.978 6 and a root mean square error of 0.380 3, and the variable number of wavelengths involved in modeling was reduced from 2 860 to 45, greatly speeding up model calculation. The proposed method could achieve rapid and accurate water acidity detection.