Based on three-dimensional first-order derivative fluorescence spectrometry, an analysis method for detecting dissolved organ matter in water is proposed in the present paper. By using simplified least squares differentiation methods presented by Savitzky and Goly, the first-order partial derivatives for emission wavelength and excitation wavelength were calculated. As the fitting polynomial has the smoothing function in the calculation of derivative spectra, a separate smoothing method is not required to remove spectrometry noise. The regression model was calculated by partial least square for 4-dimension fluorescence data including emission wavelength, excitation wavelength and their first-order derivatives. The Experimental results for detecting total organic carbon (TOC) in water show that the proposed method has obvious advantage over the conventional fluorescence spectrometry analysis methods in the aspect of the root mean square error of prediction and correlation coefficient.