Three dimensional fluorescence spectroscopy combined with multivariate calibration analysis for petroleum pollutants multicomponent determination method has problems of complex spectra aliasing, and is susceptible to blank fluorescence and interference fluorescence reducing the accuracy of result. Temperature information as one dimension added to the traditional three dimensional fluorescence spectrum to construct excitation wavelength-emission wavelength-temperature-samples four dimensional fluorescence spectrum data array (excitation-emission-temperature-sample data array, EEM-temperature data array) was proposed, and four linearity component model combined with high dimensional fluorescence spectrum qualitative and quantitative analysis was applied. Experiments demonstrate that fluorescence spectral shape of mineral oil does not change with the change of temperature in 15 to 25 ℃ range, but the intensity changes linearily, satisfying the requirement of four linearity, providing possibility for developing four dimensional fluorescence spectra with third-order correction to extract more useful information from the high dimensional data. The third order correction not only has “second order advantage”, namely to quantitatively determine interesting constituents in the presence of interferences, but also has higher selectivity and sensitivity, higher resolution ability for colinearity and background interference of the overlapping spectra, namely, “third order advantage”. The complex petroleum pollution system composed of 0# diesel, 97# gasoline and engine oil as components to be determined and humic acid as water interference component are experimental samples. The parallel factor (PARAFAC) algorithm and the alternating penalty trilinear decomposition (the alternating penalty trilinear decomposition, APTLD) algorithm are applied to the three dimensional fluorescence spectra for the second-order calibration analysis; the four dimensional fluorescence spectra data array containing temperature information is constructed by stacking three dimensional fluorescence spectra along temperature direction dimension, and is analyzed by four dimensional parallel factor algorithm (4-PARAFAC) and alternating penalty quadrilinear decomposition (alternating penalty quadrilinear decomposition, APQLD) for third-order correction analysis. The prediction results of 0# diesel, 97# gasoline and engine oil are compared and show that the four way fluorescence spectrum with adding the affecting factor of temperature increases the extraction ability for effective information and the four dimensional fluorescence spectroscopy combined with high-order correction algorithm can improve the oil spectrum recognition and concentration detection precision and improve the recovery rate and the root mean square prediction error (root mean square error of prediction, RMSEP) compared with the traditional three dimensional fluorescence spectrum analysis, advantageous to the effective, accurate, real-time, green environmental detection for petroleum pollutants. At the same time, the characteristics of 4-parafac and APQLD algorithms and their different applicable environments are pointed out, which can provide a basis for the algorithm selection for the detection of oil pollutants. The four-dimensional fluorescence spectra with introduction of the temperature parameters combined with third-order correction algorithm detection technology, no matter in qualitative resolution of constituent spectra or quantitative concentration determination of the complex system of oil pollution compared with three-dimensional fluorescence spectrum technology, is capable of realizing fast, effective, green and pollution-free detection, thus “mathematical separation” can replace “chemical separation” more effectively.