Ultraviolet (UV) absorption spectroscopy is used to detect the concentration of water chemical oxygen demand (COD). The UV absorption spectra of COD solutions are analyzed qualitatively and quantitatively. The partial least square (PLS) algorithm is used to model COD solution and the modeling results are compared. The influence of environmental temperature and turbidity is analyzed. These results show that the influence of temperature on the predicted value can be ignored. However, the change of turbidity can affect the detection results of UV spectra, and the COD detection error can be effectively compensated by establishing the single-element regression model.