Four potential interference factors involved in colorimetric determination of dissolved oxygen were studied. Those factors include the concentrations of oxygen and iodide in the sample and reagent, the wavelength and temperature at which the detection is made. Evaluations were made both theoretically and experimentally on the linearity of the calibration curve. The results show that concentrations of dissolved oxygen and iodide cause variation of the ratio between molecular iodine and tri-iodide, thus leading to a bending phenomenon of the calibration curve. Detection made at near the isobestic point (466~469 nm) may reduce the above interference at the cost of losing up to 30% of the sensitivity. The colorimetry can be carried out at 456 nm but it is important to ensure a high strength of iodide reagent (e.g. the final strength = 33 mM) so as to minimize the bending effect. The temperature of the sample is found essential because it alters the spectrum of the final color and causing a shift of the isobestic point (ca. 0.5 nm per 5 oC). The shifting is more significant at a relatively higher wavelength. In conclusion, when the final temperature is to be controlled within a ±1 oC error and the iodide reagent strength is sufficient, the colorimetric approach can provide a precise and reliable tool for the determination of oxygen at the naturally occurring concentration range (up to 300 μM) in the ocean environment.