Abstract: Tritium metrology is critical for nuclear material accountability and safeguards, as accurate knowledge of tritium inventory is fundamental to both proliferation resistance and operational safety. Calorimetry offers a nondestructive and direct approach for tritium quantification by measuring the decay heat power. Among various calorimeter configurations, the isothermal differential calorimeter is widely employed due to its stability and simplicity. However, under working conditions, ambient temperature fluctuations are unavoidable, and such non-ideal thermal environments introduce systematic measurement deviations. Therefore, understanding how temperature disturbances affect calorimetric performance and establishing a practical correction strategy are of significant importance for improving measurement reliability. Box-Behnken response surface methodology was selected to design a three-factor, three-level experiment. The independent variables investigated were temperature variation amplitude, temperature variation duration, and sample power. The response variable was the relative measurement deviation of the calorimeter. A quadratic polynomial regression model was then fitted to the experimental data to quantify the main effects, interaction effects, and quadratic effects of the three factors on the deviation. Model adequacy was verified through analysis of variance (ANOVA) and residual diagnostics. The results reveal that temperature variation amplitude is the most influential factor and exhibits a positive effect: As the amplitude increases, the absolute value of the measurement deviation increases. In contrast, sample power exerts a negative effect; Higher-power samples, which possess larger heat capacity and consequently greater thermal inertia, effectively dampen the impact of external temperature fluctuations and thus significantly reduce the deviation. The duration of temperature variation, while less influential individually, shows a notable coupling effect with amplitude: Longer duration combined with larger amplitude exacerbates the original deviation, indicating that the thermal history of the environment cannot be neglected. Based on these findings, a three-dimensional coupled correction formula incorporating temperature amplitude, duration, and sample power was derived. The formula takes the form of a polynomial function that directly maps the measured raw deviation to a corrected value. Validation experiments conducted within the same parameter space demonstrate that the correction method reduces all deviations that initially exceed the precision requirement to below 1%, meeting the typical accuracy threshold for tritium accounting. In conclusion, the proposed correction approach provides an effective and practical engineering solution for isothermal differential calorimeters operating under non-ideal temperature conditions. It requires no hardware modification and relies only on recording parameters during measurement. Future work will focus on expanding the correction formula library by accumulating more calorimetric data under diverse disturbance patterns, including cyclic and stepwise temperature changes, and extending the method to other calorimeter types. With further refinement, this strategy will enhance the traceability and accuracy of tritium calorimetry in routine safeguards.