InGaAs NIR photodetectors are widely used due to their high responsivity, low noise, low dark current, fast response time, and large spectral range, which covers a range of 900 to 1700 nm and can be extended up to a 2600 nm cutoff. However, thermal drift is a major challenge that can affect the responsivity of these photodetectors, especially in miniaturized systems, where the thermal management problem is challenging. InGaAs photodetectors exhibit a highly nonlinear increase in responsivity near the cutoff, with an increase of about 4%/°C and a nonlinear reduction of about 0.25%/°C in the middle of the spectral range. This nonlinear drift cannot be corrected by common pre-processing methods used in spectroscopy. The change in responsivity near the cutoff is due to thermal-assisted bandgap reduction, while the change in responsivity in the middle of the spectral range is not well described in the literature. To address this issue, we applied the Urbach tail formula to model the nonlinear reduction of responsivity in the middle of the spectral range. The model showed an accuracy of approximately 90% compared to experimental thermal drift, allowing us to deduce the root causes of this phenomenon. Finally, we proposed compensation methods for the thermal drift, which were investigated using MOEMS FTIR spectral sensors as a case study. Some of these methods successfully reduced the drift that occurs due to a 60°C temperature change to less than 3%.