Under the assumption that a gradient system is linear and time-invariant (LTI), accurate gradient field waveforms can be predicted by gradient response functions. However, time-invariance can be violated due to heating of system components. Temperature sensors can be used to assess heating of the gradient coils. To assess the predictability of gradient response function based on temperature measurements, the temperature dependence of gradient response functions is analyzed using an NMR probe based field camera and optically connected temperature sensors. From this data a prediction model is generated and tested for its application in image reconstruction.