Dara, it might be worth looking at the Laplacian divided by dT/dt. This should help highlight external forces, since it is normalizing for the effect of the heat equation. This will probably be noisy when dT/dt is very small.

Maybe the bettter would be to first estimate the a linear model of the Laplacian vs dT/dt and then visualize the residual error.

Maybe the bettter would be to first estimate the a linear model of the Laplacian vs dT/dt and then visualize the residual error.