When the temperature is low, the molecules move slowly and as a result their collisions are weak and infrequent. This allows the molecules to stay mostly aligned with the electric field. 

When the temperature is high, the molecules move much faster and so they collide more frequently and with greater intensity. These collisions "knock" molecules out of alignment with the electric field.

If the temperature is high enough, then the time it takes for the electric field to align the molecule can be much longer than the average time between violent collisions. In this case we do not expect much net alignment at all. This appears to be the case with water at room temperature, where even in large electric fields the average dipole moment is a small fraction of a water molecule's permanent dipole moment. 

We would expect then that the polarizability of water is lower at higher temperatures, since the average dipole moment for a given electric field will be smaller. Since the relative permittivity decreases with decreasing polarizability, we then would also expect the relative permittivity to decrease with increasing temperature. This is indeed what we see from the plot of the measured permittivity vs. temperature.