That sounds good. You don't think there will be much electrical energy generated at turn-off by the pump as it runs down? I guess the run-down would be pretty short and not able to generate much power?
Immediately after the relay opens the motor armature will still be rotating. The motor will act as a generator (assuming it's a permanent magnet motor), generating a voltage just about equal to the applied voltage. The polarity of this voltage will match the applied 12 volts--in other words, it will be positive at the positive terminal of the motor. This voltage therefore will produce essentially no voltage across the relay contacts, and it will therefore cause no arcing at the relay contacts and doesn't need to be suppressed.
On the other hand, the armature does possess some inductance. When the relay contacts open, that inductance will produce a negative voltage spike at the positive terminal of the motor. Its duration will depend on the distributed capacitance of the armature winding and other parasitics.
So, when the relay contacts open there will be a very short, relatively high voltage, negative going spike of voltage at the positive terminal of the motor and this spike will cause arcing at the relay contacts if it isn't suppressed. The relatively low voltage due to the spinning of the armature can be ignored.
Are you suggesting this across the pump?
How can you estimate the amount of energy that the pump will generate at turn-off due to its inductive behaviour and its run-down? For surge durations short enough that the burst of heat can be dissipated, the tranzorb's limits would be in terms of peak instantaneous power, and amount of energy, but the only dissipation rating stated seems to be 1500W peak dissipation using a standard 10/1000 us surge waveform.
To estimate the energy of the voltage spike we would need to know the apparent inductance of the armature; it varies somewhat as the armature turns, The "run-down" energy doesn't need to be suppressed as I explained above.
The OP's motor is fairly large--20 amps at 12 volts is 240 watts which suggests to me a motor of about 1/4 horsepower. I don't have a DC motor that large, but I do have a small motor and I measured its armature inductance as about 1 mH. If the OP's motor had a 1 mH armature inductance, the energy stored in the inductance would be 1/2 * L * I^2 = 1/2 * .001*400 = .2 joules of energy at a rate less than 1500 watts, well within the capability of a transorb (assuming the time constant of the inductance isn't too long; a measurement really should be made, but I would guess it would be ok). Even with an inductance of several millihenries a single transorb of the size in the datasheet I linked could handle the spike. Larger transorbs are available, or several lower voltage units of the ones I referenced could be connected in series for more energy handling capability.