Program disturb is reduced in a non-volatile storage system during a programming operation by switching from using programming pulses of a longer duration to programming pulses of a shorter duration, partway through the programming operation. A switchover point can be based on temperature, selected word line position and/or tracking of storage elements to a trigger state. The switchover point occurs sooner for higher temperatures, and for drain side word lines. The trigger state can be selected based on temperature. A portion of storage elements which are required to reach the trigger state to trigger a switchover can also be set a function of temperature. Programming pulses of a shorter duration improve channel boosting for inhibited storage elements, thereby reducing program disturb for these storage elements.