Dask dataframe apply with manipulation on time-based dtypes

Question

Say I have the below data, where the timestamp b is always after a

numb  a                     b                     c
40583 2020-08