Applying a function on a groupby table that depends on the value of another column and returns all other columns intact

Question

I have a dataframe df and its rows are basically logs of record temperatures T from a given station ID at a given time date