There is a lot that is common in between $L_1$ and $L_2$. In fact, out of observation it is clear that $L_2 \subset L_1$.
Again from observation, we can exactly identify the strings that are in $L_1$ but not $L_2$. These strings are the one which belongs to the regular expression (r.e) $\epsilon(ab+a)$
Now, this r.e is obtained (as it might be obvious to you) by generating $\epsilon$ from $a^*$ in $L_1$. This is true because the only string that $a^*$ can generate but not $a^+$ is $\epsilon$ i.e empty string. Thus,
$$L_1 - L_2 = (ab+a)$$
HTH