A power-bounded operator T satisfying
$\sup _n n\lVert T^n-T^{n+1}\rVert <\infty $ is a Ritt operator. For such operators, we study the generalized square function
$Q_{\alpha ,s,r}^Tf=(\sum _n n^{\alpha } |T^n(I-T)^rf|^s)^{1/s}$. It is known that when T is a positive contraction and a Ritt operator on
$L^p$,
$1<p<\infty $, then for any integer
$r\ge 1$, the square function
$Q_{2r-1,2,r}^Tf$ defines a bounded operator [17] on
$L^p$. In this work, we extend the theory to the endpoint case
$p=1$. We show that if T is a Ritt operator on
$L^1$, then the generalized square function
$Q_{\alpha ,s,r}^Tf $ is bounded on
$L^1$ whenever
$\alpha +1<sr$. In the particular setting where T is a convolution operator of the form
$T_{\mu }=\sum _k \mu (k) U^kf$, with
$\mu $ a probability measure on
$\mathbb Z$ and U the composition operator induced by an invertible measure-preserving transformation, we provide sufficient conditions on
$\mu $ under which
$Q_{2r-1,2,r}^{T_{\mu }}f$ is of weak type
$(1,1)$, for
$r>0$. We also establish bounds for variational and oscillation norms,
$\lVert n^{\beta } T^n(1-T)^r\rVert _{v(s)}$ and
$\lVert n^{\beta } T^n(1-T)^r\rVert _{o(s)}$, for Ritt operators, highlighting endpoint behavior.