1: \begin{abstract}

2: \noindent

3: This paper studies best finitely supported approximations of

4: one-dimensional probability measures with respect to the

5: $L^r$-Kantorovich (or transport) distance, where either the

6: locations or the weights of the approximations' atoms are

7: prescribed. Necessary and sufficient optimality conditions are

8: established, and the rate of convergence (as the number of atoms goes

9: to infinity) is discussed. In view of emerging mathematical and

10: statistical applications, special attention is given to the case of

11: best uniform approximations (i.e., all atoms having equal weight). The

12: approach developed in this paper is elementary; it is based on best 

13: approximations of (monotone) $L^r$-functions by step functions, 

14: and thus different from, yet naturally complementary to, the classical 

15: Voronoi partition approach.

16: \end{abstract}