1: \begin{abstract}

2: The discretization of optimal transport problems often leads to large

3: linear programs with sparse solutions. We derive error estimates for

4: the approximation of the problem using convex combinations of Dirac

5: measures and devise an active-set strategy that uses the optimality

6: conditions to predict the support of a solution within a multilevel

7: strategy. Numerical experiments confirm the theoretically predicted

8: convergence rates and a linear growth of effective problem sizes 

9: with respect to the variables used to discretize given data. 

10: \end{abstract}

11: