Tail dependence models for distributions attracted to a max-stable law are tted using observations above a high threshold. To cope with spatial, high-dimensional data, a rankbased M-estimator is proposed relying on bivariate margins only. A data-driven weight matrix is used to minimize the...