Statistical considerations in network design

Following some general remarks on the importance and unimportance of optimization in spatial network design, we take up, in modest detail, how one might exploit spatial autocorrelations and covariateinformation. We point out that spatial autocorrelations themselves require care in their estimation andthen proceed with two illustrations to show how probabilistic error calculations are made for mappingproblems using network station data. One illustration uses a quantitative mapping variable, and the otheruses a qualitative mapping variable.