Redesigning continental-scale monitoring networks

To address the problem of redesigning an existing continental-scale monitoring network, a method is given for optimally adding sites to a subregion of the continent so that both the mean relative error of estimation (estimate standard error divided by estimate) over the subregion and the standard deviation of the relative error estimate at the subregion's center are minimized.The method consists of two steps. First, the region is divided into disjoint subregions having homogeneous climatic/ecological characteristics. Then, for each subregion, optimal site locations are found by minimizing the dual objective function composed of the subregion's mean relative error and the standard deviation of a simulated relative error estimate at the subregion's center. The relative error with new sites is calculated by a moving window kriging algorithm that accounts for spatial trebd and heterogeneous spatial covariance through the use of data from the existing network.This method is applied to the problem of adding new sites to the U.S. National Atmospheric Deposition Program/National Trends Network using wet sulfate deposition data. It is found that for many of the subregions employed, on the order of a hundred new sites may be needed to reduce the subregion mean relative error by 5%. The subregion-center relative error standard deviation, however, can be reduced by about 50% in most subregions by the addition of no more than 10 sites.A subregion-based redesign method may be preferable to a region-based method when (a) network design goal priorities differ across subregions, and/or (b) network performance measures calculated over the entire region are insignificantly affected by the addition of a small number of new sites.