Geostatistical estimation variance approach to optimizing an air temperature monitoring network

Monitoring is important and necessary for any scientific study and optimal monitoring makes the project cost-effective. There are many approaches and constraints on the basis of which a network could be optimized. Geostatistical estimation variance reduction is one of the unbiased ways of optimizing a network with a desired degree of accuracy. A monitoring network in Colarado (USA) for measuring air temperature had 21 measuring locations fairly but randomly distributed in the area. The temperature data received from these stations were analysed geostatistically to infer the spatial variability of the parameter. Then a threshold on the maximum S.D. of the estimation error was arbitrarily fixed, by comparing with the S.D. of the data to maintain the order of magnitude. Using the advantage that calculation of the geostatistical estimation error is independent of the measured values, the variance of the estimation error on a uniform grid was calculated to compare with the threshold. A number of additional measurement points were incorporated, a couple of them were removed and some points were also shifted. Thus, an optimal monitoring network evolved. The result shows that with a slight increase in the measurement cost, the accuracy of estimation from the optimized network can be increased considerably.