Minimum error estimates of global mean temperature through optimal arrangement of gauges

This paper considers the minimum mean squared error (MSE) incurred in estimating an idealized earth's global average temperature with a finite network of point gauges located over the globe. We use a spectral MSE formalism to find the optimal locations for N gauges in the problem of estimating the earth's global average temperature. Limiting MSE configurations are obtained as the limiting least error case for randomly distributed samples of size N . Our results suggest that for N greater than about 60, one can obtain estimates such that the amount of measured variance due to sampling error is less than 10%, a result likely to be acceptable to climatologists.