Fossil climate indicators are often quite fragmental. Further, sampling points of fossil indicators do not have uniform distribution. Latitudinal distribution of the terrestrial area are changing because of plate tectonics. All these circumstances may bring about strongly biased set of sampling points, which in turn cause errors in the reconstruction of the paleo-climate during the Phanerozoic. If the number of the sample points equals 100 or less, and when there is no information on altitude of each sample point, our simulations estimate that inferred global average surface temperature is accompanied with an uncertainty of more than 2 K. Results of our simulation indicates the inferred temperature can strongly vary with the change of the land-sea distribution. This result suggests that the reconstructed climate change in precedent studies can be a result of the varying land-sea distribution under the same global climatic condition during the Phanerozoic.