1985 Volume 63 Issue 6 Pages 1147-1156
Climatic noise is an important concept, because it is a measure of statistical sampling error in time mean and of potential predictability of climate. However, reliable method of its estimate has not yet been established mainly because of difficulties in separation of climatic signals from noise. Noticing year-to-year fluctuations of intra-month variances, an estimate of the noise is attempted in the present paper, under the following three assumptions: i) anomaly of daily weather can be expressed by sum of anomalies of changes due to internal dynamics (n) and those due to external or boundary forcing (s) ; ii) intra-month variance may be equal to sum of squares of deviations of (n) and (s), iii) effect of signal on noise estimate may vanish when intra-month variance is the smallest in 30-year time series.
Year-to-year fluctuations of intra-month variances are confirmed by analysis of 30-year time series of daily mean surface air temperature at 19 stations in Hokkaido. And no appreciable influence of the signal on autocorrelation could be found in the data analysis.
Estimate of the noise is made by utilizing autocorrelation of autoregressive model of the second order and the smallest intra-month variance during 30 years. The noise thus estimated is 0.3-0.8°C, and clearly less than noise by method of other authors. Comparison with the observational error shows that the noise is greater than observational error, and implies that consideration of the noise alone is sufficient for ambiguity of time mean. It is noted that the noise of spatial mean over an area (ca.400km×400km) has no appreciable difference from that of individual stations, and for this fact an interpretation is given.