Effects of secular changes in frequency of observations and observational errors on monthly mean MSLP summary statistics derived from ICOADS

A Monte Carlo technique has been employed to assess how sextile mean sea level pressure (MSLP) statistics derived from ship observations can be affected by changes in the frequency of observations. The results show that when the number of observations is small (less than 20 per month), the estimates of the first sextile as well as the intersextile range, which is considered to be a resistant estimate of the standard deviation, can contain large biases. The results also suggest that, while changes in, the frequency of observations do not have strong impacts on the standard way of estimating the standard deviation, such statistics are strongly affected by secular trends in observational error statistics. The results are applied to examine the increasing trend in cool season (December-March) Pacific cyclone activity during the second half of the twentieth century. The results show that the trends in sextile statistics derived front the NCEP-NCAR reanalysis data are only consistent with those derived from the international Comprehensive Ocean-Atmosphere Data Set (ICOADS) summary statistics if biases due to changes in the frequency of observation are not taken into account. When such biases are accounted for, the trends derived from the observations are significantly smaller than those derived from the reanalysis data. As for the increasing trend in MSLP variance, the trends derived from the ICOADS statistics are smaller than those derived from the reanalysis regardless of whether corrections are made to account for the secular trend in MSLP error statistics. In either case, the corrections that have to be applied have the same order of magnitude as the observed trends. The two main conclusions are that 1) climate statistics can be strongly affected by changes in frequency of observations as well as changes in observational error statistics and 2) the trends in North Pacific winter cyclone activity, as derived from NCEP-NCAR reanalysis data, appear to be significantly larger than similar trends computed from ICOADS sextile and variance statistics, when biases due to changes in frequency of observations and observational error statistics have been taken into account.