Avoiding Inhomogeneity in Percentile-Based Indices of Temperature Extremes

Abstract

Using a Monte Carlo simulation, it is demonstrated that percentile-based temperature indices computed for climate change detection and monitoring may contain artificial discontinuities at the beginning and end of the period that is used for calculating the percentiles (base period). This would make these exceedance frequency time series unsuitable for monitoring and detecting climate change. The problem occurs because the threshold calculated in the base period is affected by sampling error. On average, this error leads to overestimated exceedance rates outside the base period. A bootstrap resampling procedure is proposed to estimate exceedance frequencies during the base period. The procedure effectively removes the inhomogeneity. Abstract Using a Monte Carlo simulation, it is demonstrated that percentile-based temperature indices computed for climate change detection and monitoring may contain artificial discontinuities at the beginning and end of the period that is used for calculating the percentiles (base period). This would make these exceedance frequency time series unsuitable for monitoring and detecting climate change. The problem occurs because the threshold calculated in the base period is affected by sampling error. On average, this error leads to overestimated exceedance rates outside the base period. A bootstrap resampling procedure is proposed to estimate exceedance frequencies during the base period. The procedure effectively removes the inhomogeneity.