Resolving the Dry Period Projection Paradox: Treat ‘Consecutiveness’ as a Nonlinearity

Abstract

This paper provides a worked example of how the property of consecutiveness, or continuity, can be lost when computing climate indices such as consecutive dry days (CDD) or dry periods from model simulations of the future. That essential continuity property can easily be lost if such indices are computed from future projections of bias-corrected daily precipitation time series. A bias-correction algorithm such as quantile-mapping typically adjusts daily time series to remove overall precipitation bias, but takes no account of consecutiveness, and so can introduce occasional wet-day interruptions into otherwise dry periods. This can lead to inconsistencies between the raw and bias-corrected projections of such indices. To obtain consistent projections, CDD and related indices should be treated as independent parameters and bias-corrected directly in their own right. Such indices should be counted first, and bias-corrected later. In this sense, consecutiveness should be treated as a nonlinearity to be computed before performing any other mathematical operation such as bias correction. This paradox and its resolution are demonstrated using future climate projections from the TRANSLATE project, all of which are derived from global CMIP5 simulations as downscaled over Ireland by two separate regional model ensembles.