Instrumental agreement and retrospective analysis of trends in precipitation extremes in the French Mediterranean Region

Abstract

In this letter we show the emergence of an agreement between the instruments of a rain-gauge network to point toward a positive trend in daily precipitation extremes since 1960 in the French Mediterranean Region. We identify for each gauge the time varying parameters of the generalized extreme value distribution of annual maximum precipitation over incremental time-windows. These distributions provide for each station of the network a trend assessment over a chosen period that can be interpreted for instance as a trend of the mean or as the trend of a chosen quantile. The incremental window, i.e. a window containing the series of data available at a given date, mimics the annual assessment of the trends that could have been made through time. Each year we thus have one trend per gauge that we can look in distribution through the network in order to assess the level of consensus among instruments. We show how the increasing size of the datasets used over a period of possible climate non-stationarity progressively leads from a dissensus anarchically pointing to no trend (before the 2000s) to a consensus where a majority of gauges points toward a positive trend (after the 2000s). The detected trend in this Mediterranean Region is quite substantial. For instance the 20 year return period precipitation in 1960 turns out to become a 8 year return period precipitation in 2020. Using a simulation basis we try to characterize the effect of decadal variability that is quite readable in the consensus evolution. The proposed metrics is thought to be a good candidate for the assessment of the local time and rate of emergence of climate change that has important implications in regards to adaptation of human and natural systems.