Non‐stationary peaks‐over‐threshold analysis of extreme precipitation events in Finland, 1961–2016
There is an urgent need to understand and predict how extreme precipitation events (EPEs) will change at high latitudes, both for local climate change adaptation plans and risk mitigation and as a potential proxy “anticipating” the impact of climate change elsewhere in the world. This paper illustra...
Published in: | International Journal of Climatology |
---|---|
Main Authors: | , |
Other Authors: | |
Format: | Article in Journal/Newspaper |
Language: | English |
Published: |
Wiley
2018
|
Subjects: | |
Online Access: | http://dx.doi.org/10.1002/joc.5867 https://api.wiley.com/onlinelibrary/tdm/v1/articles/10.1002%2Fjoc.5867 https://rmets.onlinelibrary.wiley.com/doi/pdf/10.1002/joc.5867 |
Summary: | There is an urgent need to understand and predict how extreme precipitation events (EPEs) will change at high latitudes, both for local climate change adaptation plans and risk mitigation and as a potential proxy “anticipating” the impact of climate change elsewhere in the world. This paper illustrates that a combination of non‐stationary modelling approaches can be adopted to evaluate trends in EPEs under uncertainty. A large database of daily rainfall events from 281 sparsely distributed weather stations in Finland between 1961 and 2016 was analysed. Among the tested methods, Poisson distributions provided a powerful method to evaluate the impacts of multiple physical covariates, including temperature and atmospheric circulation patterns (ACPs), on the resulting trends. The analysis demonstrates that non‐stationarity is statistically valid for the majority of observations, independently of their location in the country and the season of the year. However, subsampling can severely hinder the statistical validity of the trends, which can be easily confused with random noise and therefore complicate the decision‐making processes regarding long‐term planning. Scaling effects have a strong impact on the estimates of non‐stationary parameters, as homogenizing the data in space and time reduces the statistical validity of the trends. Trends in EPE statistics (mean, 90 and 99% percentiles) and best‐fitted Generalized Pareto parameters in the tails of the distributions appear to be stronger when approaching the Polar region (Lapland) than away from it, consistent with the Arctic amplification of climate change. ACPs are key covariates in physically explaining these trends. In particular, the Arctic Oscillation (AO) and North Atlantic Oscillation (NAO) can explain statistically significant increases in extreme precipitation in Lapland, Bothnian and South regions of Finland, particularly during summer and fall seasons. |
---|