AUTHOR=Parey Sylvie 

TITLE=Generating a Set of Temperature Time Series Representative of Recent Past and Near Future Climate

JOURNAL=Frontiers in Environmental Science

VOLUME=Volume 7 - 2019

YEAR=2019

URL=https://www.frontiersin.org/journals/environmental-science/articles/10.3389/fenvs.2019.00099

DOI=10.3389/fenvs.2019.00099

ISSN=2296-665X

ABSTRACT=Temperature plays an important role in electricity demand and generation; heat waves and cold waves represent potential threats to the energy system due to the sustained nature of such events. In the current climate change context, it is important to anticipate possible changes in the occurrence of the most severe events to adapt energy system planning and management. However, because such events are rare, any estimation of their frequency is uncertain and a large data sample is required to reduce any sampling uncertainty.
In this paper, we present a way of combining past observations and climate model simulations to generate very large samples for a period extending from the recent past to the near future. It is based on the decomposition of the temperature signal into deterministic parts (smooth trends and seasonality in the mean and the standard deviation) and stochastic residuals. Once the observed signal is decomposed for a long enough time period (at least 30 years), new time series can be built using bias adjusted climate simulation trends, observed seasonality and simulated stochastic residuals.
The approach is first described and applied to an observed 32-station weighted average temperature time series in France, used as an indicator for the role of temperature in electricity demand, in a cross-validation setting. Using the observed time series of this thermal indicator over the period 1950-2017, the approach is calibrated over the period 1955-1986 to generate similar time series over the whole period 1955-2017 using different CMIP5 GCM simulations. Then the generated set of time series are validated for the mean annual cycle and distributions of heat waves and cold waves over the period 1987-2017. The choices made for generation and reconstruction are detailed and motivated. The methodology is used to generate a large set of indicator evolutions covering recent past and near future periods, which is used to identify changes in the frequencies of the most severe heat waves or cold waves. The increased frequency of very severe heat waves is found in agreement with previous studies, and estimates of the decreased frequency of the most severe cold waves are also provided.