Similarities within a multi-model ensemble: functional data analysis framework

Abstract

Abstract. Despite the abundance of available global climate model (GCM) and regional climate model (RCM) outputs, their use for evaluation of past and future climate change is often complicated by substantial differences between individual simulations and the resulting uncertainties. In this study, we present a methodological framework for the analysis of multi-model ensembles based on a functional data analysis approach. A set of two metrics that generalize the concept of similarity based on the behavior of entire simulated climatic time series, encompassing both past and future periods, is introduced. To our knowledge, our method is the first to quantitatively assess similarities between model simulations based on the temporal evolution of simulated values. To evaluate mutual distances of the time series, we used two semimetrics based on Euclidean distances between the simulated trajectories and based on differences in their first derivatives. Further, we introduce an innovative way of visualizing climate model similarities based on a network spatialization algorithm. Using the layout graphs, the data are ordered on a two-dimensional plane which enables an unambiguous interpretation of the results. The method is demonstrated using two illustrative cases of air temperature over the British Isles (BI) and precipitation in central Europe, simulated by an ensemble of EURO-CORDEX RCMs and their driving GCMs over the 1971–2098 period. In addition to the sample results, interpretational aspects of the applied methodology and its possible extensions are also discussed.

Changes in temperature and precipitation in the instrumental period (1951–2018) and projections up to 2100 in Podgorica (Montenegro)

“…Holtanová et al . (2019) indicated that projections and interpretations of past and future climate change would be increasingly complicated, as there were significant differences in the results of both global climate models (GCMs) and regional climate models (RCMs) between the scenarios. Belda et al .…”

“…Other authors also noted that the temperature projections reported by the IPCC in its reports were overestimated, especially for the more extreme variants (Michaels et al, 2002;Ongoma et al, 2018). The uncertainties of climate models was discussed in the work of Laprise et al (2008), Holtanová et al (2014), Sanderson et al (2017). Holtanová et al (2019) indicated that projections and interpretations of past and future climate change would be increasingly complicated, as there were significant differences in the results of both global climate models (GCMs) and regional climate models (RCMs) between the scenarios.…”

“…The uncertainties of climate models was discussed in the work of Laprise et al (2008), Holtanová et al (2014), Sanderson et al (2017). Holtanová et al (2019) indicated that projections and interpretations of past and future climate change would be increasingly complicated, as there were significant differences in the results of both global climate models (GCMs) and regional climate models (RCMs) between the scenarios. Belda et al (2017) indicated the degree of uncertainty of a model depended on several factors-model structure, parameterization conditions, and at RCM and boundary conditions.…”

Using high-resolution regional climate models to estimate return levels of daily extreme precipitation over Bavaria

“…Multi-model experiments using the same boundary and starting conditions yield deviating simulations of the climate (Holtanová et al, 2019;Solman et al, 2013). Yet, also the same model applying differing physics options and parametrization schemes can lead to significant variability in the model results (Laux et al, 2019).…”

Evaluation of some distributional downscaling methods as applied to daily precipitation with an eye towards extremes

“…While temperature is eventually expected to rise everywhere the sign of the precipitation response varies by location, season, and between models. However, different RCM forced by the same GCM can yield opposite signed precipitation responses (Karmalkar, 2018; Holtanova et al ., 2019). Hence, projecting changes in precipitation is particularly challenging, especially at the regional scale.…”