A strategy to effectively make use of large volumes of climate data for climate change adaptation

Abstract

A strategy is suggested for presenting high-resolution temperature maps based on projections from large multi-model ensembles with minimal requirement of data space. This ability to reduce data volumes may be useful for climate services. We present a web-based solution that provides maps with seasonal mean temperatures at 5-min spatial resolution. The maps were generated from downscaled groups of 223 stations from the Barents region, and were based on results from principal component analysis (PCA) for which the five leading modes represented most of the variance and enabled the extraction of salient features while significantly reducing the data volume. A demonstration of the concept showed how different aspects can be distilled, such as ensemble means, ensemble member differences, point-wise time series, probabilities, number of hot/cold days, and various quality aspects. The demonstration included three different types of emissions scenarios: the RCPs 2.6, 4.5, and 8.5. This way of organising data is instrumental to extracting relevant information for decision-making, but does not alone imply actionable adaptation information. The question of reliability and robustness depends on the quality of the data rather than the way it is organised.

Keywords

• Big data
• Ensembles
• Downscaling
• Distillation