| Popis: |
To describe earth’s former and predict its expected future climate in a general way, we need to understand at least two basic characteristics of the distribution of earth’s temperature, its mean state and its temporal and spatial variance. While there is some confidence in projections of the mean state, earth system models (ESMs) models are suspected to underestimate natural climate variability on longer time scales, as they are developed and evaluated mainly on observational data that only cover a short time period. At low frequencies they show up to two orders of magnitude less temperature variability than data from proxy archives.However, while climate proxies, such as oxygen isotopes from foraminifera retrieved from marine sediments, provide long records, they are sparse in time and space and contain additional non-climate components such as measurement error. A fundamental criticism of proxy data is that correlation between even nearby records is low and therefore they cannot represent the shared climate signal between the two locations.So how to decide whether climate models simulate enough variability at long timescales? We here suggest a new evaluation method for existing climate models. The idea is to combine proxy data, proxy understanding and climate models in such a way that we can derive a plausibility check on the ability of the climate models to reproduce natural climate variability. Considering the high noise levels that we find, what if the proxy records don’t replicate well but still replicate too well enough to have come from a relatively low-variability climate like the climate models?This is the question we answer in this combined data analysis and simulation study. We do this by developing a hypothesis test where transient ESM simulation TraCE21k forms the null hypothesis on the nature of climate variability. Our proxy data base are multiple and high-resolution Holocene oxygen isotopes and Mg/Ca records from marine sediment cores from the Eastern tropical Indian Ocean - a region with a strong variability mismatch between proxy data and climate model simulations.The result of the hypothesis test is that, based on the correlation we find in our records, we have to reject the null hypothesis. That means the possibility that the climate signal that was imprinted in our records was generated by a climate system with TraCE21k-like variability characteristics is very unlikely. We therefore reject the hypothesis that TraCE21k reproduces the climate variability characteristics.We will expand the analysis on different simulations of last generation of ESMs to see if they perform better. |
