1. What are the key features and extensions of the MESMER-X emulator for climate extremes and water cycle variables?
The MESMER-X emulator is a model developed to replicate regional changes in climate extremes and water conditions of Earth System Models (ESMs) at a lower computational cost. It evaluates annual extreme temperatures (TXx) for every land grid point of the Earth, over an arbitrary number of emulations, reproducing the natural variability and the local statistical distributions of TXx. MESMER-X was trained on each available ESM of the Climate Model Intercomparison Project Phase 6 (CMIP6) and accounts for the spatial and temporal correlations in TXx. The emulator has been extended to emulate annual indicators of interest for fire weather and soil moisture, which are relevant for assessing the potential of nature-based solutions to mitigate climate change. These extensions stress-test the emulator's capacity in various situations and are of high relevance for further development of the MESMER-X emulator.
read more
2. How does MESMER-X extend MESMER?
MESMER-X is an extension of MESMER, designed to represent impact-related variables, including climate extremes. It allows the emulation of monthly local temperatures, expanding the capabilities of MESMER. The emulator has been applied to various scenarios, such as evaluating the contributions of emitters to regional warming and integrating spatial observational constraints to improve local temperature projections. MESMER-X has also been coupled with the simple climate model MAGICC, enabling efficient calculations of local responses to emissions scenarios, considering uncertainties in modeling and natural variability. This extension enhances the versatility and applicability of MESMER in climate research.
read more
3. What is the MESMER-X emulator approach and how does it emulate spatially resolved climate variability?
The MESMER-X emulator approach emulates spatially resolved climate variability by sampling from conditional distributions. It replaces the pattern scaling of MESMER using conditional distributions for a more flexible 'distribution' scaling. The training of spatio-temporal correlations is similar to MESMER, but performed not on the residuals of the pattern scaling, but by projecting the sample onto a standard normal distribution using a probability integral transform. The emulator assumes that a climate variable can be represented locally by a probability distribution, such as a Generalized Extreme Value distribution for block-extrema or a normal distribution for averages. To account for spatio-temporal correlations, the emulator parametrizes internal climate variability using an autoregressive process with spatially correlated innovations. The normalized residuals are characterized using this process, and the resulting covariance matrix is regularized using localization. The emulations are obtained through a back probability integral transform. This approach allows for emulating spatially resolved climate variability by considering conditional distributions and spatio-temporal correlations.
read more
4. What is the Continuous Rank Probability Score (CRPS) used for in MESMER-X configuration?
The Continuous Rank Probability Score (CRPS) is used to measure differences in the cumulative distribution functions of the emulations and the training set in MESMER-X configuration. It is a commonly used score in atmospheric sciences to assess and compare the performance of different configurations. The CRPS is calculated using equations (5) and (6) and is used to define the Continuous Rank Probability Skill Score (CRPSS) by comparing the CRPS of a configuration to the CRPS of a benchmark 0. A high CRPS for the benchmark indicates significant differences between the cumulative distribution functions, suggesting potential issues with the sample properties. To simplify comparisons, the CRPS is averaged over space, time, and scenarios.
read more