Evaluating historical climate extremes in the FGOALS-g3 large ensemble in the presence of internal climate variability

The observed climate includes both the response to external forcings and internal climate variability. Thus, more reasonable approaches of model evaluation which fully take into account the influence of internal variability are desirable, with the availability of initial-condition large ensembles (LEs). Here we evaluate the newly developed FGOALS-g3 LE (with 110 realizations which differ only in the initial conditions) in terms of the climatology and historical changes of temperature and precipitation extremes, by comparing multiple observational datasets with the full ensemble spread (representing a wide variety of possibilities of internal variability). We show that internal variability does not significantly affect the evaluation of model simulated climatology, but it does for the evaluation of long-term changes. For climatology, the FGOALS-g3 LE reasonably reproduces the global spatial distributions of temperature and precipitation extremes, with pattern correlations exceeding 0.90 for temperature extremes and around 0.70 for precipitation extremes. Specifically, the FGOALS-g3 LE shows an overall warm bias in warm extremes and cold bias in cold extremes aggregated over global land; the biases in extreme precipitation intensity only exist in limited regions while the biases in extreme precipitation amount are more widespread. For the long-term change since 1950, taking internal variability into account, the FGOALS-g3 LE can reasonably simulate the observed changes in cold extremes and precipitation extremes for global land average and for most land regions; while it generally overestimates the warming trend in warm extremes. With the growing availability of initial-condition LEs, it is desirable to apply these more reasonable approaches in model evaluations to fully consider the influence of internal variability.