New Analysis Methodology, with ARM Measurements, Identifies Reasons Behind Climate Model Biases

Summary

To make confident predictions about future global and regional climate, global climate models (GCMs) must be capable of reproducing the present-day distribution of global heat and moisture. However, many GCMs exhibit a persistent bias in temperature over the mid-latitude continents, which is present in both short-range forecasts as well as long-term climate simulations. A common approach to evaluating model biases is to focus on the model-mean state, but this approach makes an unambiguous interpretation of the bias origins difficult, given that biases are often the result of the superposition of impacts of different processes over multiple time steps in the model.

A team of scientists funded in part by the Department of Energy’s (DOE) Atmospheric System Research and Regional and Global Climate Modeling programs developed a new methodology to objectively disentangle and quantify contributions from clouds and other processes in the creation of a surface warm bias in climate models. A unique feature of this approach is its focus on the growth of the temperature error at the time-step level. Compositing the error growth by the coinciding bias in total downwelling radiation provides unambiguous evidence for the role that clouds play in the creation of the surface warm bias during certain portions of the day. Furthermore, application of an objective cloud-regime classification allows for the detection of the specific cloud regimes that matter most for the bias creation. The new model evaluation methodology relies heavily on the availability of high-temporal resolution observations of temperature, cloud properties, and surface radiation from DOE’s Atmospheric Radiation Measurement (ARM) Climate Research Facility.

The scientists applied their new method to two state-of-the-art GCMs that exhibit a distinct warm bias over the ARM Southern Great Plains (SGP) site. The analysis finds that in one GCM, biases in deep-convective and low-level clouds contribute most to the temperature-error growth in the afternoon and evening, respectively. In the second GCM, deep clouds persist too long in the evening, leading to a growth of the temperature bias. The reduction of the temperature bias in both models in the morning and the growth of the bias in the second GCM in the afternoon could not be assigned to a cloud issue, but are more likely caused by a land-surface deficiency. This new analysis approach provides specific guidance to model developers about the processes on which they should focus development efforts to resolve existing model biases.

References

Van Weverberg, K., C. J. Morcrette, H.-Y. Ma, S. A. Klein, and J. C. Petch. 2015. “Using Regime Analysis to Identify the Contribution of Clouds to Surface Temperature Errors in Weather and Climate Models,” Quarterly Journal of the Royal Meteorological Society, DOI: 10.1002/qj.2603.