Does Increasing Horizontal Resolution Improve Seasonal Prediction of Indian Summer Monsoon?: A Climate Forecast System Model Perspective.

The seasonal prediction skill of climate forecast system model with two different resolutions, namely T126 and T382, is studied using hindcast data. Using novel diagnostic tools such as total variation distance and two‐state Markov Chain analysis, it is shown that increasing the horizontal resolution of the model has minimal impact on the quality of seasonal prediction. The underlying rain distribution and associated transition probabilities are very similar in both model versions. The Markov chain analysis also provides critical clues about the issues associated with convective processes in the model. Both the models produce longer (shorter) wet (dry) spells compared to the observations. Models are trying to bring the atmosphere closer to convective quasi‐equilibrium, leading to a substantial departure from observed features. Although the conventional error metrics are helpful to assess the prediction skills, the new metrics used in the study provide further insights on possible pathways to improve model moist physics. Plain Language Summary: The state of the art in respect to the prediction of Indian summer monsoon uses coupled climate models. Models, in general, have a hard time producing an accurate forecast of seasonal mean monsoon. Serious effort is being put globally to improve the quality of the forecast. One of the simple and easy to comprehend approaches among the scientific community is to increase the model's horizontal resolution to improve the accuracy of the predictions. The present study indicates that this approach alone is not very useful as there is an only marginal improvement with an increase in the resolutions. Key Points: Novel metrics have been used to compare model hindcast at different horizontal resolutionsPrevious studies claiming an improved simulation of monsoon with higher resolutions climate forecast system model require further investigationOnly a marginal improvement is seen in the fidelity of the model with an increase in horizontal resolution [ABSTRACT FROM AUTHOR]