Intercomparison of bias correction approaches for simulated temperature by multiple climatic models over southern India.

Climate model, a complex numerical representation of the global climate system, has been developed to simulate current climate and used to project future climatic conditions. Simulated climatic variables from climate models often exhibit significant deviations from observations. In climate projections, different approaches were introduced to deal with systematic deviations and random model errors. This paper demonstrates the intercomparison of four bias correction approaches (linear scaling, delta change correction, distribution mapping, and variance scaling) underlying the assumptions of stationary output from climate models. Mean monthly temperatures derived from five global climate models were corrected by four bias correction approaches for five states of southern India. The suitability of correction approaches depends upon the climate models and regional framework. The applied approaches improve the mean values and other statistical properties. The results show that all four bias techniques significantly improved the simulated data, but distribution mapping and variance scaling were more effective in removing systematic model biases. [ABSTRACT FROM AUTHOR]