A quantile mapping approach‐based bias correction in Coupled Model Intercomparison Project Phase 5 models for decadal temperature predictions over India

• Published in: International journal of climatology 2022-03, Vol.42 (4), p.2455-2469
• Main Authors: Patel, Jayshri, Gnanaseelan, Chellappan, Chowdary, Jasti S., Parekh, Anant
• Format: Article
• Language: English
• Subjects: Agriculture; Air temperature; Atmospheric models; Bias; bias correction; Climate models; climate over India; Climate prediction; Climate variability; Community planning; Corrections; decadal predictions; Intercomparison; Lead time; Mapping; Modelling; Ocean models; quantile mapping; Spatial distribution; surface air temperature; Surface-air temperature relationships; Uncertainty
• ISSN: 0899-8418; 1097-0088
• DOI: 10.1002/joc.7376

Decadal climate prediction has been recognized as the important information for policy makers for agriculture, health and energy sectors and the general public for any near‐term planning. The modelling community is determined to put forward a reliable near‐term climate forecasting system, to predict annual to decadal state and variability of climate. However, deriving reliable information from the decadal prediction/hindcast is still a challenge. This study examines the decadal hindcast simulations of surface air temperature (SAT) over India in seven different ocean–atmosphere coupled models from the Coupled Model Intercomparison Project Phase 5 (CMIP5). Each decadal hindcast is available for the next 10‐years period from the initialized climate states of 1961–2006. The performance of models is assessed using different evaluation metrics, such as absolute mean difference, root mean square error, skill score and uncertainty in terms of the range of hindcasts. The multimodel ensemble mean displayed considerable skill in representing the spatial distribution of SAT over the Indian region except over the western Himalaya and Northeast India. Our results indicate that the decadal hindcasts skills improved noticeably when quantile mapping (QM) approach is used for the bias corrections. The major improvements are seen in terms of reducing absolute mean difference and uncertainty, regardless of lead time and region. The present study advocates that QM approach is useful not only for reducing bias but also for improving the decadal hindcast skill for SAT over India in the coupled models. The raw and bias‐corrected decadal hindcast simulations of surface air temperature (SAT) are examined. The multimodels ensemble mean from raw general circulation models is reasonably well in representing the spatial distribution of SAT over the Indian region except western Himalaya and northeast India. Quantile mapping (QM) is able to reduce absolute mean difference, unconditional bias and uncertainty, regardless of lead time and regions, thereby supporting that the QM Approach is useful in improving decadal prediction skill of coupled models.Figure: Schematic of a decadal prediction system, its skill and bias correction.