Open AccessAssessing the performance of bias correction approaches for correcting monthly precipitation over I ndia through coupled models
Author(s) -
Singh Ankita,
Sahoo Raj Kumar,
Nair Archana,
Mohanty U. C.,
Rai R. K.
Publication year - 2017
Publication title -
meteorological applications
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.672
H-Index - 59
eISSN - 1469-8080
pISSN - 1350-4827
DOI - 10.1002/met.1627
Subject(s) - precipitation , quantile , statistics , econometrics , climatology , environmental science , mathematics , monsoon , transformation (genetics) , meteorology , geography , chemistry , biochemistry , gene , geology
The objective of the present study was to investigate the inter‐annual variation and error structure in the prediction of monthly precipitation through two global coupled models, the N ational C enters for E nvironmental P rediction C limate F orecast S ystem version 2 ( CFSv2 ) and the G eophysical F luid D ynamics L aboratory model. In view of the consistent systematic bias (dry bias during summer monsoon months and wet bias during pre‐monsoon months in CFSv2 ) a requirement to correct the inherent error is inevitable. For this purpose, a few bias correction methods, standardization−reconstruction ( Z ), quantile−quantile mapping ( QQ ) and nonlinear transformation ( NL_Zi ), are explored. The methods are applied to the outputs of the dynamic models and the efficiency is examined through different statistical skill measures. A maximum error reduction is noticed for M arch, J uly, S eptember and D ecember. A decreasing tendency for rainfall in J uly is represented by the raw model and its biased counterpart. The observed probability is noticed to be overestimated (underestimated) corresponding to below normal (above normal) precipitation in the raw model. The varying relationship between monthly precipitation and the NINO3 .4 index might be a reason for misleading prediction during extreme years. Among the bias correction methods, NL_Zi showed maximum improvement in terms of predicting the precipitation amount and probability distribution all through the year irrespective of the selection of the coupled model.