Correcting for systematic biases in multiple raw GCM variables across a range of timescales

- Corrects for biases in first and second order moments and dependence.
- Corrects for biases on many variables simultaneously.
- Corrects for biases over all the time scales considered.
- Use of the approach is expected to deliver better results.

SummaryMany hydro-climatological applications require use of General Circulation Models (GCMs) outputs. However, the raw information as available from GCMs often contains significant systematic biases when compared with observations. This necessitates some kind of statistical adjustment to be carried out on the GCM fields before their use in any application. It is common to correct the GCM simulations by removing the systematic biases at the time scale of interest, usually individually for each GCM variable that is needed. The outcome is a set of bias corrected variables that are not assessed for bias in their joint dependence structure, nor biases in all attributes at other time scales (such as daily and monthly and annual) except the one under consideration. In this paper, we present a bias correction approach that simultaneously adjusts for the biases in multiple variables across multiple time scales. The proposed Multivariate Recursive Nesting Bias Correction (MRNBC) approach simultaneously corrects many GCM variables and repeats the procedure across different levels of temporal aggregation to impart observed distributional and persistence properties at multiple time scales. The bias corrected series exhibits improvements across all variables and over all the time scales considered. The use of the approach in hydrology and water resources related downscaling applications is expected to deliver better results.