Correcting biases in atmospheric variables prior to impact studies or dynamical downscaling can lead to new biases as dynamical consistency between the 'corrected' fields is not maintained. Use of these bias corrected fields for subsequent impact studies and dynamical downscaling provides input conditions that do not appropriately represent intervariable relationships in atmospheric fields. Here we investigate the consequences of the lack of dynamical consistency in bias correction using a measure of model consistency—the potential vorticity (PV). This paper presents an assessment of the biases present in PV using two alternative correction techniques—an approach where bias correction is performed individually on each atmospheric variable, thereby ignoring the physical relationships that exists between the multiple variables that are corrected, and a second approach where bias correction is performed directly on the PV field, thereby keeping the system dynamically coherent throughout the correction process. In this paper we show that bias correcting variables independently results in increased errors above the tropopause in the mean and standard deviation of the PV field, which are improved when using the alternative proposed. Furthermore, patterns of spatial variability are improved over nearly all vertical levels when applying the alternative approach. Results point to a need for a dynamically consistent atmospheric bias correction technique which results in fields that can be used as dynamically consistent lateral boundaries in follow-up downscaling applications.
Fig. 3. Time-averaged variograms showing semivariance versus great circle distance for the PV field of ERA-I (PVE), and bias correction
methods PV′, and PV′′ over vertical levels (a) 70 hPa, (b) 400 hPa, (c) 600 hPa, and (d) 925 hPa for the validation period. Difference between
PV′ and PVE (e) and PV′′ and PVE (f) standardized by the maximum difference between PV′ or PV′′ and PVE on each level.
This page is maintaind by Jason Evans |
Last updated 29 November 2013