Bias in dynamically downscaled rainfall characteristics for hydroclimatic projections

Potter, Nicholas J.; Chiew, Francis H. S.; Charles, Stephen P.; Fu, Guobin; Zheng, Hongxing; Zhang, Lu

doi:https://doi.org/10.5194/hess-24-2963-2020

Articles | Volume 24, issue 6

https://doi.org/10.5194/hess-24-2963-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

https://doi.org/10.5194/hess-24-2963-2020

© Author(s) 2020. This work is distributed under
the Creative Commons Attribution 4.0 License.

Articles | Volume 24, issue 6

Research article

|

08 Jun 2020

Research article |

| 08 Jun 2020

Bias in dynamically downscaled rainfall characteristics for hydroclimatic projections

Nicholas J. Potter, Francis H. S. Chiew, Stephen P. Charles, Guobin Fu, Hongxing Zheng, and Lu Zhang

Download

Final revised paper (published on 08 Jun 2020)
Preprint (discussion started on 17 Apr 2019)

Interactive discussion

Status: closed

AC: Author comment | RC: Referee comment | SC: Short comment | EC: Editor comment

- Printer-friendly version

- Supplement

RC1: 'Comprehensive QQM case study', Anonymous Referee #1, 17 May 2019
- AC1: 'Authors' comment', Potter Nick, 11 Jul 2019
RC2: 'Review', Anonymous Referee #2, 20 May 2019
- AC1: 'Authors' comment', Potter Nick, 11 Jul 2019
RC3: 'Review', Anonymous Referee #3, 05 Jun 2019
- AC1: 'Authors' comment', Potter Nick, 11 Jul 2019

Peer-review completion

AR: Author's response | RR: Referee report | ED: Editor decision

ED: Reconsider after major revisions (further review by editor and referees) (14 Jul 2019) by Nadav Peleg

AR by Potter Nick on behalf of the Authors (30 Aug 2019) Author's response Manuscript

ED: Referee Nomination & Report Request started (02 Sep 2019) by Nadav Peleg

RR by Anonymous Referee #3 (03 Oct 2019)

Suggestions for revision or reasons for rejection

The authors missed the point regarding bias with respect to spatial correlation or persistence. They discuss the "inflation" issue as raised by Maruan (2013). This is regarding the inflation of the subgrid aggregated precipitation amount. This stems from applying the same quantity from a lower-resolution RCM/GCM to all of the points or grids that fall within each low-res grid cell. So, for example, when the model simulates 100 mm for a day, this amount is similarly bias corrected for all subgrid cells contained within the larger model grid cell. This often will lead to inflating the area-average, because the subgrid cell variability is not well represented.

Entirely separate, but equally important, is the issue of modeled versus observed spatial correlation (or what the authors refer to as spatial structure) when they are on the same grid (Bardossy and Pegram, 2012). Consider an example where both the model and observations are on the same 10x10 grid. Also, consider the model to have, on average, larger spatial correlation (correlation in time between each grid cell, this would be a 100x100 correlation matrix) than observations (i.e., when it rains at one grid cell, it is much more likely to rain at an adjacent grid cell than what observations show). This effect will translate to inflating the area-averaged precipitation amounts of the bias-corrected modeled data. This is because the model has larger covariance or spatial correlation or coherence or spatial structure than the observations, so the extent of individual model storms are, on average, greater than those seen in observations. The impact of the model over- or under-estimating the spatial correlation matrix, with respect the observations, is not effectively dealt with by bias correction. And again, this is an entirely different issue than the “inflation” issue discussed by Maraun (2013). The authors end the paragraph that begins page 10 with, “However, the issue of using a bias correction methodology that corrects daily amounts (and more generally temporal structure) while preserving spatial structure across catchments and basins remains a challenge and is a direction for further research.” I would agree that a multivariate method that effectively deals with temporal and spatial biases simultaneously is a direction for further research. However, a researcher can correct for spatial structure via recorrelation (Bardossy and Pegram, 2012) and temporal structure via a commonly used univariate bias correction method. The result of correcting the biases of time and space independently will not perfectly correct all possible types of bias within the modeled data, but it will be superior to doing one or the other alone. At the very least, the impact that model bias (in spatial structure) has on area-averaged precipitation needs to be further acknowledged and discussed.

Hide

RR by Anonymous Referee #2 (09 Oct 2019)

ED: Reconsider after major revisions (further review by editor and referees) (14 Oct 2019) by Nadav Peleg

AR by Potter Nick on behalf of the Authors (13 Jan 2020) Author's response Manuscript

ED: Referee Nomination & Report Request started (22 Jan 2020) by Nadav Peleg

RR by Anonymous Referee #2 (06 Feb 2020)

ED: Publish subject to minor revisions (review by editor) (14 Feb 2020) by Nadav Peleg

AR by Potter Nick on behalf of the Authors (24 Feb 2020) Author's response Manuscript

ED: Publish subject to technical corrections (26 Feb 2020) by Nadav Peleg

AR by Potter Nick on behalf of the Authors (02 Apr 2020) Author's response Manuscript

Short summary

There is a growing need for information about possible changes to water resource availability in the future due to climate change. Large-scale outputs from global climate models need to be translated to finer-resolution spatial scales before hydrological modelling. Biases in this downscaled data often need to be corrected. We show that usual bias correction methods can retain residual biases in multi-day occurrences of rainfall, which can result in biases in modelled runoff.