the Creative Commons Attribution 4.0 License.
the Creative Commons Attribution 4.0 License.
The impact of lateral boundary forcing in the CORDEX-Africa ensemble over southern Africa
- 1Department of Meteorology and Climatology, School of Geology, Faculty of Sciences, Aristotle University of Thessaloniki, Thessaloniki, Greece
- 2NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway
- 3CETEMPS—Department of Physical and Chemical Sciences, University of L’Aquila, L’Aquila, Italy
- 4Department of Physical and Chemical Sciences, University of L’Aquila, L’Aquila, Italy
- 5Rossby Centre, Swedish Meteorological and Hydrological Institute, Norrköping, Sweden
- 1Department of Meteorology and Climatology, School of Geology, Faculty of Sciences, Aristotle University of Thessaloniki, Thessaloniki, Greece
- 2NORCE Norwegian Research Centre, Bjerknes Centre for Climate Research, Bergen, Norway
- 3CETEMPS—Department of Physical and Chemical Sciences, University of L’Aquila, L’Aquila, Italy
- 4Department of Physical and Chemical Sciences, University of L’Aquila, L’Aquila, Italy
- 5Rossby Centre, Swedish Meteorological and Hydrological Institute, Norrköping, Sweden
Abstract. The region of southern Africa (SAF) is among the most exposed climate change hotspots and is projected to experience severe impacts on multiple economical and societal sectors. For this reason, producing reliable projections of the expected impacts of climate change is key for local communities. In this work we use a set of 19 regional climate models (RCMs) performed in the context of the Coordinated Regional Climate Downscaling Experiment (CORDEX) – Africa and a set of 10 global climate models (GCMs) participating in the Coupled Model Intercomparison Project Phase 5 (CMIP5), that were used as the driving GCMs in the RCM simulations. We are concerned about the degree to which RCM simulations are influenced by their driving GCMs, with regards to monthly precipitation climatologies, precipitation biases and precipitation change signal, according to the Representative Concentration Pathway (RCP) 8.5 for the end of the 21st century. We investigate the degree to which RCMs and GCMs are able to reproduce specific climatic features over SAF and over three sub-regions, namely the greater Angola region, the greater Mozambique region and the greater South Africa region. We identify that during the beginning of the rainy season, when regional processes are largely dependent on the coupling between the surface and the atmosphere, the impact of the driving GCMs on the RCMs is smaller, compared to the core of the rainy season, when precipitation is mainly controlled by the large-scale circulation. In addition, we show that RCMs are able to counteract the bias received by their driving GCMs, hence, we claim that the cascade of uncertainty over SAF is not additive, but indeed the RCMs do provide improved precipitation climatologies. The fact that certain bias patterns over the historical period (1985–2005) identified in GCMs are resolved in RCMs, provides evidence that RCMs are reliable tools for climate change impact studies over SAF.
- Preprint
(6179 KB) -
Supplement
(10659 KB) - BibTeX
- EndNote
Maria Chara Karypidou et al.
Status: final response (author comments only)
-
RC1: 'Comment on gmd-2021-348', Anonymous Referee #1, 19 Mar 2022
The impact of lateral boundary forcing in the CORDEX-Africa ensemble over southern Africa by Karypidou et al
This manuscript attempts to answer the challenging question regards the extent to which RCM simulations can reduce biases present in GCM simulations for the regional climate of southern Africa. The manuscript is well-written with clearly documented methods, a well-justified aim, and appropriate figures.
There are a few minor and major queries or difficulties I had in understanding these results, which I detail here.
Major
- Variance analysis presented in Fig.8 and 13.: These results are crucial to the stated aims of the manuscript. However, I am left wondering to what extent the physical interpretation made in this analysis is undermined by the low number of RCM members (three) along with a sensitivity to total rainfall in the following way:
- Variance is constrained to be lower in October than January. Indeed much of southern Africa only experiences full onset of rains by November. So any intermodel variability is constrained by total rainfall. This seems reflected by Fig. 8 C and D, where GCM variance is higher in the respective region’s wetter months. Similarly, in October in the future [Fig. 13] all variance is very low, reflecting the lack of rain in this month.
- The behaviour set out in a. would be seen in a well-sampled system (e.g. 12 GCMs) but with 3 RCMs there the is a high risk that either they all look the same and there is no variance or there is one that is very different and the variance is substantial. This happens here as CCLM4 is biased dry in October with the others two biased wet.
- The behaviour in b. is strongly dependent on the behaviour of only one RCM. If that RCM comes in line with the others later in the season, as CCLM does with similar wet/dry bias patterns to the other RCMs in January, the variance is much lower. Indeed in region B and D it is surprisingly almost zero in some months.
- Part of this query may be unpicked if the rainfall is standardized per month in order to remove problem (a). I’m not sure if this will address b and c though.
- Based on my understanding of recent literature for the region, I disagree with the interpretations about the regional climate drivers.
- Munday & Washington 2018 demonstrated the heat low to tropical low switch of the Angola Low. Howard & Washington showed the tropical Angola low was in fact the monthly aggregrate of frequent tropical depressions crossing southern African from Mozambique and stalling in Angola. This is at odds with the interpretation provided in line 143-145, where the Angola Low is viewed separately.
- In the Heat Low phase, the Angola Low is not directly driving rainfall. It cannot because it can only develop under subsiding clear-sky conditions. Rainfall in the early season happens when the heat low is temporarily displaced/dissolved. The leading candidate for this displacement is large-scale synoptic westerly waves. See c.
- Work from the early 1990s by D’Abreton and picked up by others, including recently Hart et al 2018, suggested southern African rainfall is controlled by mid-latitude westerly wave dynamics (large-scale) earlier in the season. Thish then gives way to more local processes later in the season as the moist thermodynamic environment becomes more tropical (less subtropical) by the height of summer. Your speculation (Line 25-28) counters this, which is fine, but see d. below.
- Speculation about land-surface coupling seems key to your argument, but at least as far as I am aware this is not well-established for early season over southern Africa. Please include references which point to this if you have them available. I am not sure if the literature, as yet, has shown that for example the soil-moisture – rainfall coupling seen in Indian and the Sahel does play a role in southern Africa. And it is an open question whether this is true in the real-world, let alone whether it is resolved in RCMs.
Minor:
- Line 97-98, (line 300 too). Excess surface heating is surely even greater in peak summer months? Furthermore, heating is insufficient for convection when surface environments are moisture limited as they are during October.
- Line 208-209: I am not quite sure how to understand this statement about smooth topography. I read this to imply that the topography should be smooth, but is not the point of RCMs to include more detailed “jaggedness” that the real world topography contains?
- Line 358,359 linked to line 388. The only truly unambiguous signal, already well-established in literature, is this early season drying for southern Africa. So these statements about models struggling with transition (October) seem paradoxical with the clear drying signal. In the revision of the manuscript, hopefully this can be rethought and rewritten.
- Line 383 talks about observation products being kept in sight. I suspect you mean, in mind. Taking this quite literally, Kendon et al 2019 made this observation uncertainty visible by display TRMM-CMORPH bias alongside other bias or change plots (panel d in Figures 2-6). Including such a figure in your manuscript would help stay true to your line 383 statement and give the reader and indicate of magnitude of changes relative to obs. uncertainty.
- For clarity remove region letters in text and figures. Just go with SAF-All, Angola, East Coast, SAfr, or similar throughout to make for easier reading.
- Bias plots will be much clearer to interpret if in expressed % bias from climatology (with mask for negiligible rainfall areas, e.g. <1mm/month) and if colours were white for low/no bias increasing to dark red (dry) or blue (wet).
- A paper you may have missed by which is relevant to some of your analysis here is Munday, C., & Washington, R. (2019). Controls on the diversity in climate model projections of early summer drying over southern Africa. Journal of Climate, 32(12), 3707-3725.
References:
D’Abreton, P., and J. Lindesay, 1993: Water vapour transport over southern Africa during wet and dry early and late summer months. Int. J. Climatol., 13, 151–170, https://doi.org/10.1002/joc.3370130203.
Hart, N. C. G., Washington, R., & Reason, C. J. C. (2018). On the Likelihood of Tropical–Extratropical Cloud Bands in the South Indian Convergence Zone during ENSO Events, Journal of Climate, 31(7), 2797-2817. Retrieved Mar 19, 2022, from https://journals.ametsoc.org/view/journals/clim/31/7/jcli-d-17-0221.1.xml
Kendon, E.J., Stratton, R.A., Tucker, S. et al. Enhanced future changes in wet and dry extremes over Africa at convection-permitting scale. Nat Commun 10, 1794 (2019). https://doi.org/10.1038/s41467-019-09776-9
- Variance analysis presented in Fig.8 and 13.: These results are crucial to the stated aims of the manuscript. However, I am left wondering to what extent the physical interpretation made in this analysis is undermined by the low number of RCM members (three) along with a sensitivity to total rainfall in the following way:
-
RC2: 'Comment on gmd-2021-348', Anonymous Referee #2, 22 May 2022
This is an interesting paper assessing the impact of the GCM as compared to the RCM on a domain in Southern Africa. However, I think there are some concerns and questions that need to be addressed before the paper is ready for publication.
Variance between the models (RCMs or GCMs) has been employed to compare the model performance for the simulation of the monthly precipitation. In my reading, this study aims to show how much each RCM output is affected by its driving GCM. However, some concerns that may be critical should be addressed if the information provided in this study will eventually be used by both climate and non-climate scientists, as the authors mentioned.
• RCM output can be affected not only by the input GCM datasets but also by its parameterization, configuration, options, and setup. A simple comparison of the variance of monthly data between the complex models does not guarantee whether or not the impact of the driving GCM is dominant. All possible outcomes or limitations should be discussed to show the relationships between the RCMs and GCMs.
• How the relaxation zone has been defined to simulate each RCM should be addressed. Bias can be amplified through the lateral boundary conditions.
• The reasons why the three RCMs have been chosen should be addressed in the data section. If they have shown good performance in the domain, I suggest adding related references.Specific comments:
l57 - The authors should perhaps also cite the RCM lateral boundary papers focussing on Southern Africa authored by Ditiro Moalafhi, given the focus there is the impact of changes to lateral boundary conditions, with bias assessed based on a reanalysis dataset. These papers are:
1 Moalafhi, D. B., Sharma, A., Evans, J. P., Mehrotra, R. & Rocheta, E. Impact of biasâcorrected reanalysisâderived lateral boundary conditions on WRF simulations. Journal of Advances in Modeling Earth Systems (2017).
2 Moalafhi, D. B., Sharma, A. & Evans, J. P. Reconstructing hydro-climatological data using dynamical downscaling of reanalysis products in data-sparse regions–Application to the Limpopo catchment in southern Africa. Journal of Hydrology: Regional Studies 12, 378-395 (2017).
3 Moalafhi, D. B., Evans, J. P. & Sharma, A. Influence of reanalysis datasets on dynamically downscaling the recent past. Climate Dynamics 49, 1239-1255 (2017).
4 Moalafhi, D. B., Evans, J. P. & Sharma, A. Evaluating global reanalysis datasets for provision of boundary conditions in regional climate modelling. Climate Dynamics 47, 2727-2745, doi:10.1007/s00382-016-2994-x (2016).l89 - I would like to point the authors to papers on correcting lateral and lower boundary variables focussing on Australia, where the focus was the representation of drought, and the RCM used was WRF. While here the authors are using an ensemble of GCMs, these papers focussed more on what the representation of different attributes in the lateral and lower boundaries did to the overall monthly outcomes. These papers are:
1 Rocheta, E., Evans, J. P. & Sharma, A. Correcting lateral boundary biases in regional climate modelling: the effect of the relaxation zone. Climate Dynamics 55, 2511-2521, doi:10.1007/s00382-020-05393-1 (2020).
2 Rocheta, E., Evans, J. P. & Sharma, A. Can bias correction of regional climate model lateral boundary conditions improve low-frequency rainfall variability? Journal of Climate 0, null, doi:10.1175/jcli-d-16-0654.1 (2017).Of special interest to this study is the first paper that assessed the progressing deterioration in the corrections as one focussed deeper into the domain (away from the relaxation zone). I think the authors have missed this entire volume of work as I see no references to these papers. Please also note the different levels of impact lower versus lateral boundaries end up having on the simulations.
I also urge the authors to read the additional more recent papers:
1 Kim, Y., Evans, J. P., Sharma, A. & Rocheta, E. Spatial, temporal, and multivariate bias in regional climate model simulations. Geophysical Research Letters 48, e2020GL092058 (2021).
2 Kim, Y., Rocheta, E., Evans, J. P. & Sharma, A. Impact of bias correction of regional climate model boundary conditions on the simulation of precipitation extremes. Climate Dynamics 55, 3507-3526, doi:10.1007/s00382-020-05462-5 (2020).Here the focus was on extremes, which were found to be impacted to a greater extent by the lateral boundary corrections, than th e overall monthly attributes. Please also note the more recent of the two papers that attempted to quantify the impact of multivariate dependence bias in the lateral and lower boundaries, noting that the lack of this plays a significant role in the over quality of simulations.
I must confess that I am an author to the above papers, and leave it to the authors (and editor's) judgement whether my suggestions above are essential to the present study. However, in my reading of the current paper, I did feel the above referenced works do add to the story the authors are trying to tell, as they focus on a similar domain (Moalafhi) and altered lateral boundaries (all).
l172 - please clarify if this is for the mean or the monthly series.
l184 - there seems to be a mistake in the notation here, or also in equation 2. Please check. GCMi should refer to the variance of all RCMs driven by GCMi? Also, what is N?
l187 - similar confuision about the notation as earlier - please check and correct.
Figure 2 and 3 - It would have been nice to also show the observed climatology in this figure.
Remaining figures and ANOVA analysis - I found this quite comprehensive and interesting to read. I realise the authors are already reporting a lot of information here, but I was interested on their comments on the following:
(a) what was the impact on Temperature simulations and how the intra-RCM variances there compared against the precipitation?
(b) there is little focus on variability, although the main advantage a RCM brings is the added variability in both space and time. Could the authors comment on within grid variabilities across the RCMs, and change in variability in time at each grid cell?
(c) Our results showed significant impact on precipitation extremes (from altered lateral and lower boundaries). It would be interesting if the authors could comment on this aspect of the RCM simulations compared to the GCM simulations.
Maria Chara Karypidou et al.
Maria Chara Karypidou et al.
Viewed
| HTML | XML | Total | Supplement | BibTeX | EndNote | |
|---|---|---|---|---|---|---|
| 524 | 118 | 18 | 660 | 35 | 9 | 9 |
- HTML: 524
- PDF: 118
- XML: 18
- Total: 660
- Supplement: 35
- BibTeX: 9
- EndNote: 9
Viewed (geographical distribution)
| Country | # | Views | % |
|---|
| Total: | 0 |
| HTML: | 0 |
| PDF: | 0 |
| XML: | 0 |
- 1