Skip to content

Latest commit

 

History

History
313 lines (278 loc) · 17.6 KB

interpreting_regional_climate_projections.rst

File metadata and controls

313 lines (278 loc) · 17.6 KB

Interpreting regional climate projections

General remarks

In this section we highlight a few topics that are important to keep in mind when using climate change projections for practical applications. It might be necessary that the reader briefly reviews previous sections.

How to interpret small-scale structures?

There is still a debate to what extent small-scale structures at the spatial scale of individual climate model grid cells can and should be interpreted (e.g. Grasso 2000). On the one hand, several grid cells (typically more than four) are required to resolve atmospheric structures. Smaller scale variations are partly removed by the numerical filters of climate models in order to ensure numerical stability. Spatial smoothing is also frequently applied to the surface forcing fields of a climate model, in particular to orography in order to avoid steep topographic gradients and corresponding numerical instabilities. On the other hand, the surface forcing is grid cell specific and directly influences the simulation results over the individual grid cells. This is true for forcings such as topography, land use and soil type. Neighbouring grid cells, for instance, are typically located at different elevations which directly leads to differences in near-surface air temperature and to differences in the presence or absence of cryospheric features at the grid cell scale.

The most appropriate manner to analyse small-scale structures probably lies somewhere in-between the above described cases - i.e., interpretation of simulated atmospheric features at the grid cell scale should be avoided unless a direct and consistent influence of specific surface forcings can be expected.

Further reading

  • Grasso L. D. (2000), The differentiation between grid spacing and resolution and their application to numerical modeling. Bull. Amer. Meteorol. Soc. 81.3, pp. 579-580

How to interpret divergence between models?

Climate models differ in their model details as well as in the respective model setup (see chapter *What are limits of climate modelling?*). This explains differences in the output of different models even if the initialisation or the lateral boundary conditions of the model are the same. Further, differences in the results of regional models can be attributed to different regionalisation methods (e.g. dynamical vs. statistical downscaling).

It has to be noted that there is a priori no criteria for which chain of global to regional model is the best one. Therefore, the first recommendation is to treat all model combinations as equal. Only after an in-depth evaluation of all regional model results (see chapter *How can climate model simulations be evaluated?*) there may be clear indications to sort out particular models. This must be done consistently for all ensemble analyses.

How to interpret time and space variability?

Individual components of the climate system (including their interactions and feedbacks) operate on very different time scales. Variability of the climate system is observed at time scales ranging from intra-seasonal to inter-decadal (or even centennial and millennial) scale. Furthermore, local physical processes may induce large spatial variability of the climate characteristics, that typically appear more robustly over complex surfaces, e.g., mountainous or coastal regions.

The climate system is vastly complex, there is a number of governing processes that we have deficient knowledge about (e.g., turbulence, cloud microphysics, surface fluxes, aerosols). Due to this fact and the limited computational capacity, a certain level of simplification in climate models is inevitable. It means that climate models are able to provide information about climate variability only on limited spatial and time scale, depending on their resolution, described physical processes, and on their domain-size in case of regional simulations.

Global climate models currently have 100-500 km horizontal resolution. These models are dedicated to project the future climate change and its variability on larger time and spatial scale (e.g., large scale atmospheric or ocean internal modes of variability like NAO and ENSO, respectively). Due to such long-memory phenomena existing in the atmosphere and ocean, to distinguish climate change from climate variability, statistics of the meteorological conditions should be considered over at least a 30-year long period. Local processes may significantly alter the general large scale signals, as a result changes on smaller scale may be amplified or lessened, or even be in contrast with the global tendencies.

Detailed information about climate change and its variability for a smaller area (e.g., a country) can be obtained from regional climate models, which are applying 10-50 km resolution nowadays and are therefore describing dynamical processes in the atmosphere in more detail. However, it must be considered that the effective model resolution is at least 2-3 times coarser than the grid spacing, i.e., only phenomena with a characteristic size bigger than this effective resolution should be examined. Intra-annual or intra-seasonal variability, with special emphasis on climate extremes can be adequately estimated with the use of fine resolution regional climate models. However, it should be taken into account that going towards finer time and spatial scale of specific investigations, the model results tend to become noisy. Consequently, the ensemble approach for evaluation of the climate models is especially important as much as it is important to have the ensemble approach for projections about future climate characteristics.

In climate projections, main sources of uncertainties are (i) the internal variability, (ii) the scenario uncertainty and (iii) the model uncertainty. Based on Hawkins and Sutton (2009, 2011), it is concluded that model uncertainty is of great importance both for temperature and precipitation projections at all time scales. The choice of emission scenarios is relevant rather in temperature projections and on multi-decadal time scales. In precipitation projections, total uncertainty is basically composed of the internal variability and model uncertainty, especially when focusing on smaller regions.

Further reading

  • Hawkins, E., Sutton, R., 2009: The potential to narrow uncertainty in regional climate predictions. Bull. of Amer. Meteor. Soc. 90, 1095–1107, *https://doi.org/10.1175/2009BAMS2607.1*
  • Hawkins, E., Sutton, R., 2011: The potential to narrow uncertainty in projections of regional precipitation change. Climate Dynamics 37, 407–418, *https://doi.org/10.1007/s00382-010-0810-6*
  • Deque, M.; Somot, S.; Sanchez-Gomez, E.; Goodess, C. M.; Jacob, D.; Lenderink, G. & Christensen, O. B. The spread amongst ENSEMBLES regional scenarios: regional climate models, driving general circulation models and interannual variability Climate Dynamics, 2012, 38, 951-964, *https://doi.org/10.1007/s00382-011-1053-x*

How to interpret and adjust model biases?

There is a growing demand for regional climate change information for use in impact modelling, which in turn provides downstream inputs for decision-making. Such information generated by climate models has a number of uncertainties and these affect the ability of climate models to accurately simulate changes in the complex climate system. All models are only an approximation of the real climate system and have different simplifications resulting in biases of the simulated climate when compared to the observed one. It has been widely recognised that raw climate model output cannot always be used directly as input to, e.g., impact models. As a result an adjustment (also referred to as “bias correction”) towards the observed climatology is necessary. Alternatively, one may use results from ESD, which are calibrated against observations, provided its assumptions are justified and observation density is sufficient.

Nowadays, bias adjustment has become an integral part of the pre-processing of climate simulations for the use in impact modelling studies. However, bias adjustment is generally a statistical approach missing physical arguments, and applying bias adjustment to climate model simulations introduces a new often unexplored level of uncertainty. Moreover, often bias-adjusted simulations are ‘blindly’ used, even though their limitations are very well documented. In short, bias adjustment should be considered only as a statistical post-processing approach, while the reduction of model biases can only be done by continuous model development.

The two main questions regarding bias adjustment are: i) What in general can be bias adjusted and what not and ii) How can bias adjustment modify future climate projections? Most bias adjustment methods are based on the quantile mapping approach (e.g., Piani et. al. 2010) which generally provides very good results in terms of seasonal means and percentiles but does not take directly into account time-dependent statistics as for example consecutive dry/wet days (Addor and Seibert, 2014). Additionally it has to be noted that such a point wise approach is not supposed to correct spatial displacements of atmospheric phenomena such as the positioning of the simulated rain belt associated with the Inter Tropical Convergence Zone (ITCZ).

With respect to the second question it was also found that bias-adjusted climate simulations alter the projected climate change signals when compared to non-adjusted ones (Maurer and Pierce, 2014). A number of different approaches (modifications) are used to deal with this issue as for example the ISI-MIP method, which tries to preserve monthly mean trends (Hempel et al. 2013). However, future climate change can be expected to not only affect monthly means but also the different higher-order statistics (trends in extremes etc.).

Despite these very well known problems and fragmented recommendations, there are no systematic assessments of bias-adjustment-related uncertainties and no general guidance on the use of bias-adjusted climate simulations.

Bias-adjusted CORDEX simulations should be used carefully with full understanding of all potential limitations of the bias adjustment approach. It’s strongly recommended to read following report describing for what applications bias adjusted climate simulations can be used and for what not:

  • Breakout Group 3bis: Bias Correction (pp. 21-23) in IPCC, 2015: Workshop Report of the Intergovernmental Panel on Climate Change Workshop on Regional Climate Projections and their Use in Impacts and Risk Analysis Studies [Stocker, T.F., D. Qin, G. -K. Plattner, and M. Tignor (eds.)]. IPCC Working Group I Technical Support Unit, University of Bern, Bern, Switzerland, pp. 171. (*https://www.ipcc.ch/pdf/supporting-material/RPW_WorkshopReport.pdf*)

Further reading

  • Addor, N. and J. Seibert, 2014. Bias-correction for hydrological impact studies – beyond the daily perspective. Hydrol. Process., 28, 4823-4828, *https://doi.org/10.1002/hyp.10238*
  • Hempel, S., Frieler, K., Warszawski, L., Schewe, J. and F. Piontek, 2013 A trend-preserving bias correction – the ISI-MIP approach, Earth Syst. Dynam., 4, 219-236, *https://doi.org/10.5194/esd-4-219-2013*
  • Maurer, E. P. and D: W. Pierce, 2014. Bias correction can modify climate model simulated precipitation changes without adverse effect on the ensemble mean, Hydrol. Earth Syst. Sci., 18, 915-925, *https://doi.org/10.5194/hess-18-915-2014*
  • Piani, C., Haerter, J., and E. Coppola, 2010. Statistical bias correction for daily precipitation in regional climate models over Europe, Theor. Appl. Climatol., 99, 187–192, *https://doi.org/10.1007/s00704-009-0134-9*

How can climate change results be communicated?

Climate change results are in most cases communicated to persons who are not familiar with climate modelling. Therefore, it is necessary not only to deliver the results but explain what the results are based on and what the processing methods are. All results - text, tables, figures - must contain the full information. The necessary points are listed below.

  • Always give the full result information (scenario, global and regional models, time slice, region, spatial resolution...), in the case of climate change information the reference and the future period must be given. The user should be able to reproduce the steps taken based on the methods and data description.
  • Always communicate climate change result and uncertainty range
  • Differentiate scenarios - e.g. in case of low radiative forcing scenario point out that this can be used to highlight mitigation effects.
  • Make clear that the result is not a forecast but a projection.
  • Statements concerning trends, robustness and exceeding probability must be based on state-of-the-art analysis and statistical methods.
  • Climate change signals can be communicated as absolute differences or as relative differences. It depends on the meteorological parameter which value is suitable, e.g. relative changes are not sensible for temperature. For parameters like precipitation or wind speed it may depend on the context whether the climate changes is communicated as absolute or relative value, e.g. for low-wind regions an increase of 10 % in wind speed still may be a negligible change.
  • Interpretation of climate change results must take into account the information in the chapters above, particularly limits of modelling, suitability of the used data in time and space scales, bandwidth or probability statements.
  • Visualisation of results is an essential tool in communication, therefore the figures must be clear and not suggest wrong conclusions. Helpful hints are given in Kreienkamp et al., 2012. Some examples of visualisation of ensemble results are documented in Hennemuth et al., 2013, based on "How to read a climate-fact-sheet" *http://www.climate-service-center.de/imperia/md/images/csc/projekte/climatefactsheets/manual_cfs-update_march2016.pdf*

Further reading