You are here

Climate Change Dynamical Downscaling for South and Southeast Asia: Developing Training and Capacity

Body: 

Climate Change Dynamical Downscaling for South and Southeast Asia: Developing Training and Capacity

Part 1

  • Description of Global and Regional Climate Models
  • How to choose a Domain and set up a Present Day Control Run with WRF
  • Developing Climate Change Simulations

Part 2

Processing and Analyzing Model Output
Using the Model Output for Impacts Assessment
Construction of Regional Scenarios of Climate Change
Quality of Climate Data used as input in Models
Hands on Exercises

Part 3

  • Handling and Managing Regional Climate Model
  • Model Output
  • NCL (NCAR Command Language) & Graphics Sources of Climate Data
  • NCO (netCDF Operators)
  • Hands on Exercises

Training Outline

This training module will introduce students to: (1) how to set-up a regional climate model in order to dynamically-downscale large-scale global model simulations of climate change; (2) how to process and analyze the dynamical downscaling results for their specific regions of interest; (3) how to use the results in conjunction with impact assessment modeling (IAM) and other tools to understand and adapt to climate change in their region.

Learning Outcomes

On completion of this session the student should be able to:
1. Understand how to set up WRF for a domain of their choosing
2. Understand how to process and analyze the model output
3. Understand how to use the results in conjunction with impacts assessment

Training Framework

  1. Description of global and regional climate models
    How to choose a domain and set up a present-day control run with WRF
    Developing climate change simulations

    1. Choosing the global model
      Determining the time period
      Making control and experiment runs and comparing

  2. Processing and analyzing model output
    Using the model output for impacts assessment

Introduction/ Theory

Global climate models are the most powerful tool we currently have with which to understand and predict anthropogenic climate changes for the coming decades. Due primarily to computational limitations (these are very complex models that must be run on the most powerful supercomputers), these models can only be run at a fairly coarse spatial resolution, typically 150 km (IPCC AR4) or more recently 100 km (IPCC AR5). A key issue immediately arises: As stressed by IPCC, results at the global scale are useful for indicating the general nature and large-scale patterns of climate change, but not very robust at the local or regional scale (typically 4-12 km). Indeed, our experience, as well as that of many others, has shown that climate change projections from global models are spatially too coarse for most impact studies - this is especially true in regions of complex topography or land use, where resolutions of 4km or finer may be required. This issue is especially important for the complex, mountainous terrain of south and southeast Asia.

Some type of downscaling approach must be used, either statistical downscaling (which will be covered in other modules) or dynamical downscaling, which is the topic of this module.

Dynamical downscaling requires the use of a regional climate model. These models can be run at the very high resolution required to simulate local climate changes, but only for limited domains. Furthermore, while the global models can be run for centuries of simulated climate, the regional model can only sample a few years within that period. Because climate is inherently global in nature, a limited-domain model must be driven at its lateral boundaries by either observations or output from a global model. Output from a global model is required if scenarios of future climate change (e.g., due to greenhouse gas warming) are needed, while observations (in reality the proxy reanalyses) can be used if, for example, the effects on regional climate of changes in land use patterns is under investigation.

WRF Model Description

The regional climate model we use is the Weather Research and Forecasting (WRF) model [Skamarock et al. 2008], which has become widely used over the last few years as both a mesoscale weather model and a regional climate model. A technical description of the WRF model, as well as an online tutorial on setting up and using it and a User’s Guide, can be found at the WRF Users Page (http://www.mmm.ucar.edu/wrf/users/).

How Regions Such as South and Southeast Asia are Dynamically Downscaled

Whenever we work with a new region, the first step is always to make a (three-year) control run with WRF forced by reanalyses [to date, we have used the NCAR/NCEP global reanalysis to provide the necessary forcing, but the new NASA MERRA reanalysis could be utilized if it is not used for validation]. This control run can be expected to simulate actual weather events, and can therefore be evaluated using actual observations (e.g., from individual stations). The run identifies model strengths and weaknesses in simulating the climate of a given region, and hence can be used to identify model biases. The minimum length is three years; these can be either consecutive or non- consecutive, and longer is better, but these very high-resolution simulations are also very expensive for both computer and human resources.

Next we make a pair of ‘climate change’ simulations, forced by output from a global climate model [most typically, we have used the NCAR Community Earth System Model (CESM) to provide this large-scale forcing, but any CMIP5 model could be used]. These simulations include a ‘present-day baseline’ and a future climate scenario (typically, 50 years after the present-day baseline). Each simulation needs to be a minimum of five years in length. While it is possible to select non-consecutive, ‘representative’ years from time slices (e.g., 20-year periods) 50 years apart for this purpose, the use of consecutive years yields a better measure of interannual variability. The motivation for going out 50 years (even if more recent periods are of interest) is to generate a model climate change signal (i.e., response to the imposed transient forcing) that is large relative to the ‘noise’

(in this context climate variability on interannual to decadal time-scales). Once this ‘50- year’ climate change is identified at each point, it can be scaled back linearly to earlier periods (e.g., a 25 year change).

Theory, cont.

We are fully aware that the above constitutes the bare minimum necessary for reliable, useable results. Ideally, one would like ensembles of five year control and climate change runs, both from one model combination and from multiple models (while only a few regional climate models are currently in widespread use, the IPCC AR4 and AR5 reports are based on over 20 global climate models). Indeed, such a robust slate of simulations is possible but, even with state-of-the-art computational resources, only if the spatial resolution is fairly coarse (e.g., 25 km or greater). Our previous results suggest that anything coarser than 4-12 km (depending on topography and land surface type) is essentially worthless when it comes to predicting the change in key quantities, such as surface temperature and precipitation, at a resolution and accuracy that makes them useable to the stakeholders involved. Therefore, we advocate the use of relatively short, single-model scenarios with high spatial resolution over longer, multi-model, coarse- resolution ensembles. By determining, however, where in the ensemble spread the chosen forcing lies, we can provide a reasonable context for the downscaled scenarios.



Figure 1: Domains for the ongoing UNU/Control simulations. The entire figure comprises a 36km transitional domain (d01); d02 and d03 are 12km regional domains; and the remaining areas (d04-d07) are 4km domains over areas of special interest.

Practical Training on Using Downscaled Model Output

Participants will learn how to set up appropriate WRF domains, balancing resolution versus the size of the domain. They will then learn how to make a control run forced by observations (reanalyses) to evaluate how well WRF simulates the present-day climate of their region. Next, they will learn how to choose the global model output and use that to drive the regional model climate change simulations. Using the reanalysis simulation or the control and climate change simulations, the students will learn the crucial task of processing and analyzing the voluminous output produced by WRF, through hands-on exercises using appropriate tools and software packages.

Training Exercises

  1. Creating and plotting daily averages from RCM output
    Comparing reanalysis simulation results with observations creating monthly averages, climatologies and climate change differences plotting monthly averages

Suggested readings

IPCC, 2007: Climate Change 2007: The Physical Science Basis. Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change [Solomon, S., D. Qin, M. Manning, Z. Chen, M. Marquis, K.B. Averyt, M. Tignor and H.L. Miller (eds.)], Cambridge University Press, Cambridge, United Kingdom and New York, NY, USA, 996 pp. The next IPCC report (AR5) should be out late in 2012 or 2013 Skamarock, W.C., Klemp, J.B., Dudhia, J., Gill, D.O., Barker, D.M., Duda, M.G., Huang, Xiang-Yu, Wang, W. and Powers J.G., (2008) A Description of the Advanced Research WRF Version 3, NCAR Technical Note NCAR/TN475+STR. Oglesby, R. J., T. L. Sever, W. Saturno, D. J. Erickson, III, and J. Srikishen (2010), Collapse of the Maya: Could deforestation have contributed?, J. Geophys. Res., 115, D12106, doi:10.1029/2009JD011942.
Not all climate change is global in nature.

UN-CECAR Training Programme on Climate Change Downscaling Approaches and Applications Asian Institute of Technology, Thailand
9-20 November 2012

Day 1Afternoon Session
(Group 1- 9 Nov; Group 2- 10 Nov) Prof. Robert Oglesby, UNL

Dynamical Downscaling

  1. Description of Global and Regional Climate Models
    How to choose a Domain and set up a Present Day Control Run with WRF
    Developing Climate Change Simulations

Day 2Morning Session and Afternoon Session (Group 1- 10 Nov; Group 2- 11 Nov)
Prof. Clint Rowe and Prof. Robert Oglesby, UNL

Dynamical Downscaling

  1. Data for Climate Change Scenarios
    Processing and Analyzing Model Output
    a. ncdump b. NCL

    •   plotting model output (3-hourly)
        creating and plotting daily averages (Exercise 1)
        comparing to observations (Exercise 2)

o obtaining daily climatological data

o using local data c. NCO

  •   creating monthly averages, climatologies and differences (Exercise 3)
    plotting monthly averages (Exercise 4)

5. Using the Model Output for Impacts Assessment

a. Identifying end-user needs
b. Shared learning among researchers, practitioners (such as engineers

and water managers), and local stakeholders