CGE TRAINING MATERIALS - VULNERABILITY AND ADAPTATION ASSESSMENT CHAPTER 8 Climate Change Scenarios Objectives and expectations Having read this presentation, in conjunction with the related handbook, the reader should: Be familiar with key terms, concepts and an overview of climate change scenarios; Have a general understanding on the approaches for construction climate scenarios for impact assessment Be familiar with the concept of General Circulation Models (GCM) and Regional Climate Models (RCMs) and their advantages and limitations; Have a general understanding of available methods, tools and data sources necessary for generating climate scenarios. Outline What are climate change scenarios? Why we use scenarios? Climate change overview Approach to scenario development Methods, tools and data sources Future directions in scenario development Why Use Climate Change Scenarios? We are unsure exactly how regional climate will change Scenarios are plausible combinations of variables, consistent with what we know about human-induced climate change Think of them as the prediction of a model, contingent upon the greenhouse gas (GHG) emissions scenarios Estimates of regional change by models differ substantially, consequently, individual model estimates should be treated more as a scenario Scenarios help us to understand climate change impacts and determine key vulnerabilities They can also be used to evaluate and identify adaptation strategies Brief Primer on Regional Climate Change Temperatures over most land areas are likely to rise: Other factors, e.g., land-use change, may also be important Warmer temperatures mean increases in heat waves and evaporation Global-mean sea level rise: 0.18 to 0.59m by 2100 is based on IPCC 2007 findings: Modified by local subsidence/uplift Precipitation will change (increase) globally: Local changes uncertain: critical uncertainty Increase in storm intensity in some regions. Evolution of climate models GCM grid Projected Global Surface Warming Multi-model global averages of surface warming relative to 1980-1999 (Source: IPCC, 2007 WG I) Climate and IPCC Socio-economic Scenarios from Special Report on Emission Scenarios (SRES) Precipitation change Temperature change Climate model Had CM2 2050 (Source: Nakicenovic and Swart, 2000.) Projected Patterns of Precipitation Changes Relative changes in precipitation for the period 2090-2099, relative to 1980-1999 based on multi-model averages for A1B SRES scenario (Source: IPCC, 2007 WG I) What are Climate change scenarios? Climate change scenarios are tools to: Help envision how regional climates may change with increased greenhouse gas (GHG) concentrations To understand and evaluate how sensitive systems may be affected by human-induced climate change in the hope for policy-relevant information about expected changes and guidance for appropriate mitigation and adaptation measures It is critical to keep in mind that climate change scenarios are not a prediction nor forecast of future climate change. The use of regional climate change scenarios in a V&A assessment means: They must provide information on the climate variables needed for V&A at a spatial and temporal scale needed for analysis will require daily or even sub-daily spatial data What Are Reasonable Scenarios? Scenarios should be: Consistent with our understanding of the anthropogenic effects on climate Internally consistent: e.g. Clouds, temperature, precipitation Scenarios are a communication tool about what is known and not known about climate change: They should reflect plausible range for key variables. Approach to Climate Change Scenario Development Evaluate and determine the needs for climate scenario development Specify the baseline climate Develop climate change scenarios: Arbitrary scenarios, analogue scenarios, GCMs, regional climate models (RCMs), downscaling techniques etc. There is a range of existing guides to support scenario development process, which are available through UNDP/GEF: Lu (2007) http://www.undp.org/environment/docs/lecrds/applying_ climate_information.pdf Puma & Gold (2011) http://content.undp.org/go/cms-service/download/ publication/?version=live&id=3259633 Approach to Climate Change Scenario Development A range of existing guides to support scenario development process available through UNDP/GEF Evaluation and Determination of Needs for Climate-Scenario Development Choosing the right method for climate-scenario development can only be done after careful evaluation of the available approaches against the needs (application) and constraints (e.g. financial, computing, workforce, scientific, etc.) that project managers and their teams face Before embarking on a “fishing expedition” for data, models and tools, it is strongly advisable to allocate time to define clearly the scope of the climate scenario information needed within the framework of the national communication. Lu, 2006 Evaluation and Determination of Needs for Climate Scenario Development (Source: Puma and Gold, 2011 adapted from Lu, 2006) Identification of User Needs Using the UNDP Framework) The Framework provides the following: Helps decision makers identify their constraints (e.g. financial, computing, workforce, scientific, etc.) and understand the interplay among them, to better approach climate-scenario development, in particular with respect to resource allocation. Advices project managers to work together with a team of scientific and technical experts to manage uncertainties, select appropriate scenario methods and build a prospective range of scenarios A platform that fosters clear and frequent dialogue between team members: Scientific experts in charge of scenario development, are not fully aware of the managers needs and the non-scientific aspects of projects. Specification of the Baseline Climate Baseline climate is important to identify key characteristics of the current climate regime. Baseline climate data helps to identify key characteristics of the current climate regime (such as seasonality, trends and variability, extreme events and local weather phenomena) There are several questions that need to be answered to define the baseline climate: Which climate scenarios data are needed? Scale variability Which baseline period should be selected? WMO 30-year What data sources are available? Identification of Key data Sources Baseline climate is important to identify key characteristics of the current climate regime Key data sources available to define baseline climate include: National meteorological agencies archives Weather generators Climate model outputs Reanalysis data Outputs from GCM control simulations. Some Climate Data Sources IPCC data distribution center http://www.ipcc-data.org/ International Research Institute for Climate Prediction (IRI) http://iridl.ldeo.columbia.edu/docfind/databrief/cat-atmos.html Tyndall Centre for Climate Change Research http://www.cru.uea.ac.uk/cru/data/tmc.htm US National Oceanic and Atmospheric Administration (NOAA) http://www.esrl.noaa.gov/psd/data/gridded/ data.ncep.reanalysis.html Program for Climate Model Diagnosis and Intercomparison http://www-pcmdi.llnl.gov IPCC Data Distribution Centre The IPCC Data Distribution Centre is probably the best site for public-access climate model data Observed climate data 1901-1990 Gridded to 0.5 x 0.5° 10 and 30 year means (Source:http://ipcc-ddc.cru.uea.ac.uk/ http://sedac.ciesin.columbia.edu/ddc/) IPCC Data Distribution Center (continued) GCM data from CCC (Canada) CSIRO (Australia) ECHAM5 (Germany) GFDL-R30 (U.S.) HadCM3 (UK) NIES (Japan) IPSL (France) Can obtain actual (not scaled) GCM output IPCC Data Distribution Centre (continued) Contains monthly-mean data from GCMs on: Mean temperature (°C) Maximum temperature (°C) Minimum temperature (°C) Precipitation (mm/day) Vapour pressure (hPa) Cloud cover (%) Wind speed (m/s) Soil moisture. Observational Record National meteorological offices CMORPH (precip, satellite) Climate research Unit (CRU) of the University of East Anglia Réanalysis (ERA INT, NCEP) GPCC (gauge data) ISCCP (cloudiness, satellite) TRMM (precip, satellite) GPCP (precip, satellite and gauge) Development of Climate Change Scenarios Climate change scenarios need to be at a scale necessary for analysis: Spatial: e.g. to watershed or farm level Temporal: Monthly Daily Sub-daily. Options for Climate Change Scenario Development Past climates: analogues Spatial analogues Arbitrary changes; incremental Climate models. Past Climates Options: Instrumental record Paleoclimate reconstructions. Instrumental record: Pros: Can provide daily data Includes past extreme events Cons: Range of change in past climate is limited Data can be limited. Past Climates (continued) Paleoclimate reconstructions: From tree rings, boreholes, ice cores, etc. Can give annual, sometimes seasonal, climate Can go back hundreds of years Pro: Wider range of climates Cons: Incomplete data Uncertainties about values. Past Climates: Reconstruction of N. Hemisphere Temperatures (Source: Mann et al., 1998 ) Spatial Analogues Source: NAST, 2000. Spatial Analogues (continued) Pro: Communication tool: perhaps easier to understand Con: A model result must be used to choose the spatial analogue region Does not capture changes in variability. Arbitrary/Incremental Scenarios Assumes uniform annual or seasonal changes across a region for example: +2°C or +4°C for temperature +/-10% or 20% change in precipitation Can also make assumptions about changes in variability and extremes. Arbitrary/Incremental Scenarios (continued) Pros: Easy to use Can simulate a wide range of conditions Cons: It assumes a uniform change over the year or across a region and may fail to capture important seasonal or spatial details The combinations of changes in climate for different variables can be physically implausible. Climate Models Models are mathematical representations of the climate system. A model that incorporates the principles of physics, chemistry and biology into a mathematical model of climate, e.g. GCM or RCM (limited area model). Such a model has to answer what happens to temperature, precipitation, humidity, wind speed and direction, clouds, ice and other variables all around the globe over time They can be run with different forcings, e.g., higher GHG concentrations. Models are the only way to capture the complexities of increased GHG concentrations. General Circulation Models Pros Can represent the spatial details of future climate conditions for all variables Can maintain internal consistency Cons Relatively low spatial resolution May not accurately represent climate parameters Example Data from DDC – Temperature Example of GCM Output Example Data – Precipitation Program For Climate Model Diagnosis and Intercomparison (PCMDI) Has GCM Output Downscaling from GCMs Downscaling is a way to obtain higher spatial resolution output based on GCMs Options include: Combine low-resolution monthly GCM output with high-resolution observations Use statistical downscaling: Easier to apply Assumes fixed relationships across spatial scales Use regional climate models (RCMs): High resolution Capture more complexity Limited applications Computationally very demanding. Dynamical downscaling…From GCM to RCM GCM are lateral boundary conditions of RCMs or RCMs nudged in GCMs. With GCM acting as boundary conditions for an RCM, is it possible to represent regional climate with good accuracy? Combine Monthly GCM Output with Observations An approach that has been used in many studies. Typically, one adds the (low resolution) average monthly change from a GCM to an observed (high resolution) present-day “baseline” climate: 30 year averages should be used, if possible e.g. 1961-1990 or 1971-2000: Make sure the baseline from the GCM (i.e., the period from which changes are measured) is consistent with the choice of observational baseline. This method can provide daily data at the resolution of weather observation stations Assumes uniform changes within a GCM grid box and over a month: No spatial or daily/weekly variability. How Many GCM Grid Boxes Should Be Used? Using the single grid box that includes the area being examined would be ideal, however: There can be model noise at the scale of single grid boxes Many scientists do not think single box results are reliable. Hewitson (2003) recommends using 9 grid boxes: the grid being examined plus the 8 surrounding grid boxes. Need to consider the total area covered by all those grid boxes. Does it include topography or climates not similar to the area being studied? Do not use an isolated single location: it is better to do the analysis with group of stations; set of regional small scale indices. Find out whether the output is not a singular case or influenced by small size or the data sample? Or is it physically plausible and significant, meaning that you can reasonably develop predictions or real time applications? Statistical Downscaling Statistical downscaling is a mathematical procedure that relates changes at the large spatial scale that GCMs simulate to a much finer scale: For example, a statistical relationship can be created between variables simulated by GCMs such as air, sea surface temperature, and precipitation at the GCM scale (predictors) with temperature and precipitation at a particular location (predictands). There is a direct statistical relationship with sea-surface temperature (SST) indices (or other physically established predictor indices) Statistical downscaling from numerical model output is widely used in climate change downscaling from daily GCM fields “perfect prognosis” assumption. Downscaling principles Downscaling as the process of making the link between the state of some variable representing a large space (or “large scale”) and the state of some variable representing a much smaller space (or “small scale”) (Benestad, 2002) There are two main approaches to downscaling: dynamical and empirical–statistical Dynamical downscaling makes use of limited area models (RCMs) with progressively higher spatial resolution than the global climate model (GCM). Statistical downscaling is an extraction of information about statistical relationships between the large-scale climate and the local climate. Downscaling Methods Do not use isolated single location. Find out whether the output is a singular case or influenced by small size or the data sample? Or is it physically plausible and significant, meaning that you can reasonably develop predictions or real time applications? It is better to do the analysis with: A group of stations Or a set of regional small scale indices Or time series of a grid at high resolution. Statistical Downscaling (continued) Is most appropriate for: Subgrid scales (small islands, point processes, etc.) Complex/heterogeneous environments Extreme events Exotic predictands Transient change/ensembles Is not appropriate for data-poor regions Where relationships between predictors and predictands may change Statistical downscaling is much easier to apply than regional climate modelling. In climate change studies, one important question is what implications a global warming has for the local climate. The local climate can be regarded as the result of a combination of the local geography (physiography) and the large-scale climate (circulation) local climate, y = f(X, l, G) where X = Regional climate; l = local geography; G = Global climate Statistical Downscaling (continued) Statistical downscaling assumes that the relationship between the predictors and the predictands remains the same. Those relationships could change. In such cases, using regional climate models may be more appropriate Four necessary conditions must be fulfilled in Empirical-Statistical Downscaling (ESD): Strong relationship Model representation Description of change Stationarity. Spatial structure of precipitation from radar reflection and a typical size of an RCM grid box, showing spatial variations at scales smaller than the model’s spatial resolution. Statistical Downscaling Model (SDSM) Currently, this is only feasible based on outputs from a few GCMs. Global Data to Use in Downscaling with SDSM Canadian website with Global Data: Go to scenarios, then SDSM Get output for individual grid. Regional Climate Models (RCMs) These are high resolution models that are “nested” within GCMs: A common grid resolution is 50 km: Some are higher resolution RCMs are run with boundary conditions from GCMs They give much higher resolution output than GCMs Hence, much greater sensitivity to smaller scale factors such as mountains, lakes Good to investigate higher order climate variability. RCM Limitations Can correct for some, but not all, errors in GCMs Typically applied to one GCM or only a few GCMs In many applications, just run for a simulated decade, e.g., 2040s Still need to parameterize many processes May need further downscaling for some applications Needs diagnostics based on known weather and climate features RCM evaluation limited by (observations) data availability. GCM vs. RCM Resolution Extreme Precipitation (JunJulAug) Observation RCM GCM Extremes Intensity of storms sensitive to model resolution Higher resolution improves intensity of precipitation Higher resolution improves intensity location. By Now You May Be Confused… So many choices, what to do? First, let’s remember the basics: Scenarios are essentially educational tools to help: See ranges of potential climate change Provide tools for better understanding the sensitivities of affected systems. So, we need to select scenarios that enable us to meet these goals. Selected Methods and Tools MAGICC/SCENGEN PRECIS SDSM (Statistical Downscaling Model) ASD (Automated Statistical Downscaling) ClimateWizard Clim.pact (R package) http://www.cru.uea.ac.uk/projects/ensembles/ ScenariosPortal/Downscaling2.htm Climate Explorer SimCLIM Tools for Assessing Regional Model Output It is useful first to compare results from a number of GCMs that might be used to drive an RCM Normalized GCM results allow comparison of the relative regional changes Can analyse the degree to which models agree about change in direction and relative magnitude: A measure of GCM uncertainty. Tools for Assessing Regional Model Output (continued) Agreement between GCMs does not necessarily mean that they are all correct – they may all be repeating the same mistakes. Still, GCMs are the primary tool for estimating the range of future possibilities. Normalizing GCM Output Expresses regional change relative to an increase of 1C in mean global temperature (GMT): This is a way to avoid high-sensitivity models dominating results It allows us to compare GCM output based on relative regional change. Normalized temperature change = ΔTRGCM/ΔTGMTGCM Normalized precipitation change = ΔPRGCM/ΔTGMTGCM Pattern Scaling Is a technique for estimating change in regional climate using normalized patterns of change and changes in GMT. Pattern scaled temperature change: ΔTRΔGMT = (ΔTRGCM/ΔTGMTGCM) x ΔGMT Pattern scaled precipitation: ΔPRΔGMT = (ΔPRGCM/ΔTGMTGCM) x ΔGMT Tools to Survey GCM Results Finnish report: “Future climate . . .” MAGICC/SCENGEN Finnish Publication Shows regional output on temperature and precipitation for a number of models: For three time slices over 21st century Uses some scaling Useful as a look-up to see degree of model agreement or disagreement. MAGICC/SCENGEN and COSMIC provide more flexibility to users. Finnish Environment Example Source: Ruosteenoja et al., 2003, p. 55. Alma Jean - Source extracted from notes page MAGICC/SCENGEN MAGICC is a simple model of global T and sea level rise (SLR) Used in IPCC TAR SCENGEN uses pattern scaling for 17 GCMs Yield: Model by model changes Mean change Intermodel standard deviation Interannual variability changes Current and future climate on 5 x 5 grid (Source:http://www.cgd.ucar.edu/cas/wigley/magicc/) Alma Jean - Source taken from notes pages, could you please check the format?....also applies to some of the slides that follow until 82....once this is edited, I gues cut and paste is our best friend!!!! Using MAGICC/SCENGEN MAGICC: Selecting Scenarios MAGICC: Selecting Scenarios (continued) MAGICC: Selecting Forcings MAGICC: Displaying Results MAGICC: Displaying Results (continued) Running SCENGEN Running SCENGEN (continued) SCENGEN: Analysis SCENGEN: Model Selection SCENGEN: Area of Analysis SCENGEN: Select Variable SCENGEN: Scenario SCENGEN: Map Results SCENGEN: Quantitative Results INTER-MOD S.D. : AREA AVERAGE = 5.186 % (FOR NORMALIZED GHG DATA) INTER-MOD SNR : AREA AVERAGE = -.067 (FOR NORMALIZED GHG DATA) PROB OF INCREASE : AREA AVERAGE = .473 (FOR NORMALIZED GHG DATA) GHG ONLY : AREA AVERAGE = -.411 % (FOR SCALED DATA) AEROSOL ONLY : AREA AVERAGE = -.277 % (FOR SCALED DATA) GHG AND AEROSOL : AREA AVERAGE = -.687 % (FOR SCALED DATA) *** SCALED AREA AVERAGE RESULTS FOR INDIVIDUAL MODELS *** (AEROSOLS INCLUDED) MODEL = BMRCD2 : AREA AVE = 2.404 (%) MODEL = CCC1D2 : AREA AVE = -5.384 (%) MODEL = CCSRD2 : AREA AVE = 6.250 (%) MODEL = CERFD2 : AREA AVE = -2.094 (%) MODEL = CSI2D2 : AREA AVE = 6.058 (%) MODEL = CSM_D2 : AREA AVE = 1.245 (%) MODEL = ECH3D2 : AREA AVE = .151 (%) MODEL = ECH4D2 : AREA AVE = -1.133 (%) MODEL = GFDLD2 : AREA AVE = 1.298 (%) MODEL = GISSD2 : AREA AVE = -3.874 (%) MODEL = HAD2D2 : AREA AVE = -5.442 (%) MODEL = HAD3D2 : AREA AVE = -.459 (%) MODEL = IAP_D2 : AREA AVE = -.088 (%) MODEL = LMD_D2 : AREA AVE = -6.548 (%) MODEL = MRI_D2 : AREA AVE = .065 (%) MODEL = PCM_D2 : AREA AVE = -3.451 (%) MODEL = MODBAR : AREA AVE = -.687 (%) SCENGEN: Global Analysis SCENGEN: Error Analysis SCENGEN Error Analysis (continued) UNWEIGHTED STATISTICS MODEL CORREL RMSE MEAN DIFF NUM PTS mm/day mm/day BMRCTR .632 1.312 1.026 20 CCC1TR .572 1.160 -.207 20 CCSRTR .587 .989 .322 20 CERFTR .634 1.421 -1.167 20 CSI2TR .553 1.112 -.306 20 CSM_TR .801 1.044 -.785 20 ECH3TR .174 1.501 -.649 20 ECH4TR .767 1.121 -.881 20 GFDLTR .719 .954 -.553 20 GISSTR .688 .799 .123 20 HAD2TR .920 .743 -.598 20 HAD3TR .923 .974 -.883 20 IAP_TR .599 1.408 -.734 20 LMD_TR .432 2.977 -2.103 20 MRI_TR .216 2.895 -2.026 20 PCM_TR .740 1.372 -1.041 20 MODBAR .813 .879 -.654 20 PRECIS (Providing Regional Climates for Impacts Studies) Developed by the UK Met Office Hadley Centre Designed to run on PC with Linux Provides high resolution regional climate projections of all key climate variables with hourly, daily, monthly and yearly means Significant computational requirements. A PRECIS experiment can take several months to run on a standard level PC PRECIS Minimum hardware requirements PC running Linux operating system Memory: 512MB minimum; 1GB+recommended Minimum 60GB disk space + offline storage for archiving data Simulation speed proportional to CPU speed How fast does it go? 30 year integration, 100x100 grid points NEC (supercomputer processors): 5 days (1 node with 8 processors) PC (Intel Pentium 4, 3.2GHz processor): 2.5 months (1 processor) PRECIS: Outputs RCMs can provide: Climate scenarios for any region An estimate of uncertainty due to different emissions An estimate of uncertainty due to different GCMs An estimate of uncertainty due to climate variability. Data available from RCMs: Comprehensive for atmosphere and land-surface Grid-scale box average quantities Maximum time resolution one hour. PRECIS: Outputs GCM RCM: PRECIS PRECIS: Support Detailed climate scenarios using the UKCIP02 methodology for the main developing country regions Detailed simulation for the recent climate (past 50 years) for many developing country regions Basic capacity-building and technology transfer enabling mitigation and adaptation activities via: Scientific and technical support for applying PRECIS to scenario development and climate research Ad hoc advice on using scenarios in impacts assessment, developing collaborations and research proposals. PRECIS: Selected Applications PRECIS-Caribbean initiative Vulnerability and Adaptation in Cuba IIT India University of Cape Town, South Africa Uganda Study of Climate Change Impact in China Impacts, Vulnerability and adaptation to climate change in Latin America Climate change Assessment for Esmeraldas, Ecuador Georgia Second National Communication Climate Change Assessment for Sorsogon, Philippines SimCLIM: What is SimCLIM? SimCLIM is a research product from New Zealand Climate Change Impacts Studies (CLIMPACTS) An integrated computer model for climate change impact assessment Built-in customized geographic information system (GIS) to support multi-level spatial analysis Based on IPCC guidelines and upgradeable with latest scientific research information. SimCLIM: Applications Describe the baseline climate Examine current climate variability and extremes Assess risks – present and future Investigate adaptation – present and future Create climate change scenarios Conduct sensitivity analyses Examine sectoral impacts Examine uncertainties Facilitate integrated impacts analyses. SimClim: Extreme Event Analyser Using SimCLIM extreme Event Analyser and daily time-series data to calculate change in extreme temperature. This example shows a present day return period of 46.52 years for a 39°C event. Under the SRES A2 scenario at 2050 the return period for the same temperature event becomes 13.24 years, providing an example of how extremes will become more frequent under climate change for this particular location. SimCLIM: User Interface SimCLIM: Where It Has Been Used Bahamas Barbados China Cook Islands Fiji Indonesia Maldives Marshall Islands Mongolia Nauru Peru Phillipines Samoa Solomon Islands Tonga Trinidad and Tobago Tuvalu Vanuatu Vietnam ClimateWizard Developed by The Nature Conservancy, University of Washington and University of Southern Mississippi Useful screening tool to provide instant national-level snapshot of baseline temperature and precipitation as well as projections for SRES scenarios to 2050 and 2080. GCM downscaled to 50km Grid http://climatewizard.org/ ClimateWizard How to Select Scenarios Use one or several of the methods and tools to assess the range of temperature or precipitation changes. Models can be selected based on: How well they simulate current climate SCENGEN has a routine How well they representing a broad range of conditions. 1A.* How to Select Scenarios (continued) Use results from actual GCM data or scaled data Can include other sources for scenarios, e.g. arbitrary, analogue. Selecting GCMs Some factors to consider in selecting GCMs Age of the model run: More recent runs tend to be better, but there are some exceptions Model resolution: Higher resolution tends to be better Model accuracy in simulating current climate: MAGICC/SCENGEN has a routine. What to Use under What Conditions? There is nothing wrong with using combinations of different sources for creating scenarios, e.g. models and arbitrary scenarios The climate models tend to be better for longer run analyses, e.g. beyond several decades (beyond 2050) Climate analogues tend to be better for near term, e.g. within several decades (2010-2030). Scenarios for Extreme Events Difficult to obtain from any of these sources. Options: Use long historical or paleoclimate records Incrementally change historical extremes: Try to be consistent with transient GCMs These methods are primarily useful for sensitivity studies. Final Thoughts Remember that individual scenarios are not predictions of future regional climate change. If used properly, they can help us understand and portray: What is known about how regional climates may change Uncertainties about regional climate change The potential consequences. Uses If assessing vulnerability, scenarios ought to reflect a wide, but realistic range of climate change: Serves education purpose If examining adaptation, it is important to reflect a wide range of climate change If the selected uncertainty range is too narrow, this could lead to ill-informed decisions. Climate and IPCC (recent and coming activities) IPCC(2012) has published also a Special report on Managing the Risks of Extreme Events and Disasters to Advance Climate Change Adaptation. The IPCC Fifth Assessment Report (AR5) process is now underway. The AR5 will consist of three working group reports and a synthesis report to be completed in 2013/2014 Climate Change Scenarios Exercise Aim: To explore different climate change scenarios methods and tools Method: Define your region of interest Choose your climate parameter Choose tool or tools or models outputs Assess anomalies for current climate conditions Consider emission scenarios or concentration pathways Develop future climate scenarios, projections at different time horizon (2030, 2050 and 2100) Examine issues of: GCM, RCM or statistical methods Stationnarity Resolution (higher vs coarse) Trends and variability. How does one address the main gaps (data, tools, …) in respect to available means and work to undertake when looking at climate change scenarios? Presentation title Presentation title Presentation title * This is a brief overview of a very complicated and rich topic. Analysts are encouraged to read more about this. Those interested in reading more on scenario development should read Chapter 13: Climate Scenario Development (http://www.grida.no/climate/ipcc_tar/wg1/pdf/TAR-13.PDF). Presentation title * This slide gives reasons for using scenarios of climate change. Presentation title * All the climate models project increased temperatures over land. Note that other factors, such as land-use change, can have a substantial effect on climate at a regional scale. Higher temperatures will result in increased evaporation and increased likelihood of more intense and frequent heat waves. A rise in sea level is likely. For analysing what happens at the local level, the analyst must account for local coastal subsidence or uplift. Precipitation will increase globally, but there are uncertainties about changes at the regional scale. Indeed, a lot of what we will cover in the scenarios tries to address capturing this uncertainty about regional precipitation change. Storm intensity is likely to rise, but intensity will not necessarily increase among all storms or in all places. Presentation title * Presentation title * Presentation title Climate change scenarios are derived from global climate models (GCMs) driven by changes in the atmospheric composition that in turn is derived from socio-economic scenarios from the IPCC Special Report on Emission Scenarios (SRES). The figure shows as an example map of temperature and precipitation projections derived from the HadCM2 climate model of the UK Meteorological Office. A main challenge is to interpret the results derived from the climate scenarios used as inputs. In all regions, uncertainties about the magnitude of the expected changes result in uncertainties of the agricultural evaluations. For example, in some regions projections of rainfall, a key variable for crop production, may be positive or negative depending on the climate scenario used. The uncertainty derived from the climate model is related to the limitation of current models to represent all atmospheric processes and interactions of the climate system. The limitation of projecting the socioeconomic development pathways is an additional source of uncertainty. The limitations for projecting socio-economic changes not only affect the SRES scenarios but also the potential adaptive capacity of the system. For example, uncertainty of the population (density, distribution, migration), gross domestic product, technology, determine and limit the potential adaptation strategies. Presentation title * It is critical that climate change scenarios are consistent with our understanding of climate change and are internally consistent. On the latter, this means that the changes make meteorological sense. For example, it would not be plausible for cloudiness to decrease while precipitation is decreasing. It is also critical that users recognize that scenarios are a communication tool. People may interpret scenarios as reflecting what we know about the future. So, scenarios should communicate uncertainties as appropriate. If we are confident that temperatures will rise, we should not have a scenario with cooler temperatures. On the other hand, if we do not know whether precipitation will increase or decrease, then the scenarios should include increases and decreases in precipitation. Presentation title * Presentation title * National meteorological offices are the best places to start looking for observations. Presentation title * For scenarios to be useful for analyses of vulnerability and adaptation, they have to provide information on a spatial scale compatible with impacts analysis. This could be at the scale of a watershed or farm. They also need to be at the appropriate time scale. Some analyses may need only monthly data. Others may require daily or even subdaily data. Presentation title * These are the four main options for creating climate change scenarios. Presentation title * On past climates, the instrumental record is the observed climate, i.e. the recordings of temperature and precipitation that can stretch back decades. Note that these records may have many gaps or may not go far back in time. It is best to gather as long a record as possible. Paleoclimate reconstructions are based on proxy data, such as tree rings, sediments in lakes, or boreholes. They typically estimate climate conditions over the past centuries to millennia. Presentation title * The estimates of change in climate from paleoclimatic reconstructions can cover many more years than the instrumental record and could include more extreme events. But, the accuracy of paleoclimate reconstructions is far lower than that of the instrumental record. Presentation title * This is an example of a reconstruction of paleoclimate temperatures. It was constructed by Mann and colleagues (Houghton et al., 2001) and represents average Northern Hemisphere changes. Paleoclimate reconstructions rely on proxy data to estimate past temperatures and precipitation. They used such proxies as tree rings, bore holes and ice cores. Using proxies to estimate past climate has uncertainties and these techniques provide only approximations of past climates. Note that reconstructions other than the Mann et al. case shown here do not necessarily come up with the same past climates. For example, some show warmer temperatures during some past periods. All paleoclimate reconstructions of hemispheric-scale temperatures, however, show the 20th century being as warm if not warmer than any period in the past millennium. Presentation title * Spatial analogues are current climates in other locations that may be representative of the future climate of a specific location. The example here is for the future climate of Illinois. A warmer and wetter climate (as suggested by the Hadley Centre model) makes Illinois approximately like current Virginia by 2030 and like current eastern North Carolina by 2090. A warmer and drier climate (as suggested by the Canadian model) makes Illinois like Missouri by 2030 and like Oklahoma by 2090. These examples demonstrate that analogues can be good and bad as communication tools. They paint a simple picture, which is good. But future climate is not likely to be exactly like climate in another region. Furthermore, would the wetter climate of the 22nd century in the Hadley model require a spatial analogue in the ocean? So, spatial analogues should be used purely for illustrative purposes. Presentation title * Spatial analogues are attractive as a communication tool because of their simplicity. They involve assuming the climate in one location may become like the climate in another location. That may be much easier to understand than assuming that temperatures could rise by 1°C. It is unlikely, however, that the climate in one location will become just like the climate in another location. Future climates could be quite different, and there may be no place where the future climate of a region exists today. Presentation title * Arbitrary scenarios are typically made up by analysts, but should be done so only after examining model projections of regional climate change and consulting climatologists. Presentation title * Arbitrary scenarios can be the easiest to apply. Note they often do not capture variance across space or time. Also, some combinations of variables can be implausible. Presentation title * Climate models are the most sophisticated options for developing climate change scenarios, but can also be complicated to use. Because they are run assuming higher greenhouse gas (GHG) concentrations, they are the only option of the four that explicitly incorporatesclimate change conditions resulting from increased GHG concentrations. Global climate models range from simple, one-dimensional models such as MAGICC, which is briefly described in later slides, to more complex three-dimensional models such as general circulation models (GCMs). Presentation title * GCMs model the atmosphere, ocean, the cryosphere and interactions with land surfaces. They model change on a regional scale, typically estimating change in grid boxes that are approximately several hundred kilometers wide. GCMs provide only an average change in climate for each grid box, even though real climates can vary quite considerably within several hundred kilometres. Presentation title * These maps can be downloaded from the DDC. Presentation title * The HadCM3 model (from the United Kingdom Hadley Centre) is at 2.5 x 3.8 degrees spatial resolution. The slide displays regional estimates of the increase in April temperature (°C) by 2080 relative to 1961-1990 average temperatures using the SRES A2 scenario. Note that while some of the largest increases in temperature are in high latitude areas such as Canada and northern Russia, there are also relatively large increases in temperature in northwestern Mexico, the Amazon, the Sahel, southern Africa, and many parts of Asia. Presentation title * Not all PCMDI data are publicly available. Some data sets require projects that plan to use them to be registered and approved by a selection committee. Presentation title * Typically, GCM resolution is too low for impacts work and some form of downscaling is needed. By downscaling, we mean transforming GCM output to a scale useful for impacts analysis. The options are discussed in the following slides. Presentation title * Presentation title * A most common way to downscale is to combine GCM output with a sufficiently long observed data set. The data should be at a time scale necessary for analysis. If only monthly data are needed, the observed data set need only have monthly data (although weekly or daily data can always be averaged to produce monthly data). Try to obtain a 30-year observed climate record, if possible; for example, daily data from 1971-2000. The choice of a baseline climate is important; the most recent record is preferred. However, in the 1990s temperature increased substantially in many areas. An increase from the GCMs would then be combined with increases in the baseline. This may be sufficient if only average statistics are being used. Otherwise, an earlier period, e.g. 1961-1990, might be preferable. Add the average monthly change from the GCM to the observations; e.g. if January is estimated to be 2°C warmer, then add 2°C to all the daily January observations in the observed record. For precipitation, use a percentage change. Add it (a fraction, i.e. divide by 100) to 1 and multiply the appropriate daily observations for that month by this number. So, if January precipitation is estimated to increase by 10%, multiply 1.1 (1 + 10/100) by the January observed record. Days that have precipitation in the observed record will have 10% more and days with no precipitation will have no precipitation. If February is estimated to decrease by 15%, then multiply 0.85 (1 - 15/100) by the observed February record. This yields a 30-year scenario that is the result of combining average monthly changes from the GCM with observations. Daily temperature and precipitation are changed. Seasonality can also be changed because month by month changes from the GCM will most likely differ across months. Note that daily variability does not change because each day is being changed the same amount. Also, year-to-year variability does not change because specific months in each year are being changed the same amount. Although applying monthly means is the most commonly used approach, one could use other GCM output such as data on changes in month-to-month variability. Presentation title * There is no easy answer to how many grid boxes should be used. Many studies use single grid box output. Some criticize these because there is much less reliability associated with single grid box output than with the average output over a number of grid boxes. Hewitson, 2003, recommends using 9 grid boxes: the grid box containing the area being studied and the eight surrounding grid boxes. This smoothes out GCM projections. Alternatively, the analyst can examine the climate and topography covered by the grid box that the study area is in and surrounding grid boxes. Include neighbouring grid boxes with similar climate and topography. If the area is on a plateau and is relatively dry, nearby grid boxes also on the plateau and relatively dry can be included. A similar point can be made about using monthly data. Changes in monthly climate can also have some noise (high variance). This can be solved by averaging multiple months (e.g. from a season) together. Presentation title * Statistical downscaling is a technique that allows for production of very high resolution (spatially or temporally) scenarios. A key point about statistical downscaling is that it assumes that the mathematical relationship between the climate at the spatial scale simulated by GCMs and local climate does not change. Presentation title * Statistical downscaling is more appropriate for some situations than others. It is most appropriate for small scales, complex environments (e.g. mountains, coastal areas, areas with diverse microclimates), simulating extreme climate events, exotic predictands (variables that are not typically available from climate model output that is saved) and transient changes (changes over time). Developing statistical downscaling relationships generally requires several decades of daily data. Relationships are usually developed using gridded observed data from reanalyses (for the predictors; simulating the gridded data output from a GCM) and local (station level) observed data (the predictands). Presentation title * If the analyst believes the mathematical relationships between GCM-scale predictors and the local-scale predictands may change, then standard deviation may not be so appropriate. These relationships are usually controlled by topographic influences, however, most of which are unlikely to change over time. Presentation title * Guidance: http://ipcc-ddc.cru.uea.ac.uk/guidelines/StatDown_Guide.pdf Data are available only from a few GCMs to apply SDSM. Presentation title * http://www.cics.uvic.ca/scenarios/index.cgi? Scenarios provide grid box level output from HadCM3. The software also includes reanalysis data that can be used with observed data from weather stations to derive downscaling relationships. Then, gridded HadCM3 data can be used as the predictors in these relationships to estimate future climate at the station level. The SDSM guidance document (http://ipcc-ddc.cru.uea.ac.uk/guidelines/StatDown_Guide.pdf) provides step-by-step instructions on how to downscale the GCM projections. Presentation title * RCMs are much higher resolution models that focus on a region, typically at a continental or subcontinental scale. Their grid boxes are 50 km or less across. They are therefore able to capture many regional features that GCMs cannot. However, RCMs must be run with boundary conditions from GCMs (e.g. changes in pressure patterns), so there are typically RCM runs for only a few GCMs. Some applications are for limited periods of time, e.g. a simulated decade. The advantage of RCMs is that they can provide better spatial representation of climate change than GCMs, but they cannot correct for errors in boundary conditions. Presentation title * Their application tends to be limited. If an RCM is driven by just a few GCMs, this may not represent the range of potential regional climate changes. They may simulate only a limited amount of time in the future. They do not necessarily simulate all processes well. Even though they are on a small spatial scale, further downscaling may be necessary to capture even finer resolution climatic conditions. Presentation title * RCMs give much finer resolution.This can be seen in a complex landscape such as in the western U.S. This figure compares output from the Parallel Climate Model (PCM) GCM with the MM5 RCM. Source: Ruby Leung, Battelle – Pacific Northwest National Laboratory. Presentation title * Summertime extremes in precipitation are difficult to reproduce because they are driven by very small-scale processes. In some cases, errors in the large-scale circulation in the GCM may be amplified in simulating extreme precipitation in the RCM. In general, RCMs cannot be relied on to correct for errors from GCMs. Source: Ruby Leung, Battelle - Pacific Northwest National Laboratory. Presentation title * This slide is to help users better understand what can be a confusing set of choices. Presentation title * This slide reviews tools that can help users to understand the range of change in regional climates simulated by climate models. Normalization expresses regional change in climate relative to global-mean temperature change simulated by models. It is explained in the next slide. Presentation title * Normalizing GCM results involves determining the relative change in regional climate per standard change in global-mean temperature (GMT), e.g. per 1°C increase in global mean temperature. ΔTRGCM is the regional change in temperature from a GCM. ΔTGMTGCM is the global-mean temperature change in the same GCM. So the formula is the change in regional temperature relative to the change in the global-mean temperature for a GCM. The formula “normalizes” the relative change in temperature. The third bullet is the formula for normalizing precipitation change. ΔPRGCM is the percentage change in regional precipitation. It is normalized by being expressed as a percentage change per degree Celsius increase in GMT. Presentation title * Pattern scaling allows us to estimate regional changes in climate using normalized values from single GCMs, or combinations of GCMs. Pattern scaling is based on the assumption that the underlying patterns of climate change are similar (in a given model) no matter what the change in GMT. The normalized values for a region are multiplied by the change in GMT to estimate the change in regional climate for that amount of global warming. There is more confidence in pattern scaling of temperature than of precipitation. The second bullet is the formula for calculating pattern-scaled regional temperature changes. The term in the parentheses is explained in the previous slide. The final expression (ΔGMT) is the assumed change in global mean temperature. The product of ΔTRGCM/ΔTGMTGCM and ΔGMT is ΔTRΔGMT – the pattern-scaled change in regional temperature. The third bullet is for precipitation. Precipitation is expressed as % change per degree increase in global mean temperature. P is precipitation. Presentation title * Presentation title * Ruosteenoja et al., 2003, p. 84. Available at http://www.environment.fi/default.asp?contentid=103483&lan=EN. Presentation title * 2040-2069 Southern South America, December to February, relative to 1961-1990. Each point is from an AOGCM. Scenario corresponds to colour in lower left. Orange is CGCM2 model and blue is HadCM3. Filled in symbols are from model output; open symbols are from scaling. Ovals in center are 95% confidence interval ellipses for natural variability as estimated by GCM simulations of climate without increased GHG concentrations (obtained by running the model for a simulated 1,000 years without external forcing). Note that GCM output outside of the ovals does not mean it is significantly different from zero for all variables. All the model output falls to the right of the ovals on temperature. That means all the models show an increase in temperature greater than modelled natural variability. However, for precipitation, most of the models are not above or below the oval. Therefore, the estimated changes in precipitation for most models are less than modelled natural variability – so future precipitation in this region is not expected to differ noticeably from present precipitation. Presentation title * To install: Create a folder called “sg41” in the c:\ Unzip the Scengen.zip file to the c:\sg41 directory To launch MAGICC, open: C:\sg41\SCEN-41\magicc\magicc.exe MAGICC/SCENGEN can be obtained at: http://www.cgd.ucar.edu/cas/wigley/magicc/ Presentation title * First-time users should click on ‘ReadMe’ for some important information. This also directs the first-time user to the MAGICC/SCENGEN License Agreement. Updates and revisions of MAGICC/SCENGEN will be sent only to those who have completed and returned this Agreement. A copy of the Agreement is given at the end of User Manual. To save or print what is on the screen, click Alt-Prnt Scrn. Go to a blank Word document and click Ctrl-V. The picture can be saved in a Word document or copied to PowerPoint. Presentation title * To begin, click on ‘Edit’. This displays a pull-down menu with the choices Emissions Scenarios, Model Parameters and Output Years. Under Emissions Scenarios, the user can select a Reference and Policy scenario. Under Model Parameters, most of the selections are self-explanatory. Examples will be given below. New features are climate feedbacks on the carbon cycle, and (accessed by clicking on the default User model) the option to emulate a range of AOGCMs, specifically those used in Chapter 9 of the IPCC Third Assessment Report (TAR). The range of options under Model Parameters allows the user to carry out a variety of sensitivity studies. Under Output Years, the user can control the years covered and time-step interval for output to the MAG.OUT file (in the MAGICC folder/directory), which gives full simulation details for the Reference scenario. Buttons on the right of this window can be used to return to the default settings. Presentation title * The emissions scenarios selection enables the user to select a reference (typically no policy) scenario and a policy scenario. The choices for emissions scenarios are as follows: * SRES emissions scenarios: 35 scenarios represent different storylines created using different integrated assessment (IA) models. The labels identify the storyline and (abbreviated) the IA model. The ones that show the SRES scenario and a dash “-” before the emissions model are the six illustrative scenarios used in the IPCC TAR * WRE CO2 stabilization scenarios. These are the “Wigley, Richels, Edmonds” stabilization scenarios (Wigley et al., 1996). For CO2 concentrations to stabilize with these scenarios, default MAGICC parameters must be used – these include climate feedbacks of the carbon cycle. * NFB stabilization scenarios: These are CO2 concentration stabilization scenarios that will only stabilize if carbon cycle feedbacks in MAGICC are turned off (NFB stands for No FeedBack). The CO2 emissions differences between corresponding NFB and WRE scenarios show how much the implied emissions are affected by climate feedbacks on the carbon cycle. This version of MAGICC has A1B-AIM as the default reference scenario and B2-MES as the default policy scenario. Users can select any emissions scenarios from the list on the left. Presentation title * Under Edit, select Model Parameters to display default settings for a selection of the most important controls on the output of MAGICC. Users may change these. Model sensitivity (DT2x), for example, is the amount of increase in global mean temperature caused by a doubling of carbon dioxide in the atmosphere over pre-industrial levels. The default values is 2.6oC. The 90% confidence interval is about 1.5 to 4.5°C, although some papers suggest much higher values. Under Edit, Output Years allows the user to select the beginning and end of the MAGICC runs. After these choices are made, select Run. Presentation title * To view results, press View in the MAGICC control window. Graphs enable the user to see graphs of emissions, concentrations, radiative forcings, and temperature and sea level. The figure displays CO2 concentrations for the two emissions scenarios. The solid line reflects the best estimate for concentrations given the scenario and model parameters selected. The fuzzy lines reflect uncertainty. If the user changed model parameters, click on Ref.user or Pol.user to display. Presentation title * This figure displays increases in global-mean temperature relative to 1990 for the two scenarios. The solid line reflects the choice of a 2.6oC climate sensitivity (the default selection). The fuzzy lines reflect sensitivities from 1.5 to 4.5oC. Presentation title * We now move on to SCENGEN. Go back to the main MAGICC control window, click on the SCENGEN button, and then on the Run SCENGEN button. This will bring up the SCENGEN title window. Click on OK. Presentation title * Clicking OK on SCENGEN brings up the main selection window and the blank map. Presentation title * In SCENGEN, select Analysis. This allows the user to select the analysis to be run. The choices are as follows: Data Change: Change in variable for the time period selected relative to 1990. This could be added to the observed climate data to produce the downscaled climate scenario described in slides 21-22. Error: The difference between model simulation of present-day climate and observed climate. Mod.Base: The model simulation of 1990 climate (base). Mod+Change: The model simulation of base climate plus the model simulation of change. This is different from the change because it is what the model simulates the future climate to be. Because there may be errors in the model’s simulation of present-day climate, using these results directly in analyses of vulnerability and adaptation is NOT recommended. Obs.Base. This is observed climate for the base period. SCENGEN uses the globally complete CMAP (Xie and Arkin, 1997) precipitation and CRU (New et al., 1999) temperature climatologies. Obs+Change. This is observed climate with the change from the climate model (or models) added on. This is similar to the downscaling approach described in slides 21-22, but is only at the 5°5° latitude/longitude resolution of SCENGEN. The approach described in slides 21-22 can yield the variance of results within a grid box. Variability S.D.Base: Standard deviation (interannual variability) of the present-day climate as simulated by the selected GCM (or GCMs). S.D.Change: Percentage change in standard deviation of the selected climate variable simulated by the GCMs. Tempor. SNR. Signal (i.e., the change in the selected variable) divided by the average across models of the present-day interannual (time series) standard deviation. Intermodel Inter-SNR: The change in climate averaged across the models selected divided by the intermodel standard deviation of the change. This shows whether the change simulated by the models is greater than the difference across model projections. If it is, the SNR value is greater than 1; if not, the SNR is less than 1. P(Increase): Calculates the probability of a precipitation increase assuming all the models selected are equally likely to happen (and effectively reflect the range of possibilities). This is an interesting result, but should be interpreted with caution. New et al., 1999. Xie and Arkin, 1997. Presentation title * In SCENGEN, select Models to get the screen shown. Certain models (a random selection) will be lit up as default. Select any set of models, from a single model to all models, and SCENGEN will produce results averaged over the selected models. For this example, all models are used, so click on All. We recommend, however, not using WM_95 because this is a very old model. The numbers after the model label give the year that the model used in the SCENGEN data base was for. The age of the model may be older than this. Earlier run years do not necessarily mean that the results are not as good – for example, the Hadley Centre (HadCM2) model (HAD295) is still one of the best models. Next the user has the option of using Definition 1 or Definition 2 changes. Def. 1 uses the difference between the start and end of a perturbation experiment. Def. 2 uses the difference between the perturbed state and the control climate at the same time. If a model has any spatial drift (and most models do) then Def. 2 is a way of removing this drift (under the justifiable assumption that the drift is common to both the perturbed and control runs) – normally one should use Def. 2. (Drift here refers to changes in model output that occur even if there is no external forcing. Most models show negligible drift in global-mean temperature, but do show drift when one looks at the spatial pattern of temperature.) Next, the user must decide whether or not to include the spatial effects of aerosols. Normally, these effects should be included (which is done by clicking on the ‘Aerosol effects’ button). The option not to include aerosol effects is to allow the user to determine how important these effects are. The Models window corresponds to these selections. Presentation title * Next, return to the SCENGEN window and click on Region. A map will be displayed showing the regions used for the breakdown of SO2 emissions in the MAGICC emissions files, together with a set of region selections. The default region is the whole globe. The user can select from a range of ‘hard-wired’ regions, or can define - by moving the mouse - a rectangular region on the map. To do this, click on User (this will normally be highlighted already as the default) and use the mouse to define a region. Note that a minimum of two grid boxes must be selected. The latitude/longitude domain will be shown numerically on the right. The selected region appears as a red rectangle. In this case, a region covering much of southern Africa has been selected. (Note that the hard-wired regions are generally not rectangular.) Selecting a region means that most calculations will be carried out specifically for that region. This includes area averages for the selected variable (see below), and a range of other statistics. Presentation title * Now return to the SCENGEN window and click on Variable. The Variable window, shown here, will appear. The default is annual-mean temperature. Click on Ann to see the other options (seasons or months), and then return to Ann. Next click on Precipitation. Note that the Reverse light will come on, because the standard color scheme for precipitation (red for dry to blue for wet) is the opposite of that usually used for temperature (blue for cold to red for hot). This window gives the user the option to use linear or power law (exponential) scaling. The latter is a way of avoiding physically unrealistic results that can (albeit only rarely) occur with linear scaling and large global-mean warmings. Presentation title * Return to SCENGEN and click on Warming. This is where the user selects the following: (1) The emissions scenario, either the Reference or the Policy case. The names displayed show only the first letters of the headers on the emissions files. B2MES has been selected. (2) The scenario year (i.e. the central year for a climate averaging interval of 30 years, as indicated by the length of the slider bar) – a central year of 2050 has been chosen here for illustrative purposes. (3) A particular configuration for the MAGICC model, default (i.e. best guess) or user. User is used if emissions parameters in MAGICC were changed. These factors determine the global-mean temperature (red 1.33oC at top of window) that is used for scaling the normalized patterns of change. Within the code, this global-mean temperature change is broken down into four components (a GHG component, and aerosol components for the SO2 emissions in the three emissions regions shown above) and these are used as weights for the pattern scaling algorithm. At this stage, all necessary user selections for SCENGEN have been made. Presentation title * Return now to the SCENGEN window and click on Run to run the SCENGEN software. After a short time, a map will appear. This shows the percentage change in annual-mean precipitation for the 30-year interval centered on 2050 (for the B2MES emissions scenario, and best guess climate model parameters in MAGICC) averaged over all 16 AOGCMs in the SCENGEN model data base (listed to the right of the map). The changes correspond to a global-mean warming of 1.33°C, and the patterns include the effects of aerosols according to the aerosol selection made in MAGICC (i.e. the best-guess case, which corresponds to -1.3 W/m2 aerosol forcing in 1990, the central estimate used in the IPCC TAR). Most of the displayed grid boxes are brown, which means that the model average result is a reduction in precipitation of up to 3% per year. In the northwest (yellow) grid box (10 to 15°S and 20 to 25°E), the model average is an increase in precipitation in the range 0 to 3%. Note that while still in SCENGEN, the user can place the cursor on each grid box. The screen will display the latitude and longitude of the selected grid box and the change in the value of the variable (precipitation or temperature). Presentation title * To examine quantitative results, go to C:\SG41\SCEN-41\engine\imout. Open the file AREAAVES.OUT in Wordpad or ASCII. The file will display the number of models run, the simulated year, the increase in global-mean temperature, and the area selected by the user. The first set of numbers following the model names (see below) is the normalized change for each model for the area selected. It is for GHGs alone. For temperature, it would be the change in temperature for the region per degree Celsius increase in global-mean temperature. For precipitation, it is the percentage change for the region per degree increase in global-mean temperature. The final set of numbers in the file is the pattern scaled results for the models and areas selected (displayed in the slide). These results include GHGs and aerosols. The first line, INTER-MOD S.D, is the standard deviation across the 16 GCMs for the area selected. The next line, INTER-MOD SNR, is the signal to noise ratio. It is negative because on average the models estimate a reduction in precipitation for the region. Use the absolute value. If the magnitude of the SNR is greater than 1, the model average is greater than the difference across models, indicating there is more agreement among models. If it is less than 1, this generally indicates the models are not in agreement. In this case it is well less than 1. PROB OF INCREASE is the probability of an increase, assuming the model results are equally likely and represent the range of possible outcomes. This result should be used with caution. GHG ONLY is the the model-average scaled change in precipitation considering only increased GHGs. AEROSOL ONLY is the effect of aerosols for the model average. GHG AND AEROSOL is the results above combined. Then, results by model are presented. D2 in the model name means Definition 2 has been used (i.e. possible model drift has been accounted for) – see slide 54. The last line, MODBAR, is the model average. Model average or individual model results can be used. These are the models used in SCENGEN 4.1, from the CMIP2 data base: 1. (BMRC) - BMRCTR98 (Bureau of Meteorology Research Centre – Australia) 2. (CCC) - CCC1TR99 (Canadian Climate Centre) 3. (CCSR) - CCSRTR96 (Japan) 4. (CERF) - CERFTR98 (CERFACS – France) 5. (CSIR) - CSI2TR96 (CSIRO – Australia) 6. (ECH3) - ECH3TR95 (ECHAM3 – Max Planck Institute for Meteorology, Germany) 7. (ECH4) - ECH4TR98 (ECHAM4 – Max Planck Institute for Meteorology, Germany) 8. (GFDL) - GFDLTR90 (Geophysical Fluid Dynamics Laboratory – NOAA, USA) 9. (GISS) - GISSTR95 (Goddard Institute for Space Studies – NASA, USA) 10. (IAP) - IAP_TR97 (Institute for Atmospheric Phisics – China) 11. (LMD) - LMD_TR98 (Laboratoire Meteorologique Dynamique – France) 12. (MRI) - MRI_TR96 (Meteorological Research Institute – Japan) 13. (CSM) - CSM_TR98 (Climate System Model – NCAR, USA) 14. (WM) - W&M_TR95 (Washington and Meehl – NCAR, USA) 15. (PCM) - PCM_TR00 (Parallel Climate Model – NCAR, USA) 16. (UKMO) - HAD2TR95 (HadCM2 – UK Meteorological Office) (UKMO3) - HAD3TR00 (HadCM3 – UK Meteorological Office) Further information on the models is provided in the IPCC TAR Work Group I, Chapter 8; http://www.grida.no/climate/ipcc_tar/wg1/316.htm Presentation title * Go back to SCENGEN. Note that all the settings stay the same except for the map. It always reverts to the default, which is the entire globe. Run SCENGEN again and the map of the globe above appears. This is the model average projected change in precipitation for the B2 scenario in 2050 using Definition 2 and including aerosols. In SCENGEN the user can place the cursor on any grid box to see what the models project for precipitation change for that grid box. Regional patterns can be seen. Increased precipitation is projected by the models on average in higher latitudes and substantial increases are projected in south Asia. Decreases are projected on average by the models in the Mediterranean, Mexico and Central America, and southern Africa. An analysis of signal to noise ratio should also be conducted to ascertain the degree of agreement among models. Presentation title * SCENGEN can be used to assess whether some GCMs better simulate current climate than others. In SCENGEN, select Analysis, then select Error. Here a slightly larger region, 5 to 30°S and 25 to 35°E, was selected. In Warming, select the year 2000 (to get the simulation of current climate). The map displays the average model simulation of current precipitation minus observations expressed as a percentage of the observed value. Percentage precipitation errors are often large. This is partly because simulating observed precipitation is one of the most difficult model tasks. In some areas, where observed precipitation is low, percentage errors may amount to only small absolute errors. In other areas, where natural variability of precipitation is large, a large percentage error may not be statistically significant. In any event, large model errors in simulating present-day climate do not necessarily mean that the projected changes are flawed. In the illustrated case, the errors are greatest in the southern grid boxes and lowest in the northwestern grid boxes. Presentation title * To examine quantitative error analysis results, go to C:\SG41\SCEN-41\engine\imout\ and open the file VALIDN.OUT in Wordpad, Notepad, or any text editor program. The results shown here as Unweighted Statistics are actually cosine weighted – i.e. grid box results are weighted by area, which is proportional to the cosine of latitude. (There is a labelling error in the code that has not yet been corrected.) Results labelled Cosine Weighted are, in fact, unweighted. In most cases, cosine weighted statistics are preferred, but differences between weighted and unweighted statistics are usually small. The first column (MODEL) gives the model name. The second column (CORREL) gives the pattern correlation, i.e. a measure of how well each GCM simulates the pattern of current precipitation over the selected region. The third column (RMSE) displays the root mean square error, a measure of error which gives more weight to larger errors. The fourth column (MEAN DIFF) compares the model simulation of average precipitation across the selected grid boxes with the observed average. The last column (NUM PTS) gives the number of grid boxes in the selected analysis region. (Note that the columns in the display here are slightly offset: mm/day headers should line up with columns three and four.) Any or all of these statistics can be used to select which models best simulate current climate. It is strongly recommended that users also do this analysis using results for the entire globe. Models should be assessed based on how well they simulate global climate as well as climate for the region in question. In this case HAD3 (HadCM3) best matches the pattern of precipitation. HAD2 (HadCM2) has the smallest root mean square error and CCC (Canadian Climate Centre model) has the smallest mean difference. This analysis could be used to exclude models with particularly poor scores for a number of statistics. Where the line is drawn is arbitrary. In these results, ECH3, LMD, and MRI have relatively low pattern correlations and/or high RMSEs and Mean Differences. The Change analyses can be rerun in SCENGEN using only the models that best simulate current climate. Presentation title * Presentation title * Presentation title * Presentation title * Presentation title * Presentation title * Presentation title * Presentation title * Presentation title * These slides suggest how these tools can be used to select scenarios. Presentation title * Averaging across multiple models often gives more reliable results. Also, intermodel differences give some idea of uncertainties. Presentation title * By extreme events, we mean low-frequency, high-consequence events such as very intense precipitation or heat events. Presentation title * It can be very useful for analysts to routinely remind themselves of the purposes for creating climate change scenarios. Presentation title * Using scenarios to communicate what changes in climate are possible can serve an important educational purpose and help decision makers address potential future changes in climate.
Comments
Report "CGE Training materials - VULNERABILITY AND ADAPTATION Assessment CHAPTER 8"