Modelling in Climate Engineering Research

 

Embed or link this publication

Description

This brochure shall help to understand the functioning, limits and possibilities of modelling to help joining the discourse on CE. The brochure originates in the Priority Programme 1689, funded by the German Research Foundation.

Popular Pages


p. 1

A MODELLING IN CLIMATE ENGINEERING RESEARCH Wie beeinflussen Computer­ simulationen das Wissen­s chafts­ verständnis? Modelling in Climate Engineering Research Significance and Uncertainties Priority Programme 1689 of the German Research Foundation DFG SPP 1689

[close]

p. 2

This is a translation of the german brochure “Modellierung in der Climate Engineering Forschung – Aussagekräftig trotz Unsicherheiten”.

[close]

p. 3

Preface In the priority programme “Climate Engineering: Risks, Challenges, Opportunities?” (SPP 1689), we want to provide a comprehensive evaluation of the ideas that have emerged on Climate Engineering (CE) in science and climate politics in recent years. CE is the term used to describe deliberate large-scale interventions in the climate system, with the aim of mitigating the effects of climate change caused by humans. For a robust assessment of CE ideas, we take into consideration the social, political, legal and ethical aspects, in addition to the scientific and technical dimensions. The results are also discussed within the context of mitigation and adaptation strategies for climate change. Our research to assess (not develop!) CE deliberately takes a very broad interdisciplinary approach. Field experiments are explicitly excluded within the context of the SPP 1689. The research is, therefore, extensively based upon the results of computer simulations with numeric models of the climate system. In order to better understand the functioning, limits and opportunities of such models, the SPP PhD students from all involved disciplines organised a multi-day workshop on the topic of modelling under the direction of Miriam Ferrer González and Fabian Reith, together with the project coordinator Ulrike Bernitt. The questions, discussions and ideas which emerged during this workshop prompted us to summarise in writing the different aspects of modelling within the context of our priority programme for the evaluation of CE. In addition to the participants of the workshop, other members of the SPP 1689 from various disciplines, such as the fields of philosophy and economics, have contributed. With the resulting brochure, we hope to inform both scientists and interested members of the public about the basic aspects of our research, thereby making it easier to join the discourse on CE. 1 MODELLING IN CLIMATE ENGINEERING RESEARCH Significance and Uncertainties ANDREAS OSCHLIES SPP 1689

[close]

p. 4

CONTENT The Role of Modelling in Climate Engineering Research | Andreas Oschlies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 How do Computer Simulations Influence Scientific Understanding? | Martin Carrier and Johannes Lenhard . . . . . . . . . . . . . . . . . . . . . . . 6 Dealing with Uncertainties | Gregor Betz . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9 Uncertainties in Numeric Climate Simulations in a Decision Context | Hauke Schmidt and Hermann Held. . . . . . . . . . . . . . . . . . . . . . . . 12 The UVic Earth System Climate Model | D  avid Keller, Nadine Mengis, Fabian Reith and Andreas Oschlies. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15 The MPI Earth System Model | S  ebastian Sonntag, Tatiana Ilyina, Julia Pongratz and Hauke Schmidt . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18 The IPSL-CM Earth System Model | Olivier Boucher, Ulrich Platt and Christoph Kleinschmitt. . . . . 21 The Regional Model System COSMO-ART | Tobias Schad, Thomas Leisner and Bernhard Vogel . . . . . . . . . . 24 The LPJmL Vegetation Model | Tim Beringer, Lena Boysen and Vera Heck . . . . . . . . . . . . . . . . . . 27 Economic Modelling in the Context of Climate Engineering | Timo Goeschl and Martin Quaas . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 Priority Programme 1689. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 Authors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 34 Imprint | Contact. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35 2 MODELLING IN CLIMATE ENGINEERING RESEARCH Significance and Uncertainties SPP 1689

[close]

p. 5

The Role of Modelling in Climate Engineering Research Andreas Oschlies | Earth System Modelling The term “Climate Engineering” (CE) covers various large-scale technical measures, which could be used in a targeted manner either to lower the concentration of atmospheric CO2 or to directly influence the Earth’s radiation balance in order to counteract anthropogenic global warming. The CE methods discussed within the research community and in the public have, until now, only been ideas for technical methods that initially appear plausible and could in principle work. On the other hand, however, it is difficult to examine their actual effectiveness and assess unintentional side effects. As CE methods would be a targeted intervention in the climate system, which is a globally connected system of high complexity that is not yet sufficiently understood. Under laboratory conditions or during small-scale field experiments (e.g. iron fertilisation in the ocean or afforestation), the potential effectiveness and side effects of CE methods can only be tested in a very limited manner. It is also unclear to what extent the results of such small-scale and short-term experiments can be transferred to the global climate system. In order to assess the global effects and side effects of CE in an empirically reliable manner, corresponding large-scale and possibly global field experiments would be required. However, these would perhaps not differ significantly from an actual deployment of CE and could involve considerable risks since the results of scientific experiments cannot be predicted with certainty. The consequences of such experiments carried out in the natural environment could be irreversible and the observed effects, in view of our incomplete understanding of the climate system, might not even be unambiguously attributable to the experiment. Furthermore, there is a lack of governance regarding the permission, supervision and regulation of field experiments (and also for the deployment of CE methods). In consideration of the uncertainties, 3 MODELLING IN CLIMATE ENGINEERING RESEARCH The Role of Modelling in Climate Engineering Research SPP 1689

[close]

p. 6

the considerable risks involved and the large parts of the population potentially affected, it is not currently possible to carry out large-scale CE field experiments in a responsible manner. The opportunity to examine the effects and side effects of various CE methods without putting people or the environment at risk is provided by numeric models of the Earth system, which allow experiments to be carried out in a simulated world, rather than in the real natural environment. Earth system models simulate the interactions between various components of the climate system on the basis of scientific laws. These laws are presumably neither complete nor correct in every detail, which means that the simulated world is not a perfect copy of reality. Furthermore, the mathematical equations corresponding to the laws of science can often only be solved using numeric approximations (for example, the description of small-scale turbulence). The more exact the representation is, the higher the computational power required, which in practice generally limits the duration of a simulation. Coupled models containing model components describing the ocean, sea ice and the atmosphere, are generally described as climate models. Earth system models also contain modules to describe terrestrial vegetation, The description of a complex or real system through simplified mathematical formulations is called modeling. Generally these formulations are written and calculated using computer programs, with the results giving a simplified depiction of reality. A simulation describes the application of a particular model in a way that can be varied as desired. Simulated experiments allow us to gain an understanding of given questions, without having to actually conduct the experiments in nature. This enables us to make predictions, test hypotheses and illustrate causal relationships. As simplified depictions, models will always be incomplete and must, therefore, not be thought of as reality, something that has to be considered when interpreting the results. However, because of these simplifications models are also important tools for understanding single processes in complex systems. How complex a model has to be depends on what questions are being asked. 4 MODELLING IN CLIMATE ENGINEERING RESEARCH The Role of Modelling in Climate Engineering Research MODELLING SPP 1689

[close]

p. 7

soil, marine ecosystems and biogeochemical cycles. Different research groups and disciplines use different models. These models differ according to the research questions, their objectives and, depending on which components are to be represented, in their level of detail in the description of the different components. Within the context of the SPP 1689, simulations using several different Earth system models of varying degrees of complexity deliver possible scenarios, by which all working groups of the SPP 1689 can orient themselves. They thereby also form a basis for the investigation of the potential effects of CE by the humanities or social sciences e.g. with regard to the model’s epistemic value or the meaning of the modelling results in political discussions and decision-making processes. Even with this highly interdisciplinary approach, it must always remain clear that these models are only simplified representations of reality. They neglect to consider potentially important processes and include parameterizations1 of unresolved processes. They also depend on often only poorly known initial and boundary conditions. Because of this, each model simulation contains uncertainties that must be taken into consideration when interpreting the results. Despite these limitations on the validity of the models, in our view an assessment of the effectiveness and side effects of CE methods can currently only be carried out responsibly using computer simulations. 5 MODELLING IN CLIMATE ENGINEERING RESEARCH The Role of Modelling in Climate Engineering Research 1| Parameterizations are simplified descriptions of processes that are not fully described in the Earth system models (e.g. cloud development, turbulence and numerous biological processes). SPP 1689

[close]

p. 8

How do Computer Simulations Influence Scientific Understanding? Martin Carrier and Johannes Lenhard | Philosophy of Science A rapidly growing number of scientific investigations rely on computer simulations. These include investigations into Climate Engineering (CE), which gives rise to questions on the methodological characteristics of computer simulations and their effects on our understanding of science. 6 MODELLING IN CLIMATE ENGINEERING RESEARCH How do Computer Simulations Influence Scientific Understanding? What is special about computer simulations? They are based on theoretical, mathematically-formulated models, which are automatically processed by digital computers. This requires the conversion of the theoretical models into a form that can be dealt with by computers. For this purpose, differential equations, need to be solved in numerical form, that is, solutions have to be calculated point by point and for specific parameter values. Modelling of this type differs from conventional mathematical modelling with regard to methodology. The defined parameterizations and other adaptations, which are required due to the digitalisation of the models, are often not produced by the relevant theories but rather by independent modelling steps. Clouds, for example, can only be described in the simulation model by their effects at the grid points. One must, therefore, find a type of condensed description (parameterization), which works together with the remaining dynamics at the grid points in such a way that it adequately captures the main effects of the much more complicated and fine-scaled cloud formation processes. The model dynamics are often significantly influenced by the way in which the parameterization is chosen and how the corresponding parameters are set. However there is no guaranteed recipe for success for this type of modelling step. The particular behaviour of the model is not derived from the theory alone, but also depends on the auxiliary SPP 1689

[close]

p. 9

adjustments applied. Theory thereby loses part of its epistemic authority. The relationship between theoretical approach, pragmatic modifications and predictive value is not yet well understood with regard to its effects on understanding scientific modelling. Reflection on how computer simulations contribute to changing exploratory and explanatory procedures in science is still in its infancy. Important is that computer simulations also deliver results for complex, specialised circumstances, for which an analytical solution is inconceivable. Simulations are, therefore, suited to spelling out the consequences of theoretical principles that would otherwise be inaccessible. Of course, in view of the mentioned concerns, it needs to be ensured that these consequences actually originate from the theoretical principles themselves, and not from the pragmatic adaptations of these principles for making them suitable for being run on digital computers. Are the claims made by the theoretical model actually determined by its theoretical principles or by the implementation of these principles for numeric simulation processes? Under what conditions can we therefore expect that our simulation models adequately represent the future climate development? A fundamental way of assessing a model is checking it against experience. Such a test can be achieved by exploiting an advantage of simulation models related to the immense computational power of computers. Namely, one can use such simulation models experimentally. One changes certain parameters or procedures on a trial basis and examines the effects this has on verifiable consequences of the model. In this way, certain parameterizations and calculation procedures can be distinguished through experience. One is, therefore, not experimenting with nature but rather with the models. In such simulation experiments, models are tested and adapted, if so required, in order to be able to better judge their importance for specific questions. In this way, an experimental path is opened up to testing the validity of simulation models. 7 MODELLING IN CLIMATE ENGINEERING RESEARCH How do Computer Simulations Influence Scientific Understanding? SPP 1689

[close]

p. 10

8 MODELLING IN CLIMATE ENGINEERING RESEARCH However, computer simulations face a specific problem with regard to validation. In areas such as CE, while theoretically well-confirmed conceptual models are available, the aim is precise, long-term predictions. Because these predictions refer to conditions that have not yet been realised, empirical validation of CE simulation models is difficult. However, support for these models on theoretical grounds is also problematic, because of the mentioned dependence of the results on parameterizations and auxiliary calculation procedures. We simply do not know exactly on which components of the model the specific predictions primarily rely. Accordingly, computer simulations also raise special problems for the validity testing of models. Philosophers try to reconstruct the problems associated with such a validity test by tracing the conceptual structure of the models and analysing the relationships between their underlying theoretical assumptions and the empirical basis. Philosophy then strives to make the conceptual structure of the model and the corresponding validation relationships transparent. In addition to this reflection on the instruments of knowledge gain (instead of their construction and use, as is done by natural scientists), philosophers compare the relevant model characteristics to similar or dissimilar cases from other scientific disciplines or place these within the context of the historical development of science. This contextualisation may not only allow a deeper understanding of these specific characteristics, it may also provide a heuristic means of appropriately addressing problems in the validity testing of climate models. How do Computer Simulations Influence Scientific Understanding? SPP 1689

[close]

p. 11

Dealing with Uncertainties Gregor Betz | Philosophy of Science Quantitative modelling and computer simulations do not in general guarantee certainty or reliability. Model results can be more or less uncertain for various reasons: Relevant causal connections may not be identified by the model; the values of certain model parameters are poorly constrained; or the initial and boundary conditions are unknown. We possess both linguistic and mathematical means to express uncertainty in a differentiated manner. In many cases, quantitative probabilities can be reliably determined. In other situations, only relevant possibilities can be identified, e.g. by giving an interval, an order of magnitude or a development trend for a variable. It is disputed among both climate scientists and philosophers of science which type of knowledge is provided by climate models and how their (partly heterogeneous) results are to be interpreted. There are at least four suggestions for how to understand climate models and their outcomes: 1. One regards the models as competing hypotheses about the actual climate system and assumes, for practical purposes, the forecasts of the model which is empirically best confirmed. 2. One interprets the frequency distribution of the model results themselves as probability distribution in order to thus quantify the uncertainty associated with forecasts. 3. One interprets the model results as scenarios that cover the range of plausible possibilities. 4. One uses the models to identify previously unseen (not even articulated) consequences of actions (unknown unknowns). 9 MODELLING IN CLIMATE ENGINEERING RESEARCH Dealing with Uncertainties SPP 1689

[close]

p. 12

What are the practical ramifications of model uncertainties? How is one to take these uncertainties into consideration when reflecting on and making decisions? Whether a decision is right or wrong depends on the consequences associated with the various options of action. Forecasts of the consequences of actions are the central descriptive assumptions2 in derivations of policy recommendations. Evaluations of Climate Engineering (CE) options hinge, for example, on the intended and unintended consequences of the research on, and the deployment of, CE technologies. Risk preferences and the evaluation of possible consequences represent the normative assumptions2 that fuel such practical reasoning. It is therefore possible for two parties to agree on the forecast3 and the evaluation of the consequences of policy options but, due to different levels of risk aversion, to disagree on the measure to be taken. In risk ethics and decision theory, decision situations are classified according to the available foreknowledge, which can be more or less uncertain, namely as decisions  10 MODELLING IN CLIMATE ENGINEERING RESEARCH Dealing with Uncertainties under risk (probabilities can be assigned to possible consequences of actions)  under uncertainty (all possible consequences of actions are identified) under ignorance (some relevant consequences of options are unknown)  2| Descriptive assumptions describe what is (and/or was or will be) the case; normative assumptions say something about what should be the case and evaluate a situation without implying that it actually obtains. “Peter keeps his promises”, for example, is a descriptive statement, “Peter should keep his promises” however is a normative statement. 3| Probability forecast or possibility forecast. SPP 1689

[close]

p. 13

It is in principle possible, in all these situations, to argue rationally for or against an action. In decisions under risk, the principle of expected utility maximisation is often used. In case proper uncertainty prevails, the precautionary principle can be applied. Whether, in the context of CE, we face decision situations under risk or rather under uncertainty, crucially depends on how climate models and their results are interpreted. 11 MODELLING IN CLIMATE ENGINEERING RESEARCH Dealing with Uncertainties SPP 1689

[close]

p. 14

Uncertainties in Numeric Climate Simulations in a Decision Context Hauke Schmidt | Climate Science and Hermann Held | Physics and Climate Economics Numeric computer models of the Earth system are indispensable tools for estimating future climate development. The climate of the future is projected based on scenarios for future greenhouse gas emissions. How precise must such projections be? The social relevance of the accuracy of climate projections is expressed among other things in the economic benefit that could emerge (because of the increased ability to plan at that point), if the projections could be more precise. The expected benefit of maximum precision for scenarios without Climate Engineering (CE) has already been calculated: Globally, it lies somewhere between billions and hundreds of billions of Euro per year (for these results, it was simply assumed that any underdeterminedness of climate projections can be expressed through a probability distribution). Moreover, it makes sense to suggest that the strength of the reaction of the climate system to greenhouse gases plays a role in the assessment of whether the use of CE seems reasonable. The essential strategies for assessing the reliability of the climate model projections are a) evaluation of the models on the basis of observed data and b) comparison of different models. 12 MODELLING IN CLIMATE ENGINEERING RESEARCH Uncertainties in Numeric Climate Simulations in a Decision Context Model evaluation: In order to generate confidence in future projections of a model, a necessary condition is the realistic simulation of the observed climate development of the 20th century. This is not a guarantee for the correctness of the projections of the future though. Most models at large climate research centres, for example, reproduce the observed average global rise in temperature so far quite well. However, this does not necessarily mean that the underlying mechanisms are described SPP 1689

[close]

p. 15

correctly in the models. For example, the quantification of the effect of anthropogenic atmospheric aerosols is uncertain. Possible errors by a model with regard to the effect of greenhouse gases on the temperature (climate sensitivity) could, therefore, be compensated by various ways of taking aerosol effects into consideration. Because of this, it is necessary to perform the evaluation using different parameters and not only the globally averaged temperature. For simulations of CE methods, there is also the problem of the lack of experience and empirical evidence. For example, in the case of the CE suggestion to inject sulphur into the stratosphere, climate research has to make do with the analogy of large volcanic eruptions. Of course, assessment is still difficult here; since on the one hand there are only a few well-observed large volcanic eruptions (the last was the eruption of Mount Pinatubo in 1991) and on the other hand it is not clear how similar the reactions of the climate are to artificial and natural volcanic aerosol forcing. Since the mid-1990s, systematic model comparisons within the context of the “Coupled Model Intercomparison Project” (CMIP) have been carried out by the international climate modelling community, which also delivered essential input for the Assessment Reports of the Intergovernmental Panel on Climate Change (IPCC). To do so, identical climate scenarios are simulated using many different models (most recently by about twenty research institutes). If there is broad agreement on the simulated climate, one speaks of robust signals and assumes that these are primarily determined by the basic physics of the climate system and are only marginally dependent on specific model formulations. With regard to CE, climate researchers have taken CMIP as an example, and since about 2010 have been simulating the possible climate effects of proposed measures for radiation manipulation in the Geoengineering Model Intercomparison Project (GeoMIP). 13 MODELLING IN CLIMATE ENGINEERING RESEARCH Model comparison: Uncertainties in Numeric Climate Simulations in a Decision Context SPP 1689

[close]

Comments

no comments yet