Introduction: Advanced study course on climate science Hans von Storch Bologna+Mestre, 15-26 October 2018

Hans VON STORCH Climate researcher (in the field since 1971) Coastal climate (storms, storm surges, waves; North and Baltic Sea, North Atlantic, Yellow Sea); statistical analysis Director emeritus of the Institute of Coastal Research of the Helmholtz Zentrum Geesthacht, Germany Professor at Universität Hamburg and at the Ocean University of China Editor-in-chief of the Oxford University Press Research Encyclopedia on Climate Science Lead author of IPCC AR3 and AR5. Co-Chair of regional assessment reports Baltic Sea Catchment (BACC) and Hamburg http://www.hvonstorch.de/klima/

The selection of topics reflects the interests and experience of the lecturer Hans von Storch. The lectures in this series stand mostly by themselves, thus some overlap and repetition is unavoidable. The lectures work at a relatively high level of abstraction; if formula are used, they represent shorthands for linkages. They are not used to demonstrate how to come from assertion A to conclusion B. If attendees want to support this project by building illustrative examples, please let me know.

Context: Summarizing key specifics of climate knowledge making (aka science) - a lecture series Climate comprises various components, in particular the hydro- and thermodynamics as well as the chemical composition of the atmosphere and of the ocean, but also the cryosphere, the biosphere in ocean and on land, and the socioeconomic system. The linkage of society and climate creates conditions favoring politically charged scientific milieus. The system features a practically infinite number of degrees of freedom. Many of them exhibit nonlinear dynamics and linkages. Thus, “noise” is an ubiquitous, constitutive element in the dynamics of climate. In most cases, the dynamics may be ordered according to spatial and temporal scales, with larger scales dominating the smaller scales, while smaller scales influence the larger scales significantly, but in a statistical sense (parameterizations). This ordering allows robust modelling and downscaling strategies.

Chapters Exploiting the ordered dynamics: downscaling Modelling: building knowledge Statistical analysis – explanatory and confirmatory Noise – nuisance and constitutive Pattern decomposition: EOFs and CCA Detection of non-natural change and Attribution to plausible causes The political dimension of climate science: Merton vs. Postnormal Regional Climate Service

Chapters Exploiting the ordered dynamics: downscaling Modelling: building knowledge Statistical analysis – explanatory and confirmatory Noise – nuisance and constitutive Pattern decomposition: EOFs and CCA Detection of non-natural change and Attribution to plausible causes The political dimension of climate science: Merton vs. Postnormal Regional Climate Service

Chapters Exploiting the ordered dynamics: downscaling 24. October, Bologna Modelling: building knowledge Statistical analysis – explanatory and confirmatory Noise – nuisance and constitutive 26. October, Bologna Pattern decomposition: EOFs and CCA Detection of non-natural change and Attribution to plausible causes The political dimension of climate science 16 October, Mestre Regional Climate Service 22. October 2018, Bologna

Chapter 1: Exploiting the ordered dynamics: downscaling 24. October, Bologna Atmospheric and oceanic dynamics are global dynamics, subject to global forcing (the differential heating by the sun). A simulated)Earth without structures, such as continents and mountain ranges, shows the key circulation aspects, such as the meridionally organized cells, and the bands of westerly winds. By adding continents and big mountain ranges, additional detail is formed such as the land-sea contrast, the western boundary currents etc. By adding further smaller scale details regional features emerge. This is the downscaling paradigm., which is conceptually expressed by R = F(L), with some transfer function/model F. (R regional, L = large-scale) On the other hand, smaller scale dynamics, such as convection are essential for the formation of the large-scale dynamics, but not in terms of small scale detail, but in terms of an average across such details, again conditioned by the large scale state itself – this is named parameterization. The split of scales is employed when building quasi-realistic climate models, and in “downscaling” systems, for estimating regional and local change and impact, conditioned by the simulated large scale.

Chapter 2: Modelling: building knowledge The term “model” is used differently in different scientific quarters (say physics, geology, climate science). Models are not “models of” but “models for” – that is, their design depends on what added knowledge is supposed to be constructed with them. Models of minimum complexity constitute “understanding”, as they relate a phenomenon to the dominant drivers. In most cases they are represented by a simple formula. They go under the name of “conceptual models”. Models of maximum complexity are engineering tools for simulation, without immediate generation of understanding. They allow for numerical experimentation, and validation (or falsification) of conceptual models. The are named “quasi-realistic models”. Their level of complexity is limited by the computing resources. Important applications of quasi-realistic models are forecasting, scenario construction, generating statistics and data analysis.

Chapter 3: Statistical analysis – explanatory and confirmatory Climate is the statistics of weather (mainly in the atmosphere and the ocean). The state of the atmosphere and the ocean is permanently changing; part of this variability is related to external drivers, and part of it is due to the interaction of an almost infinite number of degrees of freedom in the climate system. Thus the topics “noise” and “discriminating between internal and external causes “ are significant. Statistical analysis is envelope of mathematical methods, based on the concept of a “randomness”. It suggests methods how to estimate properties of these random variables from a limited number of observations (“explanatory analysis“) and how to determine if limited observational evidence is consistent with an a-priori formulated hypothesis (“confirmatory analysis”)

Chapter 4: Noise – nuisance and constitutive 26. October, Bologna Climate maybe understood as the statistical conceptualization of the state and variability of the climate system. Noise = internally generated variability, which can not be traced to external drivers (“Smoke without fire”). „Noise“ is ubiquitous in the climate system - on all scales - due to myriads of nonlinear mechanisms It is not relevant, if this variability is really stochastic or if it is just impossible to discriminate the variability from stochastic variations. “Noise” is a - most useful - mathematical construct. The presence of noise changes the dynamics of the system – the mean circulation would be different if there were no storms; the storms were different if there were not convective cells …; Klaus Hasselmann’s “Stochastic climate model” Noise hinders the identification of externally forced signals (and the utility of forecasts), and the attribution of anomalies to specific causes. 11

Chapter 5: Pattern decomposition: EOFs and CCA The downscaling concept suggests that the field-variability of a state variable by be ordered according to patterns of different sizes (“scales”). “Empirical Orthogonal Functions” were introduced by Ed Lorenz into atmospheric sciences. It is now a ubiquitously used tool. EOFs are patterns, so that the first pattern represents a maximum of variance. The 2nd pattern is orthogonal to the first, and represents a maximum of that variance, which is not accounted for be the first EOF. In the same way, 3rd, 4th patterns are constructed. Any field may be decomposed into contributions of these patterns. If a field is decomposed, then it is characterized by its Principal Components or EOF coefficients, which are univariate coefficients of the patterns. Similarly, Canonical Correlation Analysis decomposes the linkage between two fields in a sequential manner, with patterns and coefficients. The first patterns share maximum correlation, the 2nd patterns have no correlations with the first, but maximum correlation among each other. Related techniques are Maximum Covariance Analysis and Redundancy Analysis. Sometimes the term “singular spectrum analysis” is used, which is a misnomer as this term refers to a method of solving an eigenvalue problem.

Chapter 6: Detection of non-natural change and Attribution to plausible causes For the societal debate about climate change, at least in the west, there is one key questions, which needs scientific answers: Is there a change ? What are the dominant causes for such a change, and what are the expectations for the future? For answering this question, three issues have to be dealt with: Manifestation: The found change is real and not an artifact of the data and data collection process (inhomogeneity) Detection: The found change is beyond what may be expected due to natural (not externally caused) variations. Detection is a statistically rigorous confirmatory process. Significance tests of trends are mostly useless. Attribution: A change, which was found to be beyond the range of natural variations, may plausibly and consistently be explained by a certain (mix of) external cause(s). Here, a plausibility argument is built. The “detection and attribution”-framework was suggested by Klaus Hasselmann (1979, 1993).

Chapter 7: The political dimension of climate science: Merton vs Postnormal 16 October, Mestre Climate science takes placed in a politicized milieu.; politics are scientized, science is politicized. This contrasts to the “CUDOs” norms of R. Merton (1942): communal, unpersonal, disinterested and organized sceptic. These norm were never fully adopted, but widely in the scientific community claimed as valid. Climate change appears in the public different ways, either as “scientific construct” or as a variety of “social constructions”. They are not independent: social constructions feed back on scientific constructions. and social constructions are informed by scientific constructions. For policymaking the social constructions matter. Climate science is in a post-normal phase (Funtowicz and Ravetz) = knowledge is uncertain, decisions are urgent, values are in dispute and stakes are high. In this phase, scientific results are considered useful, when they support certain political goals, while methodological rigor is less significant. Acceptance of the failing Merton-norms and the unavoidable post-normal conditioning may help science to preserve some of its authority of “objectively explaining complex phenomena to the public”.

Chapter 8: Regional Climate Service 22 October, Bologna Climate knowledge in the public / among stakeholders is socially constructed. Dialogue between public /stakeholders and science needs understanding of communications needs. Regional stakeholders deal mostly with regionally specific issues; traditions play a role. Analysis of cultural construct, including common exaggeration in the media. Determination of response options on the local and regional scale: mainly adaptation but also regional and local mitigation. Dialogue of stakeholders and climate knowledge brokers in „Klimabureaus“. 2. Analysis of consensus on relevant issues (climate consensus reports). 3. Description of recent and present changes. 4 Projection of possible future changes, which are dynamically consistent and possible („scenarios“)