Refine
Year of publication
Document Type
- Article (13)
- Working Paper (6)
- Report (4)
- Part of a Book (3)
- Other (2)
- Conference Proceeding (1)
- Part of Periodical (1)
Has Fulltext
- yes (30)
Is part of the Bibliography
- no (30)
Keywords
- Klimaänderung (4)
- Deutschland (3)
- Geschichte 1900-2000 (2)
- Hessen (2)
- Klima (2)
- Klimawandel (2)
- (anthropogene) Klimabeeinflussung (1)
- Climate change (1)
- Europa (1)
- Extremereignis (1)
Institute
- Geowissenschaften (30)
Die im Industriezeitalter und im globalen Mittel beobachtete Erwärmung der unteren Atmosphäre zeigt ausgeprägte regional-jahreszeitliche Besonderheiten (IPCC 2001, SCHÖNWIESE 2003, 2004). Dies gilt in noch höherem Maß für den Niederschlag (vgl. Kap. 3.1.2 und 3.1.8). Die Vermutung, dass eine solche Erwärmung zu einer Intensivierung des hydrologischen Zyklus führt, was im Prinzip zunächst richtig ist (vgl. Kap. 3.1.2), erweist sich jedoch als viel zu simpel, wenn daraus einfach auf eine generelle Niederschlagszunahme geschlossen wird. Dies gilt sogar innerhalb einer so kleinen Region wie Deutschland. Denn obwohl Deutschland im Mittel überproportional an der »globalen« Erwärmung teilnimmt (SCHÖNWIESE 2003, 2004), zeigen die Langzeitänderungen des Niederschlages im Detail ganz unterschiedliche Charakteristika. Dabei kann die hier vorgestellte Beschreibung der in Deutschland beobachteten Niederschlagtrends subregional noch wesentlich verfeinert werden, vgl. z.B. Analyse für Sachsen (FRANKE et al. 2004), da der Niederschlag eine nur geringe räumliche Repräsentanz aufweist (SCHÖNWIESE & RAPP 1997). Zeitliche Änderungen von Klimaelementen lassen sich nun in ganz unterschiedlicher Weise betrachten. Am meisten verbreitet sind lineare Trendberechnungen, wie sie auch einem Teil der hier vorliegenden Studie zugrunde liegen. Es können aber auch Trends anderer statistischer Kenngrößen als des Mittelwertes von Interesse sein, z.B. der Varianz. Häufigkeitsverteilungen, die in normierter Form Wahrscheinlichkeitsdichtefunktionen heißen, erlauben die Bestimmung solcher Kenngrößen in Form der Verteilungsparameter. Wird unter Nutzung geeigneter Verteilungen (z.B. Normal- oder Gumbelverteilung, vgl. unten Abb. 3.1.6-4) eine statistische Modellierung der jeweils betrachteten klimatologischen Zeitreihe vorgenommen, werden Aussagen über die Unter- bzw. Überschreitungswahrscheinlichkeiten bestimmter Schwellenwerte möglich, in verallgemeinerter Form für beliebige Schwellen und Zeiten (TRÖMEL 2004). Da dieser extremwertorientierte Aspekt von großer Wichtigkeit ist, soll auch ihm hier nachgegangen werden (vgl. alternativ Kap. 3.1.7 und 3.1.10). Die im Folgenden verwendeten Daten sind jeweils Monatssummen des Niederschlages 1901–2000 an 132 Stationen in Deutschland (teilweise unter Einbezug einiger Stationen in den angrenzenden Ländern), einschließlich der daraus abgeleiteten Flächenmittelwerte (sog. Rasterdaten; Quelle: Deutscher Wetterdienst, siehe u.a. MÜLLER-WESTERMEIER 2002; vgl. weiterhin RAPP & SCHÖNWIESE 1996, dort auch Hinweise zur Homogenitätsprüfung, sowie RAPP 2000).
The climate system can be regarded as a dynamic nonlinear system. Thus, traditional linear statistical methods fail to model the nonlinearities of such a system. These nonlinearities render it necessary to find alternative statistical techniques. Since artificial neural network models (NNM) represent such a nonlinear statistical method their use in analyzing the climate system has been studied for a couple of years now. Most authors use the standard Backpropagation Network (BPN) for their investigations, although this specific model architecture carries a certain risk of over-/underfitting. Here we use the so called Cauchy Machine (CM) with an implemented Fast Simulated Annealing schedule (FSA) (Szu, 1986) for the purpose of attributing and detecting anthropogenic climate change instead. Under certain conditions the CM-FSA guarantees to find the global minimum of a yet undefined cost function (Geman and Geman, 1986). In addition to potential anthropogenic influences on climate (greenhouse gases (GHG), sulphur dioxide (SO2)) natural influences on near surface air temperature (variations of solar activity, explosive volcanism and the El Nino = Southern Oscillation phenomenon) serve as model inputs. The simulations are carried out on different spatial scales: global and area weighted averages. In addition, a multiple linear regression analysis serves as a linear reference. It is shown that the adaptive nonlinear CM-FSA algorithm captures the dynamics of the climate system to a great extent. However, free parameters of this specific network architecture have to be optimized subjectively. The quality of the simulations obtained by the CM-FSA algorithm exceeds the results of a multiple linear regression model; the simulation quality on the global scale amounts up to 81% explained variance. Furthermore the combined anthropogenic effect corresponds to the observed increase in temperature Jones et al. (1994), updated by Jones (1999a), for the examined period 1856–1998 on all investigated scales. In accordance to recent findings of physical climate models, the CM-FSA succeeds with the detection of anthropogenic induced climate change on a high significance level. Thus, the CMFSA algorithm can be regarded as a suitable nonlinear statistical tool for modeling and diagnosing the climate system.
Observed global and European spatiotemporal related fields of surface air temperature, mean-sea-level pressure and precipitation are analyzed statistically with respect to their response to external forcing factors such as anthropogenic greenhouse gases, anthropogenic sulfate aerosol, solar variations and explosive volcanism, and known internal climate mechanisms such as the El Niño-Southern Oscillation (ENSO) and the North Atlantic Oscillation (NAO). As a first step, a principal component analysis (PCA) is applied to the observed spatiotemporal related fields to obtain spatial patterns with linear independent temporal structure. In a second step, the time series of each of the spatial patterns is subject to a stepwise regression analysis in order to separate it into signals of the external forcing factors and internal climate mechanisms as listed above as well as the residuals. Finally a back-transformation leads to the spatiotemporally related patterns of all these signals being intercompared. Two kinds of significance tests are applied to the anthropogenic signals. First, it is tested whether the anthropogenic signal is significant compared with the complete residual variance including natural variability. This test answers the question whether a significant anthropogenic climate change is visible in the observed data. As a second test the anthropogenic signal is tested with respect to the climate noise component only. This test answers the question whether the anthropogenic signal is significant among others in the observed data. Using both tests, regions can be specified where the anthropogenic influence is visible (second test) and regions where the anthropogenic influence has already significantly changed climate (first test).
Im Rahmen der Diskussion des globalen bzw. regionalen Klimawandels sind, neben Extremereignissen, Langfristtrends von besonderem Interesse. Doch erfordert die ausgeprägte Klimavariabilität in Zeit und Raum spezielle regionale Detailuntersuchungen. Daher wird hier eine solche Analyse für Deutschland und die Klimaelemente bodennahe Lufttemperatur sowie Niederschlag vorgestellt, mit besonderem Blick auf die jahreszeitlichen/monatlichen Besonderheiten der Trends in ausgewählten Zeitintervallen zwischen 1891 und 2000. Am auffälligsten ist dabei die sich verstärkende winterliche Temperatur- und Niederschlagszunahme, während im Sommer, unter ebenfalls Erwärmung, eine Trendwende von abnehmendem zu in den letzten Dekaden zunehmendem Niederschlag eingetreten ist.
Temporal changes in the occurrence of extreme events in time series of observed precipitation are investigated. The analysis is based on a European gridded data set and a German station-based data set of recent monthly totals (1896=1899–1995=1998). Two approaches are used. First, values above certain defined thresholds are counted for the first and second halves of the observation period. In the second step time series components, such as trends, are removed to obtain a deeper insight into the causes of the observed changes. As an example, this technique is applied to the time series of the German station Eppenrod. It arises that most of the events concern extreme wet months whose frequency has significantly increased in winter. Whereas on the European scale the other seasons also show this increase, especially in autumn, in Germany an insignificant decrease in the summer and autumn seasons is found. Moreover it is demonstrated that the increase of extreme wet months is reflected in a systematic increase in the variance and the Weibull probability density function parameters, respectively.
Attribution and detection of anthropogenic climate change using a backpropagation neural network
(2002)
The climate system can be regarded as a dynamic nonlinear system. Thus traditional linear statistical methods are not suited to describe the nonlinearities of this system which renders it necessary to find alternative statistical techniques to model those nonlinear properties. In addition to an earlier paper on this subject (WALTER et al., 1998), the problem of attribution and detection of the observed climate change is addressed here using a nonlinear Backpropagation Neural Network (BPN). In addition to potential anthropogenic influences on climate (CO2-equivalent concentrations, called greenhouse gases, GHG and SO2 emissions) natural influences on surface air temperature (variations of solar activity, volcanism and the El Niño/Southern Oscillation phenomenon) are integrated into the simulations as well. It is shown that the adaptive BPN algorithm captures the dynamics of the climate system, i.e. global and area weighted mean temperature anomalies, to a great extent. However, free parameters of this network architecture have to be optimized in a time consuming trial-and-error process. The simulation quality obtained by the BPN exceeds the results of those from a linear model by far; the simulation quality on the global scale amounts to 84% explained variance. Additionally the results of the nonlinear algorithm are plausible in a physical sense, i.e. amplitude and time structure. Nevertheless they cover a broad range, e.g. the GHG-signal on the global scale ranges from 0.37 K to 1.65 K warming for the time period 1856-1998. However the simulated amplitudes are situated within the discussed range (HOUGHTON et al., 2001). Additionally the combined anthropogenic effect corresponds to the observed increase in temperature for the examined time period. In addition to that, the BPN succeeds with the detection of anthropogenic induced climate change on a high significance level. Therefore the concept of neural networks can be regarded as a suitable nonlinear statistical tool for modeling and diagnosing the climate system.
Vielleicht hätte sich außerhalb der Fachwissenschaft niemand für das Weltklimaproblem interessiert, wären da nicht zwei brisante, miteinander gekoppelte Fakten: Die Menschheit ist hochgradig von der Gunst des Klimas abhängig. Es kann uns daher nicht gleichgültig sein, was mit unserem Klima geschieht. Und: Die Menschheit ist mehr und mehr dazu übergegangen, das Klima auch selbst zu beeinflussen. Daraus erwächst uns allen eine besondere Verantwortung. ...
Die Zunahme der Konzentration von CO2 und anderen "Treibhausgasen" in der Atmosphäre ist unzweifelhaft, und ebenso unzweifelhaft reagiert das Klima darauf. Christian-Dietrich Schönwiese, Professor für Meteorologische Umweltforschung und Klimatologie an der Universität Frankfurt am Main, sieht dringenden politischen Handlungsbedarf und plädiert gleichzeitig dafür, die Debatte rund um den Klimaschutz zu versachlichen.
This is a short movie about the research project No. 297 41 132 of the German Federal Environmental Agency. It is comissioned by the German Ministry for the Environment, Nature Conservation and Nuclear Safety and the German Federal Environmental Agency, Division II 6.2.
Wenn sich beim Klimagipfel in Den Haag [genauer bei der nun schon 6. Vertragstaatenkonferenz zur Klimaschutzkonvention der Vereinten Nationen] nun wieder die Delegationen aus fast allen Staaten der Welt treffen, um über Klimaschutzmaßnahmen zu beraten, dann schwingt auch immer die Frage mit: Sind solche Maßnahmen wirklich notwendig? Sollen wir nicht einfach warten, bis wir mehr, ja vielleicht alles wissen? ...
The assumption that mankind is able to have an in uence on global or regional climate, respectively, due to the emission of greenhouse gases, is often discussed. This assumption is both very important and very obscure. In consequence, it is necessary to clarify definitively which meteorological elements (climate parameters) are in uencend by the anthropogenic climate impact, and to which extent in which regions of the world. In addition, to be able to interprete such an information properly, it is also necessary to know the magnitude of the different climate signals due to natural variability (for example due to volcanic or solar activity) and the magnitide of stochastic climate noise. The usual tool of climatologists, general circulation models (GCM) suffer from the problem that they are at least quantitatively uncertain with regard to the regional patterns of the behaviour of climate elements and from the lack of accurate information about long-term (decadal and centennial) forcing. In contrast to that, statistical methods as used in this study have the advantage to test hypotheses directly based on observational data. So, we focus to the very reality of climate variability as it has occurred in the past. We apply two strategies of time series analyis with regard to the observed climate variables under consideration. First, each time series is splitted into its variation components. This procedure is called 'structure-oriented time series separation'. The second strategy called 'cause-oriented time series separation' matches various time series representing various forcing mechanisms with those representing the climate behaviour (climate elements). In this way it can be assessed which part of observed climate variability can be explained by this (combined) forcing and which part remains unexplained.
The assumption that mankind is able to have an in uence on global or regional climate, respectively, due to the emission of greenhouse gases, is often discussed. This assumption is both very important and very obscure. In consequence, it is necessary to clarify definitively which meteorological elements (climate parameters) are in uencend by the anthropogenic climate impact, and to which extent in which regions of the world. In addition, to be able to interprete such an information properly, it is also necessary to know the magnitude of the different climate signals due to natural variability (for example due to volcanic or solar activity) and the magnitide of stochastic climate noise. The usual tool of climatologists, general circulation models (GCM) suffer from the problem that they are at least quantitatively uncertain with regard to the regional patterns of the behaviour of climate elements and from the lack of accurate information about long-term (decadal and centennial) forcing. In contrast to that, statistical methods as used in this study have the advantage to test hypotheses directly based on observational data. So, we focus to the very reality of climate variability as it has occurred in the past. We apply two strategies of time series analyis with regard to the observed climate variables under consideration. First, each time series is splitted into its variation components. This procedure is called 'structure-oriented time series separation'. The second strategy called 'cause-oriented time series separation' matches various time series representing various forcing mechanisms with those representing the climate behaviour (climate elements). In this way it can be assessed which part of observed climate variability can be explained by this (combined) forcing and which part remains unexplained.
Simulation of global temperature variations and signal detection studies using neural networks
(1998)
The concept of neural network models (NNM) is a statistical strategy which can be used if a superposition of any forcing mechanisms leads to any effects and if a sufficient related observational data base is available. In comparison to multiple regression analysis (MRA), the main advantages are that NNM is an appropriate tool also in the case of non-linear cause-effect relations and that interactions of the forcing mechanisms are allowed. In comparison to more sophisticated methods like general circulation models (GCM), the main advantage is that details of the physical background like feedbacks can be unknown. Neural networks learn from observations which reflect feedbacks implicitly. The disadvantage, of course, is that the physical background is neglected. In addition, the results prove to be sensitively dependent from the network architecture like the number of hidden neurons or the initialisation of learning parameters. We used a supervised backpropagation network (BPN) with three neuron layers, an unsupervised Kohonen network (KHN) and a combination of both called counterpropagation network (CPN). These concepts are tested in respect to their ability to simulate the observed global as well as hemispheric mean surface air temperature annual variations 1874 - 1993 if parameter time series of the following forcing mechanisms are incorporated : equivalent CO2 concentrations, tropospheric sulfate aerosol concentrations (both anthropogenic), volcanism, solar activity, and ENSO (all natural). It arises that in this way up to 83% of the observed temperature variance can be explained, significantly more than by MRA. The implication of the North Atlantic Oscillation does not improve these results. On a global average, the greenhouse gas (GHG) signal so far is assessed to be 0.9 - 1.3 K (warming), the sulfate signal 0.2 - 0.4 K (cooling), results which are in close similarity to the GCM findings published in the recent IPCC Report. The related signals of the natural forcing mechanisms considered cover amplitudes of 0.1 - 0.3 K. Our best NNM estimate of the GHG doubling signal amounts to 2.1K, equilibrium, or 1.7 K, transient, respectively.