Open access
Author
Date
2017Type
- Doctoral Thesis
ETH Bibliography
yes
Altmetrics
Abstract
Extreme weather and climate events (summarised as "climate extremes" from here onwards) are a crucial aspect of Earth's climatic variability. However, climate extremes are frequently associated with adverse impacts on socio-economic and ecological systems. For example, heat in combination with drought may severely affect the functioning of terrestrial ecosystems, and in some cases these events have the potential to undo several years of ecosystem carbon sequestration. Moreover, the intensity and frequency of several types of climate extremes, such as heat, cold, and heavy rainfall, have been changing in recent years. These changes are projected to continue in the 21st century, thus raising concerns about the capacity of ecological and socio-economic systems to cope with these events in the future.
Nonetheless, our scientific understanding of climate extremes and the mechanistic pathways through which these events propagate into ecological or socio-economic systems, remains limited. The impact of climate extremes varies widely depending on their type and spatio-temporal structure, and these impacts are mediated by the vulnerability and exposure of the system under scrutiny. Therefore, the quantification of these phenomena, and the attribution to their respective drivers across space and time is often ambiguous. Accordingly, closing scientific knowledge gaps and improving methodologies to scrutinise climate extremes and their impacts constitutes a research priority of high societal relevance.
The overarching objective of the present PhD thesis is to improve the quantification of, and contribute to the understanding of climate extremes and their impact on ecosystem-atmosphere interactions. To address these objectives, the thesis relies on joint analyses and integration of observation-based datasets and model ensemble simulations. Specifically, the thesis explores (1) a wide range of generic statistical-methodological considerations, (2) approaches to enable sound process-oriented model ensemble simulations using observation-based constraints, towards (3) a comprehensive attribution of ecosystem impacts arising from climate extremes.
1. Statistical quantification of extremes in observed or simulated spatio-temporal gridded datasets (Part I).
An investigation and quantification of extremes in spatio-temporal datasets requires robust statistical methodologies and diagnostics. Therefore, the thesis scrutinises statistical methods, both empirically and analytically, to explore recent changes in temperature and precipitation extremes in gridded observations. These analyses reveal that conventional statistical methods that are based on a reference period standardisation might induce substantial biases in spatially aggregated estimates of extremes. For example, the occurrence of extremes that exceed two standard deviations in standardised data could be overestimated by 48.2% outside a given reference period of 30 years in independent and identically distributed Gaussian data. Analytical corrections for these kinds of statistical errors are derived in the thesis.
Because climate extremes are inevitably rare in temporally and spatially limited observational records, ensemble simulations constitute an indispensable and complementary tool to scrutinise climate extremes from a statistical perspective, circumventing small sample issues in observations. Hence, the thesis also illustrates how model ensembles can be used as surrogate observations to benchmark statistical methods and metrics for an accurate assessment of climate extremes in observations.
2. Observation-based constraints improve model ensemble simulations of climate extremes and ecosystem impacts (Part II).
Climate model ensemble simulations generated for the purpose of quantifying and attributing climate extremes typically exhibit biases in their output that hinder any straightforward simulation or assessment of impacts. Therefore, I develop, apply, and evaluate tools to constrain climate model ensembles based on observational diagnostics related to land-atmosphere interactions. The application of these constraints simultaneously reduces multivariate biases in model ensembles and thus might offer a novel route to bias correction for climate impact simulations and analyses of climate extremes.
3. Extremes events in the terrestrial biosphere: drivers and attribution (Part III).
Linking or attributing extreme responses in the terrestrial biosphere to climatic drivers is not straightforward because respective analyses often rely on small sample sizes or even singular events in observations. Therefore, I construct an ensemble of climate-ecosystem impact simulations, constrained by observational diagnostics developed in Part II, that is designed (a) to systematically investigate and attribute changes in the intensity and frequency of simulated ecosystem productivity extremes ("EPEs") to the respective drivers, and (b) to assess the effect of timing and seasonal interaction of EPEs in the terrestrial biosphere. Thus, a perspective centred on ecosystem impacts is adopted.
An analysis of these simulations reveals that (a) recent trends in the intensity of EPEs in Europe are contrasting seasonally, i.e. spring EPEs show consistent trends towards increased carbon uptake, while trends in summer EPEs are predominantly negative (higher net carbon release under drought and heat in summer) or close to neutral. Furthermore, the analyses reveal that (b) spring-summer interacting carbon cycle effects due to climate extremes and thus their timing plays an important role in shaping EPEs in Europe. These interacting effects include both partial compensation of drought or heat wave induced carbon losses in summer due to increased carbon uptake in the preceding spring (driven by higher temperatures), and conversely, spring "carry-over" effects into summer arising from depleted soil moisture that exacerbates summer carbon losses.
In conclusion, the thesis lays out a comprehensive framework for systematically quantifying and attributing the impacts of climate extremes in the terrestrial biosphere using joint analyses of observations and model ensembles. The thesis shows that firstly, scrutinising statistical methods and diagnostics, and evaluating observation-based constraints on model ensembles, are key to an improved understanding as well as quantification of climate extremes and their impacts. Secondly, a consequent probabilistic interpretation of climate-ecosystem model ensemble simulations offers novel perspectives on the mechanistic pathways and interacting effects of terrestrial ecosystem responses to climate extremes. Show more
Permanent link
https://doi.org/10.3929/ethz-b-000176836Publication status
publishedExternal links
Search print copy at ETH Library
Contributors
Examiner: Seneviratne, Sonia I.
Examiner: Mahecha, Miguel D.
Examiner: Heimann, Martin
Examiner: Gruber, Nicolas
Publisher
ETH ZurichSubject
Climate extremes; Land-atmosphere interactions; ensembles; Climate change; carbon cycle; Bias correctionOrganisational unit
03778 - Seneviratne, Sonia / Seneviratne, Sonia
More
Show all metadata
ETH Bibliography
yes
Altmetrics