 |
| Titel |
Calibration and Comparison of three SVAT(\textbf{S}urface-\textbf{V}egetation-\textbf{A}tmosphere\textbf{T}ransfer) models |
| VerfasserIn |
Paul Dobesberger, Antonia Zeidler, Georg Wohlfahrt |
| Konferenz |
EGU General Assembly 2011
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 13 (2011) |
| Datensatznummer |
250054548
|
|
|
|
|
|
| Zusammenfassung |
| This presentation is part of the research project RIMES (Climate Change and Natural
Hazards Risk Management in Energy Systems) funded by the Austrian Climate
Research Program (ACRP). The project aims at the optimization of risk management
procedures by addressing uncertainties of various domains (climate change, natural
hazards, economical losses) and the standardization of a method to determine the
vulnerability of an energy system. Climate change it is likely to affect the return
period and magnitude of natural hazards, such as avalanches, debris flow or sediment
transport. Important factors to be considered when researching those hazards are the
alteration of soil and snowpack properties as well as the varying conditions for the
vegetation.
Here we will use SVAT models for assessing climate change impacts on the soil, the
snowpack and the growing conditions for the vegetation on a long term perspective. For this
we will compare three SVAT models with different theoretical backgrounds and degrees of
complexity. The first objective of the present study is a systematical and comprehensible
calibration of the models with data from different Alpine stations in Tyrol/Austria. Therefore
a Bayesian model calibration framework (DREAM – Differential Evolution Adaptive
Metropolis) via Markov chain Monte Carlo method will be used. This algorithm runs
multiple chains simultaneously for global exploration, and automatically tunes the scale
and orientation of the proposal distribution in order to find the set of parameters
which fits best to the target (e.g. soil water content). Hence the outcome of these
simulations will be the range for each parameter of the model to fit the claimed
target. Here the major interest will be in how well parameters of the three models
differing in complexity are constrained by the same set of calibration data. The
second objective of this study is the comparison of the measured and simulated
water and energy balance parameters of the soil, the snowpack and the vegetation.
Therefore several well accepted statistical methods, like root mean squared error,
model efficiency or normalized mean average error will be used to compare the
simulated and observed results and to point out the advantages and disadvantages of
each particular SVAT model. Special attention will be paid to the differences in the
model design and their influence on the reliability and accuracy of the simulated
outputs.
The major objectives of this presentation will be the calibration of the models via the
DREAM algorithm as well as to highlight the strengths and weaknesses of the particular
model in simulating the processes in an Alpine environment and to point out where it is
useful to employ complex physically based models and where a simple model produces
significant results. |
|
|
|
|
|
|