|
Titel |
Long term modelling of precipitation and temperature at the continental scale |
VerfasserIn |
L. Nicòtina, A. Hilberts, S. Eppert, D. Lohmann |
Konferenz |
EGU General Assembly 2012
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 14 (2012) |
Datensatznummer |
250063375
|
|
|
|
Zusammenfassung |
Flood risk assessment at broad spatial scales and across catchments and countries requires the
formulation of coherent stochastic modelling tools to reproduce the variability of the climatic
features that drive runoff response. Such tools must i) prove robust in the reproduction of the
observed characteristics of weather forcings (e.g., rainfall and temperature in the case of
flood models); ii) allow an accurate representation of the space-time features of
the processes involved; iii) preserve the co-varying structure of the variables at
hand.
In the present work we study the space-time correlation structure of precipitation and
temperature data across Europe and we develop a stochastic model for the generation of
correlated random fields that, under the assumption of a stationary climate, preserve the
statistical characters of the measured data. Our analysis is based on the high-resolution
E-OBS gridded dataset (Haylock et al., 2008), consisting of daily precipitation and
temperature observations interpolated on a rectangular grid with a pixel size of
24Â km x 24Â km, covering the entire European continent and the observation period
1950 - 2010.
We analyze the spatial patterns of precipitation and temperature by means of a Principal
Component Analysis (PCA) performed on the monthly data anomalies (properly normalized
to guarantee the gaussianity of the dataset). The PCA method consists in decomposing the
actual data into a set of dominant modes of variation (in the form of Empirical Orthogonal
Functions, EOFs) and a vector of appropriate, normally distributed, weights (loading). In a
further generation step we use the statistical properties of the loading vector to generate a
stochastic set of precipitation and temperature monthly anomalies from which we can obtain
the relative time series of the forcing variable. We compare two methods for reproducing the
space-time correlations in the generation step: i) the target of the PCA is the combined
precipitation-temperature data matrix (i.e. a collection of time series for each spatial location)
obtained appending the individual data series for each location. This constrains the
analysis on identifying dominant patterns that reflect both the precipitation and
temperature spatial characters; ii) EOFs and the loadings for the two variables are derived
independently and the correlation matrix of the loadings is imposed in the generation
step.
Our results show that the stochastic model based on the principal component analysis
offer an effective tool for modelling atmospheric forcings at the spatial and temporal scales
involved. In particular we observed a good capability of reproducing precipitation extremes
and the patterns that characterize extreme events over broad spatial scales. We prove that the
two methods used for reproducing correlations between precipitation and temperature can be
selected depending on the desired target, the first method being more reliable in the
representation of the existing spatial correlation between the two variables, whereas the latter
one ensures a better representation of the individual fields at the expenses of the
correlation.
References
Haylock, M. R., N. Hofstra, A. M. G. Klein Tank, E. J. Klok, P. D. Jones, and M. New
(2008), A European daily high-resolution gridded data set of surface temperature and
precipitation for 1950 – 2006, J. Geophys. Res., 113, D20119, doi:10.1029/2008JD010201 |
|
|
|
|
|