 |
| Titel |
Effects of Measurement Uncertainties of Meteorological Data on Estimates of Site Water Balance Components |
| VerfasserIn |
Uwe Spank, Kai Schwärzel, Maik Renner, Christian Bernhofer |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250076988
|
|
|
|
|
|
| Zusammenfassung |
| Numerical water balance models are widely used in ecological and hydro sciences. However,
their application is related to specific problems and uncertainties. The reliability of model
prediction depends on (i) model concept, (ii) parameters, (iii) uncertainty of input data, and
(iv) uncertainty of reference data. How model concept (i) and parameters (ii) affect
the model performance is an often treated problem. On the contrary, the effects of
(iii) and (iv) are seldom tackled although their effects are of similar magnitude. It
should be considered that uncertainties of input data and reference data do not only
affect the prediction accuracy but also the parameter identification (calibration and
validation).
The uncertainty has two different reasons: (a) actual measurement uncertainties and (b)
limitations of representativeness as consequence of a scale gap between meteorological
measurement and hydrological modelling. A separate analysis of both aspects is often not
possible as most hydrological investigations operate on catchment scale where both effects
interfere with each other. Our study is focused on site scale (< 1.5 km2). Thus, the scales of
measurement and modelling are similar, and effects due to regionalisation and generalisation
can be neglected.
At site scale we take the micrometeorological perspective: primary reference is the
evapotranspiration measured via the eddy covariance technique instead of runoff. Because of
the use of evapotranspiration as a reference, it is possible to limit the investigations to
the upper parts of the soil that are influenced by root water uptake. Thus, also the
parameter uncertainty is significantly reduced as most parameters can be directly
quantified.
The analyses of effects due to input uncertainties are based on Monte-Carlo-Simulations
with perturbated input series. The Monte-Carlo-Simulations were done for two water balance
models of different complexity (HPTFs: black box model; BROOK90: process based
complex model) and for different sets of parameterisation. Our results show that seemingly
small uncertainties in daily measurements can lead to significant discrepancies on annual
scale. Uncertainties in precipitation but also in radiation measurements are especially serious.
E.g., a small offset of 5Â WÂ m-2 in measured daily radiation sum up to an uncertainty of
160Â MJÂ m-2 (equivalent to 65Â mm) on annual scale. The uncertainty of radiation
measurements is thereby the main reason for uncertainties in calculated potential but also
actual evapotranspiration.
It is also demonstrated that effects of input uncertainties are similar for both model types
and all variants of parameterisation. The typical spread of simulation results (here defined as
the range between median and upper/ lower percentile) being caused by uncertainties of
meteorological measurement is on average ±Â 25 mm (5 %) for annual actual
evapotranspiration and ±Â 30 mm (10 %) for annual seepage. However, the maximum
spread can be significantly higher particularly in years being affected by rare and
extreme weather conditions. In summary, our study shows clearly: uncertainties in
meteorological data are not negligible. Input uncertainties must be considered in the
same way as effects of parameterisation and shortcomings of the model concept. |
|
|
|
|
|
|