 |
| Titel |
Improving the spatial estimation of evapotranspiration by assimilating land surface temperature data |
| VerfasserIn |
Matthias Zink, Luis Samaniego, Matthias Cuntz |
| Konferenz |
EGU General Assembly 2013
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 15 (2013) |
| Datensatznummer |
250082222
|
|
|
|
|
|
| Zusammenfassung |
| A combined investigation of the water and energy balance in hydrologic models might lead
to a more accurate estimation of hydrological fluxes and state variables, such as
evapotranspiration ET and soil moisture. Hydrologic models are usually calibrated against
discharge measurements, and thus are only trained on the integrated signal at few points
within a catchment. This procedure does not take into account any spatial variability of fluxes
or state variables. Satellite data are a useful source of information to incorporate spatial
information into hydrologic models.
The objective of this study is to improve the estimation of evapotranspiration in the spatial
domain by using satellite derived land surface temperature Ts for the calibration
of the distributed hydrological model mHM. The satellite products are based on
data of Meteosat Second Generation (MSG) and are provided by the Land Surface
Analysis - Satellite Application Facility (LSA-SAF). mHM simulations of Ts are
obtained by solving the energy balance wherein evapotranspiration is determined by
closing the water balance. Net radiation is calculated by using incoming short-
and longwave radiation, albedo and emissivity data provided by LSA-SAF. The
Multiscale Parameter Regionalization technique (MPR, Samaniego et al. 2010)
is applied to determine the aerodynamic resistance among other parameters. The
optimization is performed for the year 2009 using three objective functions that
consider (1) only discharge, (2) only Ts, and (3) both discharge and Ts. For the spatial
comparison of satellite derived and estimated Ts fields, a new measure accounting for
local spatial variabilities is introduced. The proposed method is applied to seven
major German river basins, i.e. Danube, Ems, Main, Mulde, Neckar, Saale, and
Weser.
The results of the Ts simulations show a bias of 4.1 K compared to the satellite data. We
hypothesize that this bias is inherent to the satellite data rather than to the model simulations.
This is corroborated by the comparison of LSA-SAF Ts with measured data of air
temperature which shows a similar offset of 4.9 K. When optimizing for discharge (1) the
discharge simulations show the best fit (NSE exceeding 0.8) compared to the optimizations
using Ts (2) and Ts and discharge (3). But the spatial fields of evapotranspiration seem to
have random variability at some days. When optimizing only for Ts (2) high flows are well
represented while the estimation of low flows fails. Furthermore, this strategy reveals a
broader discharge uncertainty band compared to the discharge only optimization
(1). This indicates that optimizing for Ts (2) has predictive power regarding flood
estimations if no discharge data are available. Additionally in comparison to the
discharge only optimization (1) the spatial distribution of ET looks much more realistic.
Optimizing with discharge and Ts data simultaneously (3) preserves the narrow
discharge uncertainty band of the discharge only optimization (1) but also the more
realistic spatial distribution of ET of the Ts only optimization (2). Furthermore, the
uncertainty in the estimation of ET related model parameters is reduced by the combined
discharge and Ts optimization (3) compared to the discharge only optimization
(1).
In summary, the estimated spatial distributions of ET and connected state variables
such as soil moisture are improved by assimilating satellite data. Furthermore the
simulation of floods using only Ts in model optimization (2) is possible in ungauged
basins while the estimation of low flows and thus hydrological droughts will fail. |
|
|
|
|
|
|