![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Modeling water and heat balance components of large territory for vegetation season using information from polar-orbital and geostationary meteorological satellites |
VerfasserIn |
Eugene Muzylev, Zoya Startseva, Alexander Uspensky, Elena Volkova, Alexander Kukharsky, Sergey Uspensky |
Konferenz |
EGU General Assembly 2015
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 17 (2015) |
Datensatznummer |
250104192
|
Publikation (Nr.) |
EGU/EGU2015-5413.pdf |
|
|
|
Zusammenfassung |
To date, physical-mathematical modeling processes of land surface-atmosphere interaction is
considered to be the most appropriate tool for obtaining reliable estimates of water and heat
balance components of large territories. The model of these processes (Land Surface Model,
LSM) developed for vegetation period is destined for simulating soil water content W,
evapotranspiration Ev, vertical latent LE and heat fluxes from land surface as well as
vertically distributed soil temperature and moisture, soil surface Tg and foliage Tf
temperatures, and land surface skin temperature (LST) Ts. The model is suitable for utilizing
remote sensing data on land surface and meteorological conditions. In the study these data
have been obtained from measurements by scanning radiometers AVHRR/NOAA,
MODIS/EOS Terra and Aqua, SEVIRI/geostationary satellites Meteosat-9, -10
(MSG-2, -3). The heterogeneity of the land surface and meteorological conditions has
been taken into account in the model by using soil and vegetation characteristics as
parameters and meteorological characteristics as input variables. Values of these
characteristics have been determined from ground observations and remote sensing
information.
So, AVHRR data have been used to build the estimates of effective land surface
temperature (LST) Ts.eff and emissivity E, vegetation-air temperature (temperature at the
vegetation level) Ta, normalized vegetation index NDVI, vegetation cover fraction B, the leaf
area index LAI, and precipitation. From MODIS data the values of LST Tls, Å, NDVI, LAI
have been derived. From SEVIRI data there have been retrieved Tls, E, Ta, NDVI, LAI and
precipitation. All named retrievals covered the vast territory of the part of the agricultural
Central Black Earth Region located in the steppe-forest zone of European Russia. This
territory with coordinates 49°30’–54°N, 31°-43°E and a total area of 227,300 km2 has been
chosen for investigation. It has been carried out for years 2009-2013 vegetation
seasons.
To provide the retrieval of Ts.eff, E, Ta, NDVI, B, and LAI the previously developed
technologies of AVHRR data processing have been refined and adapted to the region
of interest. The updated linear regression estimators for Ts.eff and Tà have been
built using representative training samples compiled for above vegetation seasons.
The updated software package has been applied for AVHRR data processing to
generate estimates of named values. To verify the accuracy of these estimates the error
statistics of Ts.eff and Ta derivation has been investigated for various days of named
seasons using comparison with in-situ ground-based measurements. On the base of
special technology and Internet resources the remote sensing products Tls, E, NDVI,
LAI derived from MODIS data and covering the study area have been extracted
from LP DAAC web-site for the same vegetation seasons. The reliability of the
MODIS-derived Tls estimates has been confirmed via comparison with analogous and
collocated ground-, AVHRR-, and SEVIRI-based ones. The prepared remote sensing
dataset has also included the SEVIRI-derived estimates of Tls, E, NDVI, Ta at
daylight and night-time and daily estimates of LAI. The Tls estimates has been built
utilizing the method and technology developed for the retrieval of Tls and E from 15
minutes time interval SEVIRI data in IR channels 10.8 and 12.0 µm (classified as
100% cloud-free and covering the area of interest) at three successive times without
accurate a priori knowledge of E. Comparison of the SEVIRI-based Tls retrievals with
independent collocated Tls estimates generated at the Land Surface Analysis Satellite
Applications Facility (LSA SAF, Lisbon, Portugal) has given daily- or monthly-averaged
values of RMS deviation in the range of 2°C for various dates and months during
the mentioned vegetation seasons which is quite acceptable result. The reliability
of the SEVIRI-based Tls estimates for the study area has been also confirmed by
comparing with AVHRR- and MODIS-derived LST estimates for the same seasons. The
SEVIRI-derived values of Ta considered as the temperature of the vegetation cover has
been obtained using Tls estimates and a previously found multiple linear regression
relationship between Tls and Ta formulated accounting for solar zenith angle and land
elevation. A comparison with ground-based collocated Ta observations has given RMS
errors of 2.5°C and lower. It can be treated as a proof of the proposed technique’s
functionality. SEVIRI-derived LAI estimates have been retrieved at LSA SAF from
measurements by this sensor in channels 0.6, 0.8, and 1.6 μm under cloud-free conditions at
that when using data in the channel 1.6 μm the accuracy of these estimates has
increased.
In the study the AVHRR- and SEVIRI-derived estimates of daily and monthly
precipitation sums for the territory under investigation for the years 2009 - 2013 vegetation
seasons have been also used. These estimates have been obtained by the improved
integrated Multi Threshold Method (MTM) providing detection and identification
of cloud types around the clock throughout the year as well as identification of
precipitation zones and determination of instantaneous precipitation maximum
intensity within the pixel using the measurement data in different channels of named
sensors as predictors. Validation of the MTM has been performed by comparing
the daily and monthly precipitation sums with appropriate values resulted from
ground-based observations at the meteorological stations of the region. The probability of
detecting precipitation zones from satellite data corresponding to the actual ones
has been amounted to 70-80%. AVHRR- and SEVIRI-derived daily and monthly
precipitation sums have been in reasonable agreement with each other and with
results of ground-based observations although they are smoother than the last values.
Discrepancies have been noted only for local maxima for which satellite-based
estimates of precipitation have been much less than ground-based ones. It may be
due to the different spatial scales of areal satellite-derived and point ground-based
estimates.
To utilize satellite-derived vegetation and meteorological characteristics in the model the
special procedures have been developed including:
- replacement of ground-based LAI and B estimates used as model parameters by their
satellite-derived estimates from AVHRR, MODIS and SEVIRI data. Correctness of such
replacement has been confirmed by comparing the time behavior of LAI over the period of
vegetation as well as modeled and measured values of evapotranspiration Ev and soil
moisture content W;
- entering AVHRR-, MODIS- and SEVIRI-derived estimates of Ts.eff Tls, and Ta into
the model as input variables instead of ground-measured values with verification of adequacy
of model operation under such a change through comparison of the calculated and measured
values of W and Ev;
- inputing satellite-derived estimates of precipitation during vegetation period retrieved
from AVHRR and SEVIRI data using the MTM into the model as input variables. When
developing given procedure algorithms and programs have been created to transit from
assessment of the rainfall intensity to evaluation of its daily values. The implementation of
such a transition requires controlling correctness of the estimates built at each time step. This
control includes comparison of areal distributions of three-hour, daily and monthly
precipitation amounts obtained from satellite data and calculated by interpolation of standard
network observation data;
- taking into account spatial heterogeneity of fields of satellite AVHRR-, MODIS- and
SEVIRI-derived estimates of LAI, B, LST and precipitation. This has involved the
development of algorithms and software for entering the values of all named characteristics
into the model in each computational grid node.
Values of evapotranspiration E, soil water content W, vertical latent and sensible heat
fluxes and other water and heat balance components as well as land surface temperature and
moisture area-distributed over the territory of interest have been resulted from the model
calculations for the years 2009-2013 vegetation seasons. These calculations have been
carried out utilizing satellite-derived estimates of the vegetation characteristics,
LST and precipitation. E and W calculation errors have not exceeded the standard
values. |
|
|
|
|
|