The International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences. Vol. XXXVII. Part B7. Beijing 2008
moisture time series with regional soil moisture in situ data (e.g.
Scipal 2002, Wagner et al. 2003, Zhao et al. 2006).
Compared to synthetic aperture radar (SAR) systems
scatterometers offer multiple incidence angles for each overpass,
which enables to better account for the effects of vegetation and
surface roughness. Furthermore, contrary to SAR, lower
resolution scatterometer sensors allow to map the Earth surface
within less than three days. Here a coarse spatial resolution of
50 km (ERS Scat) or 25 km (ASCAT) is accepted, since an
excellent temporal resolution can be achieved. Soil moisture
can be investigated at two different spatial scales. The first is
the spatial scale below 100 meters, where spatial and timely soil
moisture variability are mainly driven by vegetation, soil type
and topography (Scipal et al. 2005, Vachaud et al. 1985). The
second scale at several kilometers represents soil moisture
variability induced by atmospheric forcing effects, thus mainly
being influenced by climatic conditions and large scale
precipitation events (Vinnikov et al. 1999, Ceballos et al. 2002).
Scatterometer derived soil moisture data at the scale of 25 to 50
km therefore contains information about large scale
meteorological events. Furthermore, especially spatio-temporal
changes in longer time series of data with expected seasonal
soil moisture patterns can indicate the occurrence of slow onset
natural hazards such as floods or droughts.
1.2 The TU Wien Dataset: ERS-Scatterometer derived
Surface Moisture
The active ERS scatterometer with three sideways looking
antennae collects backscatter measurements in the 5.3 GHz
domain (C band) with vertical polarization over an incidence
angle range from 18° to 57°. Global coverage is achieved every
3-4 days (Scipal et al. 2005). Strictly speaking, the
backscattered signal cP is mainly a function of dielectric
properties of materials depending on frequency, /, polarization,
pp, and incidence angle 6. The dielectric constant of a material
mainly depends on its water content. The function S( f ,0)
describes backscattering according to surface roughness and is
also influenced by frequency and incidence angle. This basic
principle of dielectric properties and geometric surface structure
is used by the majority of electromagnetic backscattering
models to derive soil moisture (Knabe 2004).
a o(PPf > 9) =D(f, pp, 0) • S(f, 0) (1)
Since one is only interested in the part of the signal, which
represents the moisture content other influences need to be
corrected for. Heavily vegetated areas like rainforests are
masked out from the TU Wien data set. In dense forest areas
volume scattering dominates and the backscattered signal from
the ground covers a too small portion of the overall
backscattered signal. Furthermore, snow covered areas are
masked to exclude areas, where no statement about soil
moisture is possible. Coastal zones and inland water bodies are
also excluded. Incidence angle dependencies and effects of
surface roughness, heterogeneous vegetation cover, and land
cover are fully accounted for with the change detection
approach implemented and presented by Wagner et al. (1999).
Thus, after corrections the relationship between backscatter
(normalized to an incidence angle of 40°, cP (40)) and soil
moisture variability is linear (Scipal 2002). The change
detection approach thus only requires a time series of data to be
available. From this time series surface soil moisture
information equivalent to the degree of saturation in relative
units, ranging between 0-100 %, can be retrieved. In this
change detection method the current backscattering coefficient
is compared to the highest and lowest measurement record
(referred to as cP wet and a° dry respectively) for this spatial
location within the available time series. If cP^ and cP^ A
represent a completely dry soil surface and a saturated soil
surface then m s is equal to the degree of saturation, equaling the
soil moisture content in percent of porosity. m s can be derived
from every backscatter measurement for a point on earth and is
thus available every 3-4 days. From the TU Wien Global Soil
Moisture Archive surface moisture data sets can be extracted on
a weekly, ten-day, or monthly basis for every defined area.
2. METHODS FOR TIME SERIES ANALYSES
2.1 Method: Anomaly Extraction and Analyses
Figure 1. “La Nina” (post El Niño event) related drought in
south-eastern China as observed in soil moisture anomaly data
(top), and highly correlated also occurring in GPCC data
(gridded precipitation data of the German Weather Service)
As a first focus of analyses, we extracted major anomalies from
the time series. The term anomaly refers to the deviation of
surface soil moisture at a given spatial location with respect to
the time series mean of all soil moisture values for this month.
The extracted strong anomalies reflect severe drought and flood
conditions over the course of over 15 years in many countries
worldwide. The following figure 1 shows an example,
representing a “La Niña” related drought situation in China in
February 1999 after the very strong 1997/1998 El Niño. In the
upper part of the figure green areas are masked out areas (snow
cover), grey areas indicate soil moisture conditions within the
normal range, while blue areas indicate wetter than normal
conditions and yellowish to brown areas indicate drier than
normal conditions.
1364