2.4 Data combination
One of the primary benefits of the EOF analysis is a small number of
orthogonal spatial patterns was identified that together explain a large
proportion of the total variability of the soil moisture data. We now
examined how closely these underlying patterns resembled regional
characteristics that might dominate the spatial variability of soil
moisture. For this analysis, we used the correlation coefficient between
the EOFs and the available regional characteristics. Statistical
analyses were conducted using SPSS for Windows (SPSS Inc., Chicago). The
Pearson’s correlation was used to investigate the correlation between
soil moisture and soil-terrain attributes. Linear regression was used to
predict soil moisture (as a dependent variable) based on using either
soil properties and/or terrain attributes (as independent variables).
Furthermore, given that the first EOFs exhibited the general patterns of
soil moisture across the whole investigated spatial coverage, if the
EOFs of the manual soil moisture measurements at some sites had the
strong correlations with its automatically recorded values, then it was
possible to apply this relationship to the first EOFs to derive the soil
moisture across the entire area at any automated monitoring time.
Consequently, we considered that this method may have provided an
appropriate way to integrate the spatially-extensive (but
temporally-limited) manual datasets and the temporally-extensive (but
spatially-limited) automated monitoring datasets. The following
narrative describes in detail the four steps involved with the data
merging methods.
First, an EOF decomposition was performed on the soil moisture dataset
to identify the patterns of covariation (the EOFs) and their importance
on each date (the ECs).
Second, statistical tests were used to determine the EOFs between whole
spatially-extensive data and automated monitored data that were
statistically significant and should be retained in the transferring
method.
Third, a multiple linear regression was performed to identify empirical
relationships between automated monitored soil moisture data and its
EOFs.
Fourth, the identified empirical relationships were employed to
calculate soil moisture at each manual measured site within the entire
catchment via its EOFs.