(
R
2
increases 10%–15% and
RMSE
decreases 0.004–0.01 cm
3
/
cm
3
). It is mainly because the method needs to simultane-
ously satisfy the constraints about the backscatter coefficient
and the surface reflectance of optical data. The respective
constraints enable the inversion results to complement each
other and obtain higher inversion accuracy. Meanwhile, it can
be seen that the inversion result using radar remote sensing
is better than that using optical data. It may be because the
radar remote sensing is sensitive to the soil moisture and can
comprehensively depict the parameters related to the back-
scatter coefficient. Therefore, it is necessary to consider the
influences of optical and radar data and make full use of their
respective advantages. Thus, the
MPSO
-
BP
method based on
radar and optical data can achieve complementary errors and
improve the inversion accuracy of soil moisture.
Soil Moisture Maps
Figure 6 is a spatial distribution map of surface soil moisture
obtained using the
MPSO
-
BP
algorithm. It can be obtained
from Figure 6 that the soil moisture distribution is entirely
consistent with the actual surface. The farmland area and
the nonfarmland area are clearly separated and the topologi-
cal characteristic is obvious to be identified. The value of
soil moisture in the farmland area is within 0.2–0.4 cm
3
/cm
3
,
which is close to the field measurement. The value of soil
moisture is relatively high in wheat-planted areas. It is mainly
because the winter wheat is in the rapid growth period and
has a large demand for soil moisture to make the farmland
area be regularly irrigated.
(a) MPSO-BP+MIXED
(b) MPSO-BP+OPTICAL
Figure 5. Validation of soil moisture retrieval calculated from
SAR
images and field measurements. The unit of
RMSE
is cm
3
/cm
3
.
Figure 6. Map of surface soil moisture content retrieval using
MPSO-BP
.
PHOTOGRAMMETRIC ENGINEERING & REMOTE SENSING
November 2019
795
