Short-Interval Monitoring of Land Use and Land Cover Change Using a Time Series of RADARSAT-2 Polarimetric SAR Images - Space-Time Integration in Geography and GIScience - page 361

urban/built-up areas and that from vegetation to urban/built-up areas are important
for detecting potential illegal land development, and they can be combined together
in practical application. The user' and producer's accuracies for changes from
barren land or vegetation to urban/built-up areas were also calculated. Figure 19.7
shows that the average user's and producer's accuracies for barren land or vegetation
to urban/built-up areas were 83.40 % and 91.25 % respectively. However, the user's
and producer's accuracies on May 08 were lower than that on the other dates
because paddy growth caused some false alarms to changes from barren land to
urban/built-up areas. Figure 19.7 also shows the user's and producer's accuracies
were stable and high for determining changes from vegetation to barren land. The
average user's and producer's accuracies for changes from vegetation to barren land
were 86.00 % and 90.90 % respectively.

19.5 Conclusions

The primary objective of the present study was to explore the potential use of
RADARSAT-2 PolSAR images in short-interval monitoring of LULC change. A
time series of RADARSAT PolSAR images was used to extract monthly LULC
changes. High accuracies were achieved for monthly LULC change detection by
using RADARSAT-2 time series images. The average detection accuracy, average
false alarm rate, and average overall error rate for detecting changed areas were
91.29 %, 1.37 %, and 1.97 % respectively. The average overall accuracy and average
kappa value for determining types of changes were 72.44 % and 0.68 % respectively.
Given that illegal land development is commonly related to LULC changes from
barren land to urban/built-up areas, that from vegetation to barren land, and that
from vegetation to urban/built-up areas, the producer's and user's accuracies for
these types of LULC change were also calculated. The average user's accuracy
and average producer's accuracy for changes from barren land or vegetation to
urban/built-up areas were 83.40 % and 91.25 % respectively. The average user's
accuracy and average producer's accuracy for changes from vegetation to barren
land were 86.00 % and 90.90 % respectively. The results indicate that it is effective
in using a time series of RADARSAT-2 PolSAR images in short-interval monitoring
of LULC change, especially in monitoring potential illegal land development.
Paddy cultivation and growth had a great influence on the use of RADARSAT-
2 PolSAR images in the monthly monitoring of LULC change. Many false alarms
were found in paddy fields due to the growth, planting, and harvesting of paddies,
and many of them were mistakenly identified as other LULC change types. This
degraded the accuracy of monthly LULC change detection using RADARSAT-2
PolSAR images. Future studies will be carried out on utilizing more multi-temporal
PolSAR images, such as three or four images, instead of two images to extract short-
interval LULC changes. Compared with LULC changes caused by human activities,
changes caused by paddy growth and cultivation change regularly according to