season. Due to the long revisiting time of Landsat and high-spatial-resolution
satellite sensors, it is hard to collect cloud-free images for mapping PML over
large geographic areas. Therefore, Lu et al. (2015) proposed a threshold model for
mapping PML from moderate resolution imaging spectroradiometer (MODIS) time
series data and proved that MODIS data can be used to map PML for large
geographic areas.
This study chose Xinjiang, China, as the research area for three reasons. First,
more than 11% of PML in China is located in Xinjiang, which accounts for over 40%
of the total annual cotton production in China and where plastic mulch is used 100%
in cotton fields. Second, southern Xinjiang is the key cotton plantation area in
Xinjiang, accounting for over 70% of cotton acreages in the province (Yang et al.
2011). Third, the size of individual cotton fields (the average width of a mulched
farmland is around 100 meters) or their aggregation in southern Xinjiang is big
enough to be classified in MODIS images. In this study, we focus on analyzing the
PML mapping methods, including the decision-tree classifier (Lu et al. 2014), the
threshold models (Lu et al. 2015), and the spatial attraction models (Lu et al. 2017),
for detecting the transparent PML over large geographic areas using Landsat/
MODIS imagery.
17.2 A Decision-Tree Classifier for Extracting PML Using
Landsat Imagery
17.2.1 Methodology
17.2.1.1 The Detectable Features of PML
In order to obtain the rules for the decision-tree classifier, the detectable features of
PML in Landsat TM images have to be discovered. To facilitate the discovery, both
the true and false color composite images and then visually inspected the composites
are synthesized. A true color composite image was synthesized by using Landsat TM
images band 3, band 2, and band 1. In addition, two false color composite images
were synthesized by using band 7, band 5, and band 3, as well as band 7, band 4, and
band 3. Figure 17.2 shows the Landsat color composite images of May 10, 2011 for
a part of the study area. From Fig. 17.2, we can find that the composite of band 7 – R,
band 4 – G, and band 3 – B (RGB: band 743) preserves ground geographic features
and surface colors better than other composites. Therefore, RGB: band 743 composite is selected as the base imagery for future inspection.
As shown in Fig. 17.2b, the PML can be easily seen in the RGB: band 743 composite. Because almost all plastic mulch used in the research area is transparent and
dependent on the crop growing condition, soil types, and soil moisture, the PML
354
L. Lu
satellite sensors, it is hard to collect cloud-free images for mapping PML over
large geographic areas. Therefore, Lu et al. (2015) proposed a threshold model for
mapping PML from moderate resolution imaging spectroradiometer (MODIS) time
series data and proved that MODIS data can be used to map PML for large
geographic areas.
This study chose Xinjiang, China, as the research area for three reasons. First,
more than 11% of PML in China is located in Xinjiang, which accounts for over 40%
of the total annual cotton production in China and where plastic mulch is used 100%
in cotton fields. Second, southern Xinjiang is the key cotton plantation area in
Xinjiang, accounting for over 70% of cotton acreages in the province (Yang et al.
2011). Third, the size of individual cotton fields (the average width of a mulched
farmland is around 100 meters) or their aggregation in southern Xinjiang is big
enough to be classified in MODIS images. In this study, we focus on analyzing the
PML mapping methods, including the decision-tree classifier (Lu et al. 2014), the
threshold models (Lu et al. 2015), and the spatial attraction models (Lu et al. 2017),
for detecting the transparent PML over large geographic areas using Landsat/
MODIS imagery.
17.2 A Decision-Tree Classifier for Extracting PML Using
Landsat Imagery
17.2.1 Methodology
17.2.1.1 The Detectable Features of PML
In order to obtain the rules for the decision-tree classifier, the detectable features of
PML in Landsat TM images have to be discovered. To facilitate the discovery, both
the true and false color composite images and then visually inspected the composites
are synthesized. A true color composite image was synthesized by using Landsat TM
images band 3, band 2, and band 1. In addition, two false color composite images
were synthesized by using band 7, band 5, and band 3, as well as band 7, band 4, and
band 3. Figure 17.2 shows the Landsat color composite images of May 10, 2011 for
a part of the study area. From Fig. 17.2, we can find that the composite of band 7 – R,
band 4 – G, and band 3 – B (RGB: band 743) preserves ground geographic features
and surface colors better than other composites. Therefore, RGB: band 743 composite is selected as the base imagery for future inspection.
As shown in Fig. 17.2b, the PML can be easily seen in the RGB: band 743 composite. Because almost all plastic mulch used in the research area is transparent and
dependent on the crop growing condition, soil types, and soil moisture, the PML
354
L. Lu