| Since the plastic mulching cultivation technology has been used in agriculture, the consumption of agricultural plastic film is up to 50 million tons around the world. In China,1.31 million tons of agricultural plastic film is used every year, the area of plastic mulched Landcover (PML) has been over 17.58 million hectares and PML has become an important agricultural landscape. Such large-scale PML inevitably has impacts on the climate and ecological environment. Therefore, how to quickly and efficiently detect the information of PML is particularly important.Xinjiang is one of main cotton producing areas in China, and it is also one of the most extensive areas of PML. This paper chooses Aksu River Oasis in southern Xinjiang as the study area and proposes a subpixel/subpixel spatial attraction models (SSSAM) based on Spatial Attraction Model to mapping PML. Initially, pure pixel index (PPI) is used to obtain PML, plant, water and bare land four endmembers from MODIS images. Furthermore, Linear Spectral Mixing Model (LSMM) is applied to obtain soft classified information. Additionally, SSSAM and traditional Sub-pixel mapping models, Such as sub-pixel/pixel spatial attraction model (SPSAM), modified sub-pixel/pixel spatial attraction model (MSPSAM), mixed spatial attraction model (MSAM) and Two-Point Histogram Model (TPHM) are developed in IDL 8.3 to mapping PML. By comparing and analyzing the classification results of SSSAM, traditional Sub-pixel mapping models, threshold method based on time series, the effectiveness of SSSAM is validated. Finally, based on surface temperature data, the heat effect of PML is calculated. Results includes:(1) The optimal classification result can be achieved after 4 iteration steps in SSSAM. The overall accuracies are achieved 77.69%,80.97%,86.43%,82.44% and the Kappa coefficients are achieved 0.56,0.59,0.66,0.66 for years 2000,2005, 2008, and 2014, respectively.(2)The overall accuracies of SSSAM are higher 7.61%,4.65%,6.27%,12.22% than SPSAM for years 2000,2005,2008, and 2014, respectively. Compared with MSPSIM and MASM, the operation time of SSSAM are saved nearly 1/3 and 2/3. Besides, the overall accuracies of SSSAM are higher 0.63%,0.27%,0.36%,0.31% than MSPSAM, and also are higher 0.07%,0.16%,0.24%,0.19% than MSAM for years 2000,2005,2008, and 2014, respectively.(3) When the land distribution is HR type, or hybrid of HR and LR type, the overall accuracies and Kappa coefficients of SSSAM are higher than those of TPHM; when the PML distribution is LR type, the Producers Accuracies and Precision Accuracies of SSSAM are lower than those of TPHM.(4) The Spatial resolution of SSSAM is higher than the threshold method based on time series. The classification accuracies of SSSAM are high when the error of soft classification is small, so SSSAM is an effective algorithm to identify PML from MODIS imagery. And SSSAM can be improved both from optimizing algorithm and improving soft classification accuracy.(5) The relative average surface temperature of PML is about 1 ℃ higher than that of canopy filed at daytime and night. But it is lower after the filed removed PML. Furthermore, with and without PML, the relative average surface temperature difference is greater than 0 at same filed. Therefore, the PML has heat effect. The area proportion of PML is positively correlated with surface temperature at both daytime and night, and the confidence levels are up to 0.01 and 0.05 respectively.SSSAM provides a new way to quickly and large-scalely detect PML in Xinjiang from MODIS imagery, and it also provides a strong basis for government to distribute fertilizer and pesticides, to manage the recycling of plastic mulches. |
PDF Full Text Request