JIA-2019-11

2638 ZHANG Xi-wang et al. Journal of Integrative Agriculture 2019, 18(11): 2628–2643 112°30´E 112°30´E 112°00´E 112°00´E 111°30´E 111°30´E 34°40´N N 34°40´N 34°20´N 34°20´N 0 25 50 12.5 km 1.0 0.0 Fig. 7 The membership for winter wheat at the 25-m scale. Fig. 8 The identification result of winter wheat in major agricultural area of Yiluo Basin, China. 112°30´E 112°30´E 112°00´E 112°00´E 111°30´E 111°30´E 34°40´N 34°40´N 34°20´N 34°20´N 0 25 50 12.5 km Border of Yiluo Basin Winter wheat N Table 3 Area accuracy of identified winter wheat 1) County name Official statistics area (ha) MLC RFC Proposed method Identified area (ha) Aa (%) Identified area (ha) Aa (%) Identified area (ha) Aa (%) Yiyang 37 208.46 44 956.90 79.18 42 835.51 84.88 42 591.19 85.53 Luoning 30 269.37 35 211.52 83.67 33 619.22 88.93 31 487.63 95.98 Yichuan 36 799.34 45 030.18 77.63 43 806.39 80.96 43 225.00 82.54 Yanshi 41 094.04 45 207.00 89.99 44 636.17 91.38 38 234.81 93.04 Total 145 371.20 170 405.60 82.78 164 897.29 86.57 155 538.63 93.01 RMSE (%) 4.47 3.52 3.07 1) MLC, Maximum Likelihood classification; RFC, Random Forest classification; Aa, acreage accuracy; RMSE, root mean square error. Table 4 Sampling accuracy of identified winter wheat 1) County name 2) Total number of validation points MLC RFC Proposed method NCIVP As (%) NCIVP As (%) NCIVP As (%) Yiyang 135 109 80.74 116 85.93 121 89.63 Luoning 138 121 87.68 128 92.75 131 94.93 Yichuan 115 91 79.13 95 82.61 98 85.22 Yanshi 112 102 91.07 104 92.86 107 95.54 Total 500 423 84.65 443 88.60 457 91.40 RMSE (%) 4.05 3.04 2.36 1) MLC, maximum likelihood classification; RFC, Random Forest classification; NCIVP, the number of correctly identified validation points; As, sampling accuracy. 2) RMSE, root mean square error. phenologies and relatively small planting acreages. The study area inclines from southwest to northeast, resulting in a large centralized planting area located in the eastern basin, whereas additional plantings are scattered in other areas. The acreages of the individual planting areas decrease with the increasing complexity of the terrain. In Luoning, Yiyang and Yichuan counties, the topography is predominantly mountainous and hilly, whereas small amounts of flat farmland are distributed on both sides of the Yiluo River. Small, scattered farmland plots lead to an increase of mixed pixels in the MODIS data. In contrast, there is more flat farmland in Yanshi County. Flat terrain results in many large areas of farmland, concentrations of planted winter wheat, and relatively high identification accuracy in Yanshi County. According to the field investigation and remote sensing images, the planting structure is simple in Luoning County. Even in the mountainous area, smaller growing areas of winter wheat can also reflect distinct characteristics. Therefore, Luoning County still possesses high identification accuracy, despite being more mountainous and having less flat farmland. In Yiyang and Yichuan counties, most farmland is distributed in the valley and the farmland acreage is small. Interval planting with other crops also results in mutual influences between the pixel spectra, and thus the identification accuracy is relatively low. Therefore, crop structure is one of the most important factors affecting crop type identification.