Abstract:Crop growth information is closely related to crop yield prediction and food security. Accurately investigating and monitoring crop growth information provides invaluable information for agricultural management, planning and decisionmaking, and further ensures sustainable development of agriculture. Guanzhong area is a main food production region of Northwest China, however, the varied topography, including loess hilly-gully region, Weihe plain and Qinling mountain, makes crop information extraction challenged and complicated. In order to obtain winter wheat planting information of Guanzhong area in 2014, thirty phases of MODIS NDVI image from September 2013 to December 2014 were used, the images were reconstructed by iterativeing Savitzky-Golay filter in time sequence, and the characteristics of NDVI profiles of the typical objects selected from filtered images were analyzed. And then the noncropland was masked based on the characteristics analyzed above. On account of the special NDVI profile of winter wheat, the twi-difference algorithm was applied to the selection of 17 phases’ image during period of November 2013 and July 2014, and the NDVI peaks were identified. From the local crop calendar and the characteristics of typical NDVI profiles, it was found that except winter wheat and rapeseeds, the rest of the crops did not show NDVI peaks during the selected period. Interestingly, winter wheat had two peaks in its growing season. The first one was appeared before the coming winter because of the snow or frost, and the second one was the normal growing peak. Rapeseeds had a single peak which was lower than that of winter wheat. Thus, the winter wheat of Guanzhong area was identified at the pixel level by combining the pixels which had double peaks and single peak with NDVI value under 0.65. For verifying the accuracy of extraction, the winter wheat growth area was measured at four typical plots of 1km×1km, which were used to compare with the remote sensing results. Besides, totally 100 winter wheat NDVI profiles were sampled to compare with the artificial interpretation at the pixel level. The results showed that the winter wheat was mainly distributed at two regions in Guanzhong area. One was located in the river valley of middle reaches of Weihe river, and the other one was located in the wide Weibei plain of the east side of Jinghe river. Meanwhile, the winter wheat presented a scattered pattern at the loess hilly-gully area in the north, and did not appear at the mountainous area in the south. The total area of winter wheat in Guanzhong area in 2014 was 8.882 × 105hm2, and its areas in Xi’an, Tongchuan, Xianyang, Baoji and Weinan were 1.958×105hm2, 2.296×105hm2, 1.704×105hm2, 2.222×105hm2 and 2.769×105hm2, respectively. Compared with the statistics of government, the total relative error of Guanzhong area was 3.70%, with variation of 1.08%~9.02% at the city level. The maximum relative error was 9.02% which was occurred at Baoji. The accuracies of four plots of Changwu, Yintai, Fufeng and Yangling were 17.3%, 15.1%, 5.6% and 7.7%, respectively. Besides, the accuracy of sampled 100 NDVI profiles at the pixel level reached 90.28%. The results indicated that exacting the crop information using NDVI time-series at a topography complicated region was feasible and it can be applied to the other regions of Northwest China for making plan of crop planting, agricultural resource allocation and monitoring crop growing.
