根据对2001年到2004年连续4年的卫星遥感影像的解译，研究了江苏省里下河农区兴化市的棉花轮作现状，建立了棉稻和稻棉轮作周期的估算模型，并给出了棉稻和稻棉轮作周期的估算公式。结果表明，在兴化市42个乡镇中，棉稻和稻棉为主要轮作模式，但轮作水平差异较大，在假定该区域的棉稻和稻棉轮作周期为平稳随机过程的情况下，可以推算出棉稻轮作周期为2.81年，稻棉轮作周期为2.89年，两者十分接近，仅相差2.85%，符合轮作的基本规律，说明本文的轮作评估模型和相关的估算公式是合理的，遥感影像解译的精度达到了试验的要求。
The objective of this paper was to estimate rotation periods of cotton-rice and rice-cotton from 2001 to 2004 in forty-two towns of XinHua City of JiangSu province of China, belonging to Lixiahe agriculture region, using satellite-based remote sensing (RS) data and geographic information system (GIS) tools. The vegetation in this region comprises mainly rice, cotton, soybean, sweet potato and vegetables. On an average from 2001 to 2004, the cotton land made up 12 488 ha/year, while the rice land area 83812 ha/year, i.e. 9.61% and 64.47% of total arable land area of XinHua City respectively. On the basis of data availability and accuracy requirements, four Landsat Thematic Mapper (TM) satellite images dated on 26/07/2001, 29/07/2002, 24/07/2003 and 26/07/2004 were selected as the data sources for the interpretation of rice and cotton. The satellite data were geographically registered onto the map. By using the knowledge about the temporal growth pattern of various crop types in this region, based on our field observations of various land cover classes, a supervised classification approach was followed to classify the scene into various land cover classes. The rice and cotton of vegetation in the images were classified by using the classification approach. The GIS was then used to integrate large amounts of interpreted image data and other related geographic details about the region to analyze the changes of land covers from cotton to rice or vice versa by the spatial polygon analysis model. The phenomenon of cotton-rice and rice-cotton rotation was found from the interpreted images. The rotation periods of cotton-rice and rice-cotton periods were supposed to obey the stationary stochastic process. The first and second models were built for estimation of cotton-rice and rice-cotton periods of each town, which was derived by cotton and rice coverage data provided by GIS that processed the interpreted images. The third model was built for computing the arithmetic average value of cotton-rice and rice-cotton rotation periods. The data provided by the models were used for analysis of cotton-rice and rice-cotton rotation periods of all the towns (Table 1 and Table 2). The results showed that the cotton-rice and rice-cotton were major rotation styles in this region. Among all the towns in XinHua City, there are obvious differences among the rotation levels or periods. The three average rotation periods of cotton-rice of all the towns from 2001 to 2004 were respectively 2.01, 3.42 and 3.01 years with an average of 2.81 year. The three average rotation periods of rice-cotton of all the towns from 2001 to 2004 were respectively 2.61, 2.70 and 3.35 years with an average of 2.89 year. The difference between the average values was very small or about 2.85%, indicating clearly that the result obeys the basic law that the cotton-rice rotation period should be statistically equal to the rice-cotton one, the models or formulas for estimation of the rotation periods are reasonable. The average rotation period of cotton-rice and rice-cotton rotation of all the towns was 2.85 year. The interpretation precision of remote sensing images arrived at the experimental requirement by validation from observations on the ground. In comparison with the traditional conclusion of 2–5 year in the past more than 400 years, the rotation period of 2.85 year given in the study is more scientific and precision.
全 文 ：