【作者】 张明伟；

【导师】 蔡崇法； 周清波；

【作者基本信息】 华中农业大学 ， 资源环境信息工程， 2006， 博士

【摘要】 由于受自然条件和经济条件的限制，农业生产很大程度上依赖于天气气候条件。这种情况下，粮食安全面临巨大威胁，已经引起广泛的关注。及时监测农作物长势以及提早做出作物产量预测对政策制订和保持可持续发展十分重要。因此，无论发展中国家还是发达国家，几十年来，作物监测和产量预测一直受到高度重视。利用遥感技术来进行作物产量预测是最有发展前途的方法之一。 Terra／MODIS是一种新型和重要的卫星传感器，其数据性能较NOAA／AVHRR数据有了较大改善，空间分辨率为1000m，500m，250m，在波谱0.4-14.5范围内有36个波段，覆盖了可见光、近红外和短波红外波段，且波段均较窄，加上其每天一次对地区覆盖的高时间分辨率，MODIS数据在大尺度作物遥感监测和估产方面具有潜在优势。 在使用遥感信息获取作物生长发育状态信息时，如何将作物信息和其它信息分离，如何解决遥感信息和农学信息的匹配和转换是仍需进一步研究的问题。本文基于2003年、2004年MODIS数据进行中国华北地区主要作物物候期信息提取和作物类型识别研究。利用MODIS NDVI和EVI数据作为遥感参数，利用非线性方程模拟，提取作物关键物候期信息，并将物候遥感监测指标与地面观测指标进行对比分析，确定两者匹配关系。通过对作物生长期内植被指数变化曲线分析，选择合适的分类特征，进行华北地区主要作物类型识别。研究主要内容有： 1．分别利用Logistic方程和高斯方程对作物生长期内植被指数曲线进行模拟，并采用最大曲率法、动态阈值法提取作物关键物候期。分别利用MODIS NDVI和EVI序列提取冬小麦关键物候期，以农业气象观测值为参考值对监测结果进行分析，结果表明，利用这两种数据源获取作物生长季始末期与参考值比较结果相近，而利用MODIS EVI序列数据提取作物生殖生长转折点提取结果，较NDVI数据更接近参考值。两种曲线模拟方法利用曲率最大值法确定的冬小麦物候期均取得了较好结果，利用动态阈值法提取作物物候期与参考值相比相差稍大。 2．物候遥感监测结果与农业气象观测值相匹配。本文通过分析作物关键生育期的农学意义，及其在遥感植被指数序列数据的反映，分析农业气象观测指标及其大田表现，由此将物候遥感指标和观测指标相匹配。冬小麦与夏玉米关键物候期监测更多还原

【Abstract】 Because of the limit of natural and economic condition, the agricultural production strongly depends on the weather and climate. In this case, the food security faces great threat and it has attracted wide attention. It is very important to real-time monitor the crop growth and to make prediction of crop production for policy making and sustainable development. So the crop monitoring and prediction of production draws great attention all the time in not only developing country but also developed country. It is one of the most promising methods to make prediction of crop production with remote sensing.Terra/MODIS is a new remote sensing sensor. It can view the entire surface of the Earth in 36 spectral bands sampling the electromagnetic spectrum from 0.4 to 14μm with a spatial resolution ranging from 250 to 1,000 meters and high time resolution. So it has potential advantage in crop monitor on large-scale.How to separate crop information from the others and how to resolve the match and transformation between remote sensing information and agricultural information in monitoring crop growth are the next-step research when the information of the crop growth is extracted from remote sensing. In this paper, based on MODIS data of 2003 and 2004, phenological information of primary crop is extracted and crop type is identified in North China. With MODIS NDVI and EVI as parameters, together with non-linear formulation, the key crop phenological phases can be extracted and the remote sensing phenological monitoring index is contrasted with the land observation index to determine the matching relationship. Selecting the proper classification features, the primary crop types in North China could be identified through the analysis of the vegetation index time series in growth season. Main research results and initiatives in the thesis are as follows:1. Modeling the vegetation index curve respectively using a series of piecewise Logistic and Gaussian functions of time and extracting the key crop phenological phase through the methods of maximum curvature and dynamic threshold. The estimated phenological stages and statistical data are compared. The result shows that EVI time series is more efficient than the NDVI time series in estimating heading date. The method of maximum curvature gets better result than dynamic threshold. The two asymmetric functions fitting methods used to extract crop phenology are all successful.2. The phenology detected using MODIS time series is linked to statistical data. The relationship between crop development stages and temporal variation in satellite derived VI data is discussed. It was realized that the estimated phenology using MODIS time series is translated to statistic data. The comparison between the key phenological phases monitoring result and agricultural statistic data of winter wheat and summer maize development stages shows that the remote sensing monitoring result matches the agricultural statistic well. The remote sensing monitoring start phase of growth is更多还原