【作者】 柏军华；

【导师】 李少昆；

【作者基本信息】 中国农业科学院 ， 作物信息科学， 2009， 博士

【摘要】 农田管理是农作的主要内容,其水平是农业生产力的集中表现,其效果直接影响农作物的产量、品质以及生产效益。因此,利用现代科学技术成果“武装”农田管理的各个环节是农业生产顺应社会需求与社会经济发展的必然要求和结果。卫星遥感技术是信息科学技术发展的璀璨结晶,作物的长势、产量、种植面积等方面是卫星遥感转入民用的重要应用领域,为农业主管单位快速和宏观的掌握生产信息提供了有效方式。随着作物遥感监测研究及应用的不断深入,遥感监测的内容已不仅仅停留在宏观结果的预报与总结上,更注重作物生产的局部环节,作物遥感监测领域正在孕育“细化监测内容,参与过程管理”的理念,使遥感的作物信息获取内容更为丰富,正适应作物农田管理信息化的时代需求。在生产需求调查的基础上,采用田间调查与参数测定、室内化学分析、遥感解译、数据估算与比较等方法,在棉花栽培专家知识、多时相遥感数据、农田背景数据等辅助数据支持下,从土壤质地、棉田保苗数、基于叶面积指数的棉花长势、棉田质量四个方面开展研究,主要研究内容、实施过程及结果如下:1)在研究区选择典型质地土壤,进行室内土壤高光谱测定和对应质地的LANDSAT-5多波段光谱提取,比较不同质地光谱反射特征,理清遥感可识别的土壤质地类型。测定了175个土壤样点的质地,分析LANDSAT-5波段反射率后,确定了不同质地土壤的最佳划分波段和划分阈值,对研究区的土壤质地进行了遥感解译和准确性检验,比较了三种土壤质地制图的特点。研究比较了主要质地土壤在棉花播种前后的土壤水、温变化特征以及棉花保苗数差异,分析了不同质地土壤对棉花萌发和破土出苗的影响,探索了影响土壤质地监测的遥感监测的理化基础,讨论得出研究区土壤质地解译的最佳时段。根据以上研究得出,不同质地类型土壤从近红外到中红外光谱反射率差异显著,可明显区分沙土-沙壤土、轻壤土-中壤土、重壤土-粘土三种质地类型土壤,因此在研究区可以使用LANDSAT-5遥感影像实现较高准确度解译,其中轻壤土-中壤土最易错分,重壤土-粘土分类准确度最高。在不同气候条件下土壤质地与土壤水分有稳定的关系,土壤水分状况的差异决定了土壤质地遥感监测的理化基础。在棉花播种前后的裸露棉田,不同质地土壤存在显著的水、温差异,严重影响棉花出苗,高空间分辨率土壤质地遥感解译结果能够指导棉花播种的空间时序。2)试验于2007年实地调查了13块棉田(630 hm2),获取了棉田生长密度、经纬度以及播种时间、出苗时间所组成的60个样区数据,每样区3个样点;从播种期到盛花期5个时相的遥感影像提取EVI和DEVI,样本等分为建模数据和模型检验数据;采取分播期和不分播期两种方式分别使用EVI和DEVI建立棉田生长密度估算模型,其决定系数经过显著水平检验后,进行模型估算准确性检验。在研究中,基于LANDSAT-5像元尺度,分析了棉田生长密度监测准确性的影响因子,并提出了改进方法,探索了减弱非棉苗背景空间差异的遥感指数,确立了棉田生长密度遥感监测的最佳时相。并将优势模型应用于2007年和2008年研究区域的棉田生长密度监测。根据以上研究得出,出苗时间和土壤等背景因素是影响监测棉田生长密度准确性的主要因素,分播期估算能显著提高棉田保苗数监测的准确性。DEVI可以使棉田留苗密度监测时间提前,现蕾期到开花期是棉田生长密度估算的最佳时段。3)于2006-2007年在新疆148团场18块标准条田开展定位试验,共获得255组不同长势棉花的叶面积指数,进行遥感数据预处理,提取棉花对应时间与样点的LANDSAT-5 NDVI、PVI和EVI植被指数,对两年数据进行统一分析,分别随机选择144组和111组数据进行叶面积指数的估算模型建立和检验,选择最佳估算模型进行叶面积指数的示例反演。为实现具有农学意义和农田管理实际意义的定量分级,总结了该区域栽培专家知识的叶面积指数知识,以2007年7月8日和2008年7月11日为例反演了研究区的叶面积指数并结合专家知识对棉花群体长势进行了定量分级。研究得出在棉花不同生长阶段选择合适的遥感植被指数有利于提高LAI估算准确度.结合棉花栽培专家LAI知识与叶面积指数遥感估算可以实现棉田长势定量分级,可为大尺度宏观管理和小尺度精细管理提供具有农学分类依据的棉花长势信息。4)将棉田质量状况划分为健康棉田、有障碍棉田和疑似有障碍棉田三类,健康棉田界定为棉花各时期生长均正常的区域,各时期棉花长势均处于不正常状态的区域界定为有障碍棉田,各时期棉花长势变异较大的区域界定为疑似有障碍棉田。不同长势的8块棉田作为定位观测区,使用棉花生长盛期多时相的LANDSAT-5各波段反射率数据,探索能较好区分不同长势棉花的波段,建立农田质量分类模型,并由此解译研究区的棉田质量,通过在定位点所测定的叶面积指数、土壤质地、土壤总盐含量验证分类结果,分析该区域棉花质量障碍的主要原因,提出应对不同土壤质量障碍的棉花栽培与耕作措施,指导该区域棉花高产与稳产。研究建立了棉花长势指标动态变化与棉田质量的关系,构建了多时相遥感数据诊断棉田质量的模型与程序,利用多时相遥感数据能够获取棉田质量信息。棉田质量诊断结果与地面调查测定相结合,能较好的分析棉田质量障碍的具体因素。更多还原

【Abstract】 Field managerment is one of the main farming measurements, and its level expresses the production ability comprehensively, the effection of managing field directly influence the yield, qualification, and benefit of crops.Therefore, it is evident that every step of field managerments should be armed with the new achievement of science and technology to meet the society and economy. The agruclutral machine take part in the field management under the industrialization, the resultion was that it liberated man hands and feet, and improve the production efficiency. Under the informatization society, which improvement would field management take? Evidence is mounting that the answer might be the collection, dealing, and decision of field management information automaticly, it would liberate our eyes and ears extremely, it would also increase the depth and breadth of monitoring the crop and field information in face of temporal and spac.Satellite remote sensing technology is the shining pearl and fruit of information science and technology. Some contents became the main application including crop growth, yield, and planting area after the technology of satellite remote sensing was used among the people, it provide the high efficient methods to let the agriculture managing unit gasp the field management information quickly and comprehensively. Following the development of rerearch and practice of remote sensing in agriculture, the works don’t just stay in foretelling and concluding the macroscopical appearance, fortunately, some detail plots were payed attention to gradually, the conception, detailing the monior content and anticipate the process, was being breeded in the crop monitoring with remote sensing, and so the information acquainted by remote sensing would be more rich, which meet the informalization era in the field management.Based on the requirement discovery, in the research, many methods was used including field investigation and measurement, analysis indoor, interpretation with remote sensing, data estimation and comparision, and much assistant data was also used ,such as the expert knowledge of cotton planting, multi-temporal remote sensing data, field background information, and so on. The four content were researched including soil texture, existing plants of cotton, cotton growth condition based on LAI, and cotton field qualification, the detailed content, its analyzing precess, and the main conclusion as follows:1) In the research area, the spectrum with representative texture soil was measured, and the corresponding multi-band spectrum of LANDSAT-5 satellite was also distrilled, these reflections were analyzed to reveal the spectrum difference with different texture soil and make the classable texture type of soil with remote sensing clear. Furthermore, the 175 soil regions in the area were sampled, these textures were tested, and the multi-band reflections of the sample regions were schemed. As a result, the idea band and threshold value, recognizing the soil texture, was ascertained. In ters of the analyzing results above, the soil texture in the research area was interpreted, and the universal ways was used to check the interpretation accuracy. Three results of cartography methods implemented through plotting, GIS, and remote sensing were compared. water content, temperature, existing plants and their changes were analyzed, the soil attribute influencing physical basis of interpreting the soil texture was also analyzed, and the situation of cotton germination was analyzed because of the influence of soil texture. The research would conclude the optimal band interpreting the soil texture.For the soil with different texture, the spectral refection had the significant difference in the NIR and MIR, three type soil, sandy soil - silty loam, light loam– medium loam, and heavy loam– clay, could be devided. The interpretation accuracy of soil texture was high with LANDSAT-5 satellite, the probability is higher of dividing light loam– medium loam into other two types, and heavy loam– clay had high classification accuracy. Because of the strong relation between the soil texture and soil moisture, the difference of soil moisture is basis of recognizing the different texture soil. During about the planting time, the different texture soil influences the change of water, temperature, and decides the number of existing plants to some extent. The interpretation result of soil texuture with high space resolution could direct the cotton planting with suitable space order.2) In 2007, Sixty group sample data, consisting of the existing cotton-seedling density, longitude/latitude, sowing time, emergence time, were obtained through investigating the thirteen fields (630 hm2), and three sample dot data in every sample area were averaged. EVI and DEVI were retrieved up from the images of five times from sowing time to full-flowering. And then sixty group sample data were divided into two equal parts to establish and text models. The linear models were established by data of the middle sowing time and the all three sowing times on the basis of EVI and DEVI, respectively, and the model accuracy was tested by RMSE and REPE. At last, the existing cotton-seedling density at the country scale was retrieved by the best model. In the study, based on the Landsat-5 cell level, having analyzed the factors affecting the estimating accuracy, having explored vegetation indexes to clear up the space information difference of the non-cotton background, having ascertained the optimal time to monitoring the existing cotton-seedling density.Emergence time and soil background were the main factors influencing the accuracy of estimating the existing plants, and the estimation of dividing into the different time stages could obviously improve the accuracy. DEVI could raise the estimation in the ealy stage of cotton growing and advace the estimation time. The optimal time estimating the existing plants was from bud stage to flowering stage.3) The experiment was carried out in 2006-2007 in Xinjiang, and the eighteen cotton fields were validated as the standard observation station, 255 group data of LAI and NDVI, PVI, EVI from LANDSAT-5 were obtained, and 144 and 111 group data were used to establish the estimation models and test the accuracy of models, respectively. In the research, the estimation levels of three vegetation indexes were compared, and the expert knowledge of LAI was concuded. LAI on the about 70th day from the emergence time were retrieved by the optimal model at the regional scale, the cotton population growth was classified by the right of the 18 person-time expert knowledge of the classic LAI. To complish the qualified classification with the signifance of agriculture and field management, the method was researched through combining the remote sensing with agricultural knowledge. the two years in 2007 and 2008 was demonstrated to retrieve the LAI and classify the cotton growth. The index relation is expressed between LAI and NDVI, PVI, and EVI, and the saturation phenomena were evident when vegetation index estimate LAI. It was idea that the suitable vegetation index was chosen to improve the estimation of LAI at the different stages. The qualitative classification would be achieved through combining the expert knowledge and the retrieved result with remote sensing, which could provide the data support with the clear agronomy significance for the growing monitor of cotton.4) The cotton field quality conditions were divided into the three styles of the healthy cotton field, handicapped cotton field and suspected cotton field with handicap. The healthy cotton field was defined as the normal growing in the whole stage, the handicapped cotton field was defined as the unnormal growing in the whole stage, and the others was divided into the suspected area with handicap. Eighty cotton fields were used as the sample regions, multi-temporal remote sensing data was used to explore the optimal band and establish the model for classifying the cotton qualification. And then, the main factors, causing the cotton handicap, were proprosed through analyzing the LAI, soil texture, total salty, and exsiting plants. At last, the measurements improving the soil qualification were raised to direct complish the high and steady yield in the region.Multi-teporal remote sensing data had the prominent action to research the field qualification, the relation was discovered between the dynamic growth condition of cotton and cotton field qualification, and the model and process were estimated to diagnose the cotton qualification with multi-temporal data. Through the diagnosing result of cotton qualification, investigation and measurement in the field, the delailed facors leading to cotton field handicap chould be found out efficiently.更多还原