【作者】 夏天；

【导师】 周勇；

【作者基本信息】 华中师范大学 ， 人文地理学， 2011， 博士

【摘要】 近年来,随着全球气候不断变暖,粮食产量受到影响。而人类的需求量不断增长,粮食缺口越来越大,粮食安全问题成为世界各国面临的主要问题之一。现代农业是以实现低能耗、高效率、高效益为目标的新型农业,它充分的利用科学技术和现代化设备,采用先进的管理方法进行农业种植。高光谱遥感技术为推进现代化农业进程提供了重要的技术支持,高光谱遥感具有实时监测能力强、监测范围广、精度高等优点,在农作物的长势监测和大面积估产、农业资源的调查与动态监测、农业灾害的预报与灾后评估等方面都发挥着重要的作用。由于遥感监测常常受到天气、影像分辨率、时间分辨率等诸多外界因素的制约,并且遥感信息只能对种植作物的表面进行观测,作物的内部结构变化、生理状况等都无法从遥感影像中直接获得,而且这些作物所反映出来的信息与作物本身、以及气候、土壤等环境因素息息相关。为了弥补这一不足,科技工作者采用了作物生长模拟模型,这种模型正是基于这些因素的影响对作物进行生长模拟,它综合了计算机技术、作物生理学、作物生态学、农业气象学、土壤学、农艺学、系统学等多学科的知识,对作物的生长、发育、产量多种状况进行动态模拟。卫星遥感技术能够宏观的监测植被状况,作物生长模型能够微观地模拟作物生长态势,两者优势互补,能够较好模拟出作物的产量。再结合自然、社会、经济等影响因素,可以对区域粮食产量进行评估,可以为区域粮食安全预警决策提供科学的依据。具体研究内容及结果如下：1)本文首先对江汉平原湖北省潜江市后湖样区的郑麦9023、皖麦369和黄淮海平原济南市长清样区的临麦2号、稳千1号和泰农18号五个品种冬小麦在不同生育期小麦冠层光谱进行分析,同时结合与光谱数据实时采集的冬小麦叶面积指数(LAI)、冬小麦叶片叶绿素值(SPAD)、冬小麦冠层有效光合辐射比率(FPAR)等冬小麦农学参量进行分析,研究发现不同区域的冬小麦在同一生育期内叶面积指数没有明显差异,冬小麦叶面积指数与高光谱720nm-760nm附近反射率相关性达到0.65。本研究选取近红外平台最大值875nm、红光波谷670nm、绿光550nm和蓝光450nm波段的光谱反射率来计算RVI、DVI、NDVI、GRVI、EVI和SAVI这6种植被指数,通过与冬小麦叶面积指数进行相关性分析发现,其相关性在0.766～0.842之间,均具有较好的相关性。将植被指数与冬小麦叶面积指数拟合建立关系模型,并进行了预测精度检验,结果显示植被指数RVI和GRVI在LAI小于3.0时拟合度较好,但随着冬小麦的覆盖度增加拟合精度下降。RVI和GRVI这两种植被指数都较敏感,特别是LAI大于3.0后,对植被的变化更敏感。DVI、NDVI、EVI和SAVI这四种植被指数模拟精度在整个生育期变化不大。利用BP神经网络选取多个典型波段反射率对冬小麦叶面积指数进行估算,能够大大地提高反演精度,拟合度高达0.959,总均方根差RMSE为0.208。但是BP神经网络如果另外增加了学习样本,“训练”好的网络就需要从头开始训练,它们对以前的权值和阀值是没有记忆的,神经网络相对传统的回归模型而言,实用性有待进一步研究。2)研究区域的冬小麦叶片叶绿素值(后简称SPAD)在整个生育期内均呈现抛物线变化,但是两样区间冬小麦叶片SPAD值存在明显差异,主要表现在SPAD峰值出现的生育期不同,后湖样区冬小麦叶片SPAD峰值出现在抽穗期,长清样区冬小麦叶片SPAD峰值出现在拔节期。这种区别在光谱反射率曲线中也得到反映,后湖样区冬小麦光谱的蓝、红光波段在拔节期出现了降低的现象,这说明叶绿素含量在升高；而长清样区的绿光波段在拔节期至抽穗期升高,说明该时期叶片的叶绿素含量降低。两研究区域冬小麦叶绿素含量的这些变化都能从光谱变化中得到印证。两样区冬小麦SPAD值与冬小麦冠层光谱反射率分别在674nm(后湖样区)和686nm(长清样区)附近达到波谷负相关最大值,760nm-1350nm近红外波段反射率与SPAD值呈显著的正相关。由于两样区冬小麦叶片SPAD存在较大的差异,故本研究分别将两样区冬小麦叶片SPAD值与植被指数GRVI、RVI、NDVI、DVI进行相关性分析,相关系数在0.686～0.901之间,其中后湖样区的SPAD值与NDVI相关系数最高,达到0.901；长清区冬小麦SPAD值与NDVI相关系数达到0.873。通过建立植被指数与SPAD值的回归模型,经模型精度检验,NDVI预测模型显示其精度相对较高,后湖样区和长清样区冬小麦SPAD反演模型均能较好的对该地区冬小麦SPAD值进行估算。3)在对冬小麦冠层光合有效辐射比率(后简称FPAR)研究中发现,两样区的冬小麦FPAR值的差异不明显,冬小麦FPAR值与冠层光谱反射率绿光波段510nm处的负相关性最大,相关系数为-0.74,与760nm附近光谱反射率相关性为0.65。通过对冬小麦FPAR与RVI、DVI、NDVI、GRVI、EVI和SAVI这6种植被指数进行相关性分析,相关系数在0.737～0.837之间,说明这6种植被指数与冬小麦FPAR相关性均较好,为了找出最佳预测FPAR模型,并比较几种模型的预测精度,将6种植被指数与冬小麦FPAR进行拟合,拟合结果显示6种植被指数中NDVI较适合进行FPAR的估算,R2为0.802。其次SAVI的估算精度较高,R2为0.734。通过精度检验结果表明,NDVI适用两样区的冬小麦FPAR估算,估算结果较好。4)为了更好地分析研究济南市长清区冬小麦产量情况,本文将微观研究与宏观研究相结合,最终选取WOFOST (World Food Studies)作物生长模型进行模拟长清区冬小麦生长,该模型能够较好的对长清区冬小麦产量进行估算。通过中国“资源环境卫星A、B星”遥感影像与本文研究的冬小麦LAI反演模型相结合,对研究区域范围内的冬小麦LAI值进行反演。由于WOFOST作物生长模型最初是为欧洲作物进行的设计,本文对该模型进行了“本地化”的优化,通过收集长清区逐日气象资料、并实测土壤数据和作物数据,在文献检索和专家咨询的基础上,对WOFOST模型进行本地化参数设置,同时利用FSEOPT优化程序,结合遥感影像反演冬小麦LAI值结果对模拟模型参数进行调整。通过“本地化”后的WOFOST作物生长模型能够较好地预测出长清区的冬小麦产量变化趋势。为评价长清区冬小麦供应状况提供了科学的数据支持。5)通过收集济南市区及长清区的社会、经济和自然条件等资料,综合分析长清区的农业、人口、消费、经济等情况,发现长清区近年来人口数量呈整体下降趋势。随着经济的发展,城市居民消费结构发生了改变,由原来以“食品为主衣着为辅”的消费结构变为如今的多元化消费结构,同时城市居民的饮食结构也在发生了变化,粮食摄入量减少,乳制品、水果等其它食品摄入量增加。农村居民的生活水平在不断提高,虽然粮食的摄入量变化不是很明显,但油脂类、肉禽类的摄入量呈上升趋势。长清区近几年冬小麦的播种面积没有大的变化,基本维持在22000公顷左右。冬小麦的单产变化较为明显,冬小麦产量受到天气、环境、管理等外界因素的影响极大。本文选取长清区及济南市区的粮食产量、人口、收入及消费分别作为自然、社会及经济因素的评价指标。通过粮食安全综合评价模型,对长清区冬小麦产量变化率进行评价,评价结果显示：2007年和2009年长清区粮食产量变化率有较大波动,冬小麦总产量出现明显降低的情况,其它年份均较上一年基本持平或有所提高。长清区冬小麦粮食安全指数分为两种情景研究,一种为对内供应型,长清区冬小麦仅供给该地区；另外一种为内外兼顾型,长清区冬小麦需要供给该地区和济南市区。第一种情况,由于长清区属于粮食生产地区,区内冬小麦产量足以满足该地区居民的需求,安全指数为安全状态。第二种情况,除2007年长清区冬小麦产量出现急剧下滑的情况,该地区冬小麦产量不能满足区内居民及济南市区居民需求,安全指数为不安全状态,其它年份冬小麦产量均能满足两地区居民需求。此评价方法对未来预测区域粮食安全状况具有一定的参考价值。更多还原

【Abstract】 In recent years, food production has affected by been global warming. With the increasing demand in grain, food security has become one of the major problems the world faced. Modern agriculture is a new-type agriculture aimed at low energy consumption, high efficiency and effectiveness through making the best use of modern science and technology and equipment, also adopting advanced management methods. Hyperspectral remote sensing is an important technology in promoting the modernization of agriculture. With a high real-time monitoring capability and a high precision to monitor a wide range of farm, it played an important role in monitoring the crops growth condition and in yield assessment of a large area, investigation and dynamic monitoring of agricultural resources, agricultural disaster forecast and post-disaster assessment. As remote sensing is often affected by the weather, image resolution, time resolution, and many other external factors. Remote sensing information can only make observations of crops on the surface of the ground, so crop’s internal structure changes, physical condition, etc. Cannot be obtained directly from the remote sensing images. Also, that information is related to the crops, environmental, climatic and pedantic factors. The crops-growth simulation model which based on those factors influence to the growth of wheat. It combining the computer technology, crop physiology, crop ecology, pedology, meteorology, systematic and so on. Then making dynamic simulation for crops growth, development and yields.Satellite remote-sensing techniques can monitor the growth situation of vegetation macroscopic. Crop-growth simulation models can simulate plant growth by macroscopic. Being united, they can into play bring their respective advantages and can better simulated crop yields. By combining the crop yields with natural, social, economic and other factors, the assessment of the food security problems within the scope of area can provide scientific data for the food security warning of region in decision-making.1)Firstly, this paper analysed five kinds of wheat’s canopy spectra, including Zhengmai 9023, Wanmai 369, Linmai2,wenqian land Tainong 18,in two plots (Houhu District in Qianjiang City of Hubei province in central China and Changqing District in Jinan City in northern China) in their different Growth stages. Then, combined the spectral data in real time with collection of wheat leaf area index (LAI), Leaf chlorophyll values (SPAD), canopy photo synthetically active radiation ratio (FPAR) and other wheat agronomic parameters. Study found that different varieties of wheat in the same growth stage did not differ significantly in leaf area index. Correlation between wheat leaf area index and hyperspectral reflectance near 720nm～760nm reached 0.65. This study selected the maximum platform near infrared 875nm, red trough 670nm,550nm green and 450nm blue band to calculate the six vegetation indexes of the RVI, DVI, NDVI, GRVI, EVI and SAVI, the correlation analysis of the wheat leaf Area index showed correlation between the 0.766～0.842, all have good correlation. If modeling vegetation indexes and the wheat of leaf area index to test models’prediction precision, results showed that when LAI is less than 3.0, RVI and GRVI are fit to each other.But with the increase in coverage and decrease in degree of fitting the two types of vegetation indexes (RVI and GRVI) are very sensitive, especially when LAI is more than 3.0, vegetation index is much sensitive. These four kinds of vegetation indexes’simulated precision just changed a little during the whole growth period. Selecting some typical band to estimate leaf area index by BP neural network can greatly improve the retrieval accuracy degree of fitting up to 0.959, with a total mean root square RMSE of 0.208. However, if the BP neural network adds more learning samples, the trained network will need to start training from scratch. Since the BP neural network has no memory of the previous values and the thresholds. Compared to traditional regression model’s the neural network’utility, needs further study.2) The values of wheat leaf chlorophyll (the following are referred to as SPAD) in two study areas during the whole growth period showed a parabola change, but the two wheat leaf SPAD values are significantly different, mainly in different growth period when SPAD peak values appear.Houhu plot’s wheat SPAD peak value appears at the heading stage, the latter at the jointing stage, this difference has also been reflected in the spectral reflectance curve. In the Houhu plot, the blue, red band of wheat spectra reduced at the heading stage. This shows that chlorophyll content increased. However, in the Changqing plot green band increased at the jointing stage, indicating that the leaf chlorophyll content decreased. The chlorophyll content change of wheat in the two plots can be confirmed from the spectral changes. Those two wheat SPAD values and canopy spectral reflectance at 674nm (Houhu plot) and 686nm (1 Changqing plot) reached through negative correlation maximum, 760nm～1350nm near-infrared reflectance and SPAD values were significantly positive. Since there is a big difference between the two plots’wheat SPAD values. So the study will do correlation analysis of the two plots’wheat SPAD values and GRVI. RVI, NDVI respectively, the correlation coefficient between 0.686～0.901.The wheat SPAD value in Houhu plot and NDVI showed the highest correlation coefficient, reaching to 0.901; In Changqing District the correlation coefficient of the wheat SPAD value and NDVI was as high as 0.873. Setting up regression models with Vegetation index and SPAD values, then testing the model that showed a relatively high precision, which means wheat SPAD inversion models of the two plots both can estimate the SPAD values suitably.3) From the study of Wheat canopy’s Fraction of photo-synthetically Active Radiation absorbed by the canopy in the PAR ratio, (the following are referred to as FPAR). We found that there is no significant difference between the wheat FPAR values in the two plots, wheat FPAR value and canopy spectral reflectance at 510nm reached the maximum negative correlation of-0.74, and spectral reflectance near 760nm correlation is 0.65. By analyzing the six vegetation indices correlation (RVI, DVI, NDVI, GRVI, EVI and SAVI) we found the correlation coefficients are between 0.737～0.837, indicating the 6 types of vegetation indices were corrdated with wheat FPAR well. To find the best FPAR prediction model and compare precision of different models, we fitted the six types of vegetation indexes into wheat FPAR values, the results showed that the NDVI is quite suitable for the estimation of with R2= 0.8021 and SAVI has a higher estimation accuracy, with R2= 0.734.From the last results of the test, NDVI is suited the estimation of the wheat FPAR value in the two plots.4) In order to make a better analysis of wheat yields. In Changqing District of Jinan city, this paper integrated the microcosmic with macroscopic research and ultimately selected the WOFOST (World Food Studies) crop growth model to simulate the growth of wheat in Changqing District. This model can estimate wheat production of Changqing District better. Combining the remote sensing image gained by "China Resources and Environment Satellite A, B" with the wheat LAI inversion model studied in this paper, we inverted the wheat LAI value within the study area. Since the crop growth model WOFOST was designed for European crops initially. This article gave the model a "localization" optimization. In Changqing District, by collecting daily weather data, measuring soil data and crop data, combining the literature with and consulting experts’advices, we made use of the remote sensing image and FSEOPT to invert the wheat LAI values to optimize the results. The crop growth model WOFOST optimized localizable can better predict wheat output trends in Changqing District. This model provided reliable scientific data for wheat output security situational in Changqing District. 5) Through collecting the social, economic and natural and other information of Changqing District of Jinan city and analyzing its agriculture, population, consumption, economy, comprehensively, we found that the population and natural population growth rate is on an overall downward trend in recent years. With the economic development, on one hand, the consumption structures of urban residents are constantly changing, from food-based supplement to the diversified consumption structure. Near urban residents’diet structure changed to they take less grains, but more dairy products, fruits and other food.On the other hand, the living standards of rural residents is also raising. Although there is no obvious changes in food intake, fats and oils, meat and poultry intake increased. The wheat acreage in Changqing District remained at around 22,000 hectares in recent years. However, obvious changes happened in wheat yields which influenced by weather, environment, personnel management and other external seriously. This paper selects food production, population and income and consumption, of Changqing District as the natural, social and economic indices evaluation respectively. With the food safety assessment model, we evaluated the change rate of wheat yield in Changqing District. The results showed the rate of change of grain production decreased in 2007 and 2009 significantly. The rates in other years, compared to the previous year, keep the same basically or increased. The wheat food security index in Changqing District indicates wheat production is in a safe condition.The research on the wheat food security index in Changqing District included two conditions:one is internal supply.That means the wheat yield only supply the Changqing District.The other is internal and external supply. That means the wheat yield supply the Changqing District and Jinan City.In the first condition the wheat yiels can need the citizen’s demand.For the Changqing District is acrops production,the wheat yields can’t need the citizen’s demand except a sharp decrease in 2007.So the food security index in a unsafe condition. The quantity can not only meet the demand of residents in Changqing District, but also will meet the residents of other areas.This method is useful for predicting the regional security condition in the future.更多还原