基于全极化雷达影像反演垄行结构土壤湿度

发布时间：2018-08-03 18:27

【摘要】：土壤湿度作为气候与环境干旱化的指示因子，是全球变化研究中的重要监测内容之一。土壤湿度不仅影响土壤理化特性与植被生长，更直接影响了粮食的质量和产量，严重制约着农牧业的发展。全球水资源的空间分布与耕地的分布尤为矛盾，40%的耕地面积位于干旱区，直接影响到全球10亿人口的生计。因此，土壤湿度的动态监测对干旱预警以及作物估产具有重要的意义。传统的基于点测量的方法很难对土壤湿度进行宏观的动态监测，通过遥感的可见光-近红外，热红外和微波波段监测土壤湿度已有近50年的历史。微波波段可以穿透云雾和一定深度的土壤，是能够定量化反演土壤湿度的最佳波段。被动微波法的空间尺度过大，不适宜农业领域的研究，本文选择主动微波法。主动微波法反演土壤湿度的基本原理是，雷达的回波强度主要受到土壤介电常数和表面粗糙度的影响，而土壤水分是影响土壤介电常数最主要的因素，以此来建立回波强度和土壤湿度的定量关系。起垄耕种的方式在全球都是普遍采用的，这种周期性地表是随机地表的一种特殊形式。目前常用的土壤湿度反演模型对周期性地表的适用性较差。中国北方的旱田作物（如玉米、大豆、高粱等）绝大多数都是采用起垄的耕种方式。本文研究区位于吉林省公主岭市境内，在松辽平原的东部，以种植玉米、大豆等旱田作物为主。本文分析了周期性地表散射的机理，模拟了雷达参数、地表参数和垄行结构参数对不同极化后向散射系数的影响，结合实测采样数据和全极化RADARSAT-2雷达影像数据，深入开展了基于主动微波法提取周期性地表土壤湿度的研究，并取得了如下创新性研究成果： 1、周期性随机地表与平坦随机地表之间的根本区别在于，当雷达脉冲入射到周期性地表时，通过周期性改变局地入射角的值而改变回波的强度。在不同方位角（即垄向与脉冲入射方向的夹角）状态下，模拟局地入射角在一个周期内的变化，分析后向散射系数对方位角的响应规律。证明在同极化模式下，后向散射系数在方位角为90o（观测方向垂直于垄向）时出现最大值，在方位角为0o或180o（观测方向平行于垄向）时出现最小值，而交叉极化对方位角的响应不敏感。 2、在不同极化方式下，模拟波长、入射角、垄高、垄距、土壤湿度、表面粗糙度对后向散射系数的影响。研究发现，波长越短，土壤湿度和表面粗糙度越大，后向散射系数越大，这个规律不会因为周期性地表而改变。在HH极化模式下，随着土壤湿度的增大，M值（方位角为90o与0o的后向散射系数的差值，用来衡量后向散射系数对方位角的敏感程度）缓慢变大然后趋于饱和；随着粗糙度的增加，M值迅速降低并趋于饱和；随着波长的增加，M值缓慢增加并趋于饱和。VV极化对方位角的敏感性则不受波长、土壤湿度和粗糙度的影响。方位角响应函数的形态在不同入射角和A/T参数（1/2的垄高与垄距的比值）下是不同的，通常可用正弦函数或一元二次函数进行模拟。而且在不同入射角下，M值对A/T的响应规律也不同。无论参数如何变化，不同极化对方位角的敏感性始终满足：HHVVVH。 3、当波长为5.55cm（C波段），入射角为45o时，在不同极化方式和不同土壤参数状态下，模拟150个采样点的后向散射系数。研究发现，垄高（或A/T的值）在一定范围内变化时，不影响后向散射系数的整体分布规律，此时可忽略垄高变化的影响。因此，去除方位角的影响是解决周期性地表土壤湿度反演的关键。根据平均土壤参数计算方位角响应曲线，发现采样点到响应曲线的距离只与土壤湿度和粗糙度有关，以此建立差值距离和比值距离参数，去除了方位角的影响。在实际应用中，地表参数平均值往往未知，研究表明对采样点分布规律进行拟合的曲线可替代方位角响应曲线。 4、基于RADARSAT-2全极化雷达数据，随机提取影像上303个采样点，分析不同极化模式下，后向散射系数对方位角的响应函数。实测的响应规律与理论模拟的规律基本一致。选取50个实测采样点中的34个点进行建模，另外16个点作为检验点。在不同极化模式下，建立特征参数与土壤湿度和粗糙度的定量关系，通过其中两种极化方式的方程进行联立可以消去粗糙度参数，获取土壤湿度的反演模型。然而此模型的精度甚至不及单极化线性模型的精度。本文通过直接模拟土壤湿度和其中两个特征参数的定量关系，粗糙度参数不参与拟合，模型的精度得到了明显提高。 5、研究发现，当雷达入射方向垂直或近似垂直于垄向时，地块易出现异常高亮度值，即相干散射亮斑。前面建立的经验模型对散射亮斑区域的反演明显是不适用的。本文通过雷达实测值和Oh模型模拟的随机地表后向散射系数，建立了周期性地表与平坦随机地表之间的同极化误差函数。误差函数修正后的交叉极化比q一定程度上去除了方位角的影响，但未去除相干散射亮斑的影响；修正后的同极化比p同时去除了方位角和相干散射亮斑的影响。最终选取同极化比和交叉极化后向散射系数与土壤参数的关系式，利用LUT表搜索法求解土壤湿度，反演结果去除了相干散射亮斑的影响。 6、提出了通过修正后Oh模型和经验模型联合反演土壤湿度的方法。Oh模型对相干散射亮斑区域的反演具有明显的优势，但由于Oh模型受到参数取值范围的限制，尤其对湿度较大地块的反演精度较差，而经验模型适用的地表参数范围更宽。16个检验点反演值与实测值误差的平均值是0.0220cm3/cm3，，相关系数为0.9706，均方根误差为0.0258cm3/cm3。4个位于相干散射亮斑区域的检验点与其他12个检验点的反演精度无明显差异。证明了本文提出的经验模型与Oh模型联合反演垄行结构土壤湿度的方法是有效和可靠的。
[Abstract]:Soil moisture, as an indicator of climate and environmental droughts, is one of the important monitoring contents in the study of global change. Soil moisture affects not only the physical and chemical characteristics of soil and the growth of vegetation, but also the quality and yield of grain, which seriously restricts the development of agriculture and animal husbandry. The spatial distribution of total water resources and the distribution of cultivated land are particularly important. Contradiction, 40% of the cultivated land is in the arid area, which directly affects the livelihoods of the 1 billion population in the world. Therefore, the dynamic monitoring of soil moisture is of great significance to the early warning of drought and the crop yield estimation.
The traditional method based on point measurement is difficult to make a macro dynamic monitoring of soil moisture. It has a history of nearly 50 years to monitor soil moisture through the visible light infrared, thermal infrared and microwave bands of remote sensing. The microwave band can penetrate the clouds and a certain depth of soil. It is the best wave band for the quantitative inversion of soil moisture. The spatial scale of the wave method is too large and is not suitable for the research in the field of agriculture. In this paper, the active microwave method is chosen. The basic principle of the active microwave method for the inversion of soil moisture is that the echo intensity of the radar is mainly influenced by the dielectric constant and surface roughness of the soil, and the soil moisture is the main factor affecting the dielectric constant of soil and soil. The quantitative relationship between wave intensity and soil moisture. The mode of ridging and cultivation is generally adopted all over the world. This periodic surface is a special form of random surface. The current commonly used inversion model of soil moisture is less applicable to periodic surface. Most of the dryfield crops (such as corn, soybeans, sorghum, etc.) in northern China are most of them In this paper, the study area is located in the Gongzhuling city of Jilin Province, in the eastern part of the Songliao plain to grow maize and soybean and other dryland crops. This paper analyses the mechanism of periodic surface scattering, and simulates the effects of radar parameters, surface parameters and ridging structure parameters on different polarization backscattering coefficients. The research on the extraction of periodic surface soil moisture based on active microwave method is carried out by measuring sampling data and fully polarizing RADARSAT-2 radar image data, and the following innovative research results have been obtained.
1, the fundamental difference between the periodic random surface and the flat random surface is that when the radar pulse is incident to the periodic surface, the intensity of the echo is changed by periodically changing the value of the local incidence angle. At the different azimuth angles (i.e. the angle between the ridges and the direction of the pulse incident), the change of the local incidence angle in a period is changed. The response law of the back scattering coefficient of the opposite angle is analyzed. It is proved that the maximum value of the backscattering coefficient at the azimuth angle is 90o (the direction of the ridge is perpendicular to the ridge) in the same polarization mode, and the minimum value occurs when the azimuth is 0o or 180o (the direction of the observation is parallel to the ridge direction), and the response of the cross polarization angle is insensitive.
2, under the different polarization mode, the influence of the simulated wavelength, incidence angle, ridge height, ridge distance, soil moisture and surface roughness on the backscattering coefficient. It is found that the shorter the wavelength, the greater the soil moisture and surface roughness, the greater the backscatter coefficient, the law will not change for the periodic surface. Under the HH polarization mode, with soil moisture With the increase of the M value (the difference between the azimuth and the backscattering coefficient of 90o and 0o), the sensitivity of the backscattering coefficient to the opposite angle of the backscattering coefficient becomes larger and then saturated; with the increase of the roughness, the M value decreases rapidly and tends to saturation; as the wavelength increases, the M value increases slowly and tends to the sensitivity of the saturation.VV polarization angle. It is not affected by wavelengths, soil moisture and roughness. The shape of the azimuth response function is different under different incidence angles and A/T parameters (the ratio of ridge height to ridge distance of 1/2), and usually can be simulated with sinusoidal function or one element two functions. And the response of M value to A/T is different at different incidence angles. The sensitivity of each other's angle is always satisfied: HHVVVH.
3, when the wavelength is 5.55cm (C band) and the incident angle is 45o, the backscattering coefficient of the 150 sampling points is simulated under different polarization modes and different soil parameters. It is found that the overall distribution of the backscatter coefficient is not affected by the ridge height (or the value of A/T) in a certain range, and the influence of the ridge height change can be ignored. Therefore, Removing the influence of azimuth angle is the key to solving the soil moisture inversion on the periodic surface. According to the calculation of the azimuth response curve according to the average soil parameters, it is found that the distance from the sampling point to the response curve is only related to the soil moisture and roughness. In this way, the difference distance and the ratio distance parameters are established, and the influence of the azimuth angle is removed. In practical application, the effect of the azimuth angle is removed. The average value of surface parameters is often unknown, and the study shows that the curve fitting the distribution of sampling points can replace the azimuth response curve.
4, based on the RADARSAT-2 fully polarized radar data, the 303 sampling points on the image are randomly extracted and the response function of the backscatter coefficient in different polarization modes is analyzed. The measured response law is basically the same as that of the theoretical simulation. 34 points of the 50 measured sampling points are chosen to be modeled and the other 16 points are used as the test points. In the same polarization mode, the quantitative relationship between characteristic parameters and soil moisture and roughness is established. Through the equation of two polarization modes, the roughness parameters can be eliminated and the inversion model of soil moisture is obtained. However, the precision of the model is not even more accurate than the single polarization linear model. The quantitative relationship between the two characteristic parameters, roughness parameters do not participate in fitting, the accuracy of the model has been significantly improved.
5, it is found that when the direction of the radar is perpendicular to the ridge or perpendicular to the ridge direction, it is easy to have an abnormal high brightness value, that is, the coherent scattering bright spot. The empirical model established in front of the radar is not suitable for the inversion of the scattered spot area. In this paper, the period of the radar measured values and the random surface backscattering coefficients of the random surface of the Oh model are established. The same polarization error function between the ground surface and the flat random surface. The cross polarization ratio modified by the error function Q does not affect the azimuth angle to a certain extent, but does not remove the influence of the coherent scattering bright spot; the modified same polarization ratio p simultaneously removes the influence of the azimuth and the coherent scattering bright spot. Finally, the same polarization ratio and intersection are selected. The relationship between polarization backscattering coefficient and soil parameters is solved by LUT table searching method. The influence of coherent scattering speckles is removed from the inversion results.
6, the method of combining the modified Oh model with the empirical model to inverse the soil moisture is proposed. The.Oh model has obvious advantages for the inversion of the coherent scattering spot area, but because the Oh model is limited by the range of parameter values, especially the inversion accuracy for the larger humidity plots, the surface parameter range of the empirical model is wider.1. The mean value of the error between the 6 test points and the measured value is 0.0220cm3/cm3, the correlation coefficient is 0.9706. The root mean square root error is 0.0258cm3/cm3.4 in the coherent scattering spot area, and there is no obvious difference between the test point and the other 12 test points. It is proved that the empirical model and the Oh model proposed in this paper are combined to inverse the ridge row structure soil. The method of soil moisture is effective and reliable.
【学位授予单位】：吉林大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：S152.7;P631.3

【参考文献】