| Soil moisture is a key variable in the global water,energy and carbon cycle.It is also a basic parameter for crop growth,crop growth monitoring,crop yield estimation and drought monitoring.Synthetic Aperture Radar(SAR)can collect all-day and all-weather data.Because SAR backscattering coefficient is sensitive to soil dielectric constant,and soil moisture is the determinant of soil dielectric constant,SAR is widely used in soil moisture inversion and monitoring.However,radar system parameters,surface roughness and other factors have great interference on soil moisture inversion.Especially in cultivated land,the interference factors are more complex,including vegetation canopy,vegetation water content and other factors.Therefore,in order to improve the precision of soil moisture inversion in cultivated land,it is necessary to eliminate these disturbances effectively.But previous studies focused on empirical,semi-empirical,physical models and other methods.The input parameters of these models are difficult to obtain,and they are only applicable to soil moisture inversion in small areas.Change detection algorithm which use radar and optical data can effectively reduce the interference of surface roughness and vegetation coverage without prior knowledge though.It has advantages in retrieving soil moisture in large cultivated land.A series of sentinel satellite data released by ESA have both radar and optical data.The data cover widely,with the high time resolution(12 days)and high spatial resolution(10m),which provides strong support for change detection algorithm.In this paper,the improved change detection algorithm is used to provide a reference for the research about soil moisture retrieval in cultivated land area.The research results are:(1)The layout of soil moisture monitoring network and the determination of water "true" value and surface roughness of monitoring points.Considering the spatial heterogeneity,vegetation type and soil texture,this paper designs a 40 km*40 km soil moisture automatic monitoring network in Dehui City cultivated land.We measured the effective "true" value of surface soil moisture(0-5 cm).The surface roughness was measured by needle roughness profile plate,and the roughness parameters were calculated,including correlation length,root mean square height,ridge width,ridge height,ridge direction,etc.Among them,ridge width is 62.8-70.4 cm,ridge height is 5.92-12.94 cm,root mean square height is 0.36-1.28 cm,exponential correlation length is 1.23-7.85 cm,ridge Gaussian correlation length is 0.94-11.1 cm,and the average incidence angle of radar system parameter ridge is 37.95-39.79 degrees.At the same time,the azimuth angles which have great influence on radar backscattering signals are measured.The azimuth angles of ridges and platforms are between 16.42 and 76.42 respectively.(2)Variation detection algorithm based on time series SAR data and determination of soil moisture optimal spatial resolution.The study introduced processing of radar data backscattering coefficient,NDVI extraction from optical data and interpolation of NDVI which is due to cloud coverage.The principle and application of several kinds of change detection are described in detail.Algorithm 1 assumes that the roughness of cultivated land changes very little during the radar transit time,the vegetation influence and the change of soil moisture are directly related to the change of backscattering coefficient.The vegetation influence factor is replaced by the NDVI correlation function to remove the disturbance of vegetation cover.Algorithm 2 improves Algorithm 1.A model IEM is used to simulate the minimum backscattering coefficient under the driest condition.According to the difference between radar data and the simulated minimum value and NDVI data,aim to determine the final soil moisture inversion algorithm and the optimal spatial scale.Algorithm 3 is assumed that the change of roughness and vegetation biomass in the adjacent radar data is usually very small,so that the difference of backscattering signals between the two periods depends mainly on the change of soil moisture.Given the boundary constraints,soil moisture can be retrieved.The each algorithm spatial resolution upscales 90 m(9*9 pixel values of backscatter coefficients are averaged).After some attempts,such as scaling up to 30m(3*3 pixels),50m(5*5 pixels),70m(7*7 pixels)and so on,the error between the inversion and the measured soil moisture values is large.Up to 90 m,the accuracy of each inversion method under this scale is the highest.And the resolution of 90 m can reduce the uncertainty caused by the heterogeneity of different maize varieties,which can satisfy the actual agricultural need.(3)Accuracy evaluation,error analysis and uncertainty analysis of error sources about several algorithmsBased on the comparison and analysis of ground measured data and inversion values about the three methods,the accuracy,advantages and disadvantages of the 3 inversion methods are evaluated by using RMSE,MRE,MAE,IA,ubRMSE and R statistical indicators.Given the applicability and application scope of the three methods,the sources of error are analyzed and discussed.According to all the indexes,the accuracy of method 2 is better than that of method 1,and method 3 is the lowest.Considering the simplicity,high efficiency,and some data have lost because of the instrument failure,thus the study choose the data with more moisture truth value to verify and analyze as far as possible.Several representative data are selected,the bare soil period(20170429),the vegetation growth period(20170628),the vegetation flourishing period(20170815)and the vegetation decline period(20170920).Among them,method 2 is more reliable in bare soil period and flourishing period.In the growth period,method 1 is more suitable.In the declining period,both methods 1 and 2 are reasonable.The above shows that the change detection algorithm based on SAR data has certain rationality and practicability.It can satisfy the needs of agriculture in a certain period of time. |
PDF Full Text Request