The Remote Sensing Inversion Study Of Soil Salinity Of The Drip Irrigation Cotton Fields Under The Membrane In Arid Areas

Facing the serious security threat of saline soil on the continued farming of land and food production, it is particularly important to invert soil salinization quickly and it has become the core and key issues of agricultural sustainable development of Xinjiang to timely monitor and prevent the soil salinization. There have been a lot of research on the inversion of soil salinity at home and abroad, but so far they have not found a more comprehensive approach to invert soil salinity. This paper tries to use Landsat8OLI remote sensing image as the data source. It quantitatively inverts soil salinity of Shihezi Reclamation District based on the data of soil salinity of the72samples measured in the field, then explores the inversion methods and techniques of the soil salinity in the study area of the Drip Irrigation cotton fields under the membrane, which uses Landsat8OLI remote sensing image as data source. The study includes:(1)The option of the methods of soil salinity inversion:it studies the correlation coefficients of saline soil spectral reflectance and soil salinity measured in cotton field, then analyzes the diagnosis index and the correlation coefficient of each band spectral reflectance and soil salinity, and finds out that the soil reflectance of Band4, Bands and Band6in Landsat8OLI remote sensing images is more appropriate to invert soil salinity. Because the inversion of soil salinity by using remote sensing image is susceptible to soil moisture, soil brightness and soil radiation level and other factors, the paper through hat or cap with tassels transform, the decomposition of mixed pixels and LBV transform to remote sensing image selects remote sensing characteristic index which represents soil salinity including soil brightness, soil moisture, soil cover and soil radiation level and other indices. To Sensitive band and remote sensing characteristic index it uses multivariate linear model approach to establish functional relationship respectively with soil salinity. Comparing to simulation results shows that the accuracy of remote sensing characteristic index to invert soil salinity is higher than the sensitive band. Therefore, we choose remote sensing characteristic index (SBI, SMI, SSCI and GRLI) as the primary method of inverting soil salinity of the Drip Irrigation cotton fields under the membrane. (2)The BP inversion based on the remote sensing characteristic index:the relationship between the remote sensing characteristic index and soil salinity is complex and changeable, so the inversion effect of general statistical model is not very satisfactory. BP neural network has superior linear and nonlinear fitting ability, which is very helpful to solve the complex relationship between the remote sensing characteristic index and soil salinity. The paper establishes inversion model of BP neural network by using soil brightness, soil moisture, soil cover and soil radiation level and other remote sensing characteristic index as input variables, soil salinity in cotton fields as output variables, established BP neural network inversion model. Through the precision validation it finds that the square of correlation coefficient of the predicted and actual values is0.917. We can know from the figure that of most validation sample the predicted soil salinity is very close to the real measured soil salinity, and relative to the measured values, the predicted values of the inversion for most sample points are small. REE and RMSM, the precision validation index of inversion model is3.07and0.34respectively, and the model accuracy is66%. The paper uses the trained neuron to invert the soil salinity of the Drip Irrigation cotton fields under the membrane.(3) According to the divided rank standard of the degree of soil salinization the paper divides the rank by the degree of soil salinity of the cotton field, then analyzes the spatial pattern of soil salinity in the study area, to find out that of the cover area of the entire cotton field, the non-salinization cotton area accounts14.02%,mild salinization cotton area accounts for21.38%, the moderate salinization cotton area accounts for21.27%, and severe salinization cotton area accounts for29.99%, saline accounts for13.3%in cotton area, only35.4%of the cotton field is suitable for farming agricultural production, other64.6%of the cotton plowland needs to continue to make improvements. The salinization impact of the cotton fields in Shawan is greatest, much higher than Manas and Shihezi in extent. The regions of high rank of salinization distribute in Shawan County, mostly locate in the alluvial plains, low-lying area, and reservoir areas.The novelty of this paper are:contrast to the common methods of remote sensing inversion of soil salinity of the predecessors, it proposes to take the remote sensing characteristic index which reflects soil salinity as input arguments, and to build high-precision remote sensing inversion model of soil salinity with the BP neural network platform, which provides scientific evidence for high-yield of cotton and the prevention and cure of soil secondary salinization.