Multi-spectral Estimation And Inversion Of Soil Salinity By Unmanned Aerial Vehicle And Satellite In Yellow River Delta Crop Planting Area

The Yellow River Delta is an important agricultural development zone in Huang-Huai-Hai region,with advantageous geographical position,abundant resources and great development potential.The coastal salinized soil in the Yellow River Delta is one of the important reserve land resources.As an important agricultural production base,the salinized soil in the Yellow River Delta destroys the agricultural ecology and has an impact on the coordinated and sustainable development of economy and ecology.Dynamically monitoring the characteristics of soil salinization in the typical crop planting area of the Yellow River Delta,understanding the different response of crop condition of soil salinization,and grasping the spatial distribution of soil salinity in the coastal crop planting areas,which have very important significance for long-term regional economic stability and sustainable development.Remote sensing monitoring has become an important means to acquire the ground ecological environment with its real-time,high efficiency and wide range.UAV airborne remote sensing has become an important technical method to acquire spectral information due to its characteristics of strong mobility,low time cost,high spatial resolution and controllable range.Therefore,the paper took Dongying city of Shandong province Kenli District as the research area,and the main research object was soil salinity in crop planting area,Kenli District field data was conducted,uniform distribution of 77 sample points in winter wheat planting area of the study area was collected,at the same time set up the experimental area of winter wheat,corn,cotton,uniform layout sample points respectively 99,110,90,to get the sample soil salt content and winter wheat plant height,SPAD and other data,using statistical methods to analyze experimental area of crops growing relationship with the change of space distribution of soil salinity,to study the response of different crops of soil salinity.Collected UAV multi-spectral images in each test area,and at the same time acquired Kenli District's Sentinel-2A satellite multi-spectral images of the same period in 2019,extracted spectral information after image preprocessing,and constructed spectral parameters and screen sensitive spectra,statistical analysis,The construction and verification of the machine learning soil salt estimation model,and the spectral correction of different scales,finally obtained the best soil salt inversion model for different crop cultivation areas in the study area,and inverted and analyzed the characteristics of soil salt distribution.It provided a scientific basis for obtaining the distribution of soil salinity levels in the coastal crop planting area and guiding the agricultural production in the research area.The main conclusions are as follows:(1)The winter wheat growth factor G was obtained by plant height and SPAD index 1:1in the winter`wheat test area,and the optimal vegetation index TDVI was obtained by correlation analysis.With the increase of soil salt content,crop spectral vegetation index showed a decreasing trend,indicating a negative correlation between soil salt content and crop growth.(2)Based on UAV multi-spectral images of winter wheat,corn and cotton experimental areas,sensitive spectral bands and spectral indexs were obtained and screened out through correlation,collinearity and significance.The correlation between the four sensitive spectral bands and the spectral indices NDVI,RVI,SI and soil salt content were greater than 0.6 and0.7,respectively,there was no multicollinearity among the parameters,and the significance of spectral parameters on soil salt content was less than 0.05.The spectral bands b _G,b _REG,b _NIRand the spectral index SAVI and GRVI were selected as sensitive spectral characteristic parameters in the corn experimental area.The correlation between the spectral bands and soil salt content was greater than 0.6 and 0.7,respectively,there was no multi-collinearity and the significance was good,while the significance of the spectral band model was higher than the indexs.The sensitive spectral bands in the cotton test area were b _R,b _REG and b _NIR,and the spectral indexs were SAVI and GRVI,with correlations greater than 0.65,no multicollinearity and significance of 0.00.Five methods including stepwise regres ion,principal component linear regression,partial least square method,BP neural network and support vector machine were used to establish a variety of soil salinity estimation models with sensitive spectral parameters as independent variables.The optimal inversion model for soil salt content of winter wheat area was Y=-9.4774×NDVI+0.4794×RVI+3.0747×SI+5.0604,and the modeling R² was 0.734,RMSE was 0.954,the verification R² was 0.784,RMSE was 0.769,R²was verified to be0.513 after ascending scale correction.The optimal model for soil salt content of corn test area and planting area was SVM support vector machine band model,and the modeling R²was 0.672,RMSE was 0.347,and the verification R²was 0.729,RMSE was 0.459.The optimal model for soil salt content of cotton test area and planting area was SVM support vector machine band model,and the modeling R²was 0.759,RMSE was 0.448,and the verification R² was 0.778,RMSE was 0.398.The obtained model has high precision and stability.(3)The ratio of average method was used to get the correction factor of each growing satellite bands image,In addition,in the winter wheat planting area,77 sample points of winter wheat growing area were used to verify four revised estimate models with higher accuracy,The model verification accuracy of index model established by partial least squares method is higher than 0.5,which has the highest precision,good stability and applicability.(4)In the study area,growing overall soil salinization degree from high to low in turn is:cotton,winter wheat,corn,show the best of cotton salt resistance and its planting area is given priority to with mild salinization soil and moderate soil salinization,area of 44.35%and41.67%respectively,the concentration distribution in the study area in central farming area,with 13.9%than the planar distribution of severe salinization of soil;The soil in the winter wheat planting area was mainly slightly salinized,accounting for 73.09%of the total area,which was concentrated in the relatively high terrain in the southwest of the study area and the northeast area affected by the fresh water of the Yellow River.The moderate salinized soil was scattered in the wheat area,while the severe salinized soil and salinized soil were distributed along the Yellow River tidal area in the central and western regions.There are80.92%mild salinized soil in maize planting area,which is widely distributed in corn planting area.