Study On Remote Sensing Inversion Of Cultivated Soil Organic Matter In Songnen Plain

Cultivated land is a key resource for human living and development.Cultivated land can provide basic material for human by farming.As an important guarantee of grain output,cultivated land resource is of great significance of national food security and social stability.In recent years,the excessive plundering of land by human beings has resulted in the obvious decline in cultivated land quality of some areas.Due to the different quality conditions of cultivated land in different regions,it is necessary to use and protect cultivated land according to local conditions.Soil organic matter(SOM)is an important index to measure soil fertility and evaluate the quality of cultivated land.SOM spatial distribution is not only conducive to analyzing the output capacity of cultivated land,but also can provide accurate and quantitative data support for the evaluation of cultivated land quality.Currently,the SOM spatial distribution acquisition method is usually based on a large number of field sample data,although this can provide the accurate SOM content of sampling points,it is time-consuming and labor-intensive,and difficult to get the spatial distribution information of SOM organic matter on a regional scale and in a large range.In other words,quantitative monitoring is still difficult.The development of remote sensing technology makes it possible to obtain the SOM spatial distribution by satellite remote sensing.In this paper,the cultivated land in the north of the Songnen Plain was taken as the research area,and 147 samples of 4 typical soil types in different regions were collected in May 2014.Meanwhile,considering the influence of soil moisture content,crop residues cover,soil surface state and other interfering factors of soil reflectance spectra,multi-temporal MODIS remote sensing images of bare soil in the study area in 2002,2009 and 2014 with differenct crop residues covers were obtained.On the basis of spectral data analysis in laboratory,remote sensing spectral indices based on single-date and multi-date MODIS images were established as model input,multi-temporal information was introduced,variables were screened by stepwise multiple linear regression method,and SOM remote sensing inversion model at regional scale was established.At the same time,the laboratory data are used to discuss and analyze different model inputs to explain the influence mechanism of interference factors of remote sensing inversion models.The results showed that:(1)the MODIS remote sensing model could be used for SOM prediction of different soil types in the Songnen Plain at regional scale.Both indoor spectral data and MODIS remote sensing image data could significantly improve the accuracy and stability of the model by building spectral indices compared with the reflectance of bands.The accuracy and stability of the inversion model based on two-date images were generally higher than single-date image model.(2)the difference between soil moisture and crop residues cover had a great influence on the SOM inversion accuracy at the regional scale.Crop residues cover,snow melting time interval,precipitation and other factors determined the input of SOM remote sensing inversion model and the selection of the optimal image data.(3)in the period with low crop residues cover and low spatial difference of soil moisture content,SOM inversion model based on the two-stage image had the highest accuracy.In 2002,the multidate images of 113 d and 121 d had high accuracy.At this time,the snow and in the study area had been melting for a long time and the rainfall had not yet arrived.The prediction models on the spectral index(difference spectral index and angle spectral index)of the first four short bands as the main input had the highest accuracy.The models based on two-date images could improve the stability and accuracy of the inversion model.(4)when crop residues cover was high,SOM inversion model with single-date spectral index after tillage as input has high accuracy.Crop residues buried after ploughing and exposed the soil,but the difference of soil moisture content still existed.The optimal SOM inversion model input included the band which was sensitive to soil moisture and the band which was sensitive to organic matter.The ratio of spectral index R61 was the main input variable.The ratio of spectral index R61 was significantly correlated with SOM,but it had low correlation with soil moisture content.So,R61 could reduce the effect of soil moisture for the inversion model.(5)SOM content in north cultivated land of the Songnen Plain decreased from northeast to southwest.Due to the higher latitude in the northeast,the temperature was lower than that in the southwest and the low temperature lasted a longer time.The harsh cold environment all year long had fixed a large amount of humus in the soil and increased the content of organic matter in the soil.Soil types with high content of organic matter such as black soil and chernozem were widely distributed.The southwest was an alluvial plain with many pastoral areas.Therefore,under the combined action of river impact,overgrazing behavior and high groundwater level,soil erosion and land desertification were serious,the SOM content level in this area was low.The results can provide accurate and quantitative SOM spatial distribution data for evaluation of cultivated land quality in regional scale and serve for making decision of cultivated land utilization and protection.