Monitoring Model For In Situ Spectroscopy Of Soil Organic Matter In Salinized Desert Soils In Southern Xinjiang,China

Soil organic matter（SOM）as a key indicator for evaluating soil fertility and an important component of terrestrial carbon pool.Rapid monitoring of organic matter content in desert soil can provide a scientific basis for the rational development and utilization of reserve arable land resources.In-situ field spectroscopy such as visible and near-infrared（Vis–NIR）spectroscopy has proved as an ideal tool to achieve rapid and efficient detection of SOM,but soil moisture interference is one of the main challenges for measurement of SOM using in-situ spectroscopy.In this study,135 surface（0–20 cm）soil samples were collected from the Aksu region of Xinjiang,Northwestern China,and their corresponding in situ spectra,laboratory spectra,organic matter content,and soil moisture content（SMC）were determined.The performance of the SOM prediction models developed by three pretreatments and three special variable selection algorithms combined with three modeling approaches was analyzed and compared.The soil samples were also wetted to a saturated moisture content state after manual removal of salts.The spectral data of 11 different moisture levels during natural air-drying of the wet soil samples were measured to analyze and compare the performance of three algorithms,EPO,DS and PDS,in removing the effect of moisture and improving the accuracy of SOM prediction.Finally,the correction ability of the optimal moisture correction algorithm for in situ spectra at different moisture intervals was investigated.The results show that:PDS has a relatively poor ability to remove the effect of moisture among the three moisture correction algorithms.When the moisture content of the soil exceeds 48%,EPO,DS and PDS cannot effectively remove the disturbance of moisture.The EPO algorithm was the most effective in removing moisture effects when the SMC is 25–48%.Using EPO,the predicted R² and RPD of SOM increased by more than 0.19 and 0.36,respectively.The DS algorithm was the best method to remove the moisture effect on soil Vis–NIR spectra when the SMC was 6–25%,and the R² and RPD predictions of SOM after removing the effect of moisture by DS increased by more than 0.09 and 0.32,respectively.However,when the moisture content is less than6%,the influence of soil moisture on the prediction accuracy of SOM spectra is negligible and SOM can be predicted with high accuracy using in situ spectra without introduce of moisture correction algorithms.The standard normal transform（SNV）is the most effective pretreatment method for predicting organic matter content based on Vis-NIR in situ spectroscopy.The performance of the organic matter prediction models built by the three feature variable selection algorithms that screened bands was better than that of the full-band spectra.And the model built by the particle swarm algorithm（PSO）has the highest accuracy,with R² and RPD improved by more than 0.34 and 0.16,respectively.The prediction model constructed with a convolutional neural network（CNN）provided the best prediction of soil organic matter in both the full band and the filtered feature band.The SNV-PSO-CNN is the optimal combined model for in situ spectral measurement of soil organic matter with R²of 0.71,RPD of 1.88 and RMSE of 1.67 g kg^-1,which can achieve quantitative in situ spectral inversion of desert soil organic matter.The in situ spectra of soils with different moisture contents were corrected using the EPO and DS algorithms to effectively remove the effect of moisture on in situ spectral measurements.In particular,the EPO algorithm was used to remove the effect of moisture for soil samples with high moisture content（25%<SMC<48%）,while the DS algorithm was used to remove moisture interference in the in situ spectra when the soil moisture content was between 6%and 25%.The organic matter prediction models R²,RMSE and RPD based on Vis-NIR in situ corrected spectra could reach 0.76,1.60 g kg^-1 and 2.02.Therefore,it is necessary to use the moisture correction algorithm for spectral correction before estimating soil organic matter using in situ spectra.