High Spectral Prediction Model for Soil moisture in Songnen Plain
The object of this paper was to raise the prediction accuracy of high spectrum reflectance in order to eliminate the effects of soil types.The continuum removal method was used to treat reflective spectral curves and extract the characteristics parameter of spectral adsorption.Reflectance was treated with various mathematical transformation models.Differences of reflectance spectral characteristics among difference soil types and soil moisture levels were analyzed in this paper.A high spectrum predictive model in different soil types was established.The result accuracy was evaluated by determination coefficient (R2) and RMSE.The larger the R2 of the model was,the more stable of the model was.The smaller the RMSE of the model was,the more stable of the model was.Result showed that:(1) It was difficult to extract effective spectral characteristics parameters for predicting soil moisture.(2) The accuracy of model established by spectral characteristics parameters was high in different soil types except for aeolian soil.(3) Unary linear regression based on spectral characteristics parameters had high accuracy and less input.The results would provide scientific bases for the prediction of soil moisture in different soil types.