Prediction Of Soil Texture Under Different Water Conditions Based On Vis-NIR Spectroscopy

It is time and energy consuming to test soil texture information based on the soil physical analysis method.It is difficult to meet the modern agriculture and the environment management of soil texture information needs.With the development of the visible near infrared spectral technology for fast access provides a new way of soil texture.Vis-NIR spectroscopy applied in field soil texture prediction,the prediction accuracy will be affected by soil moisture content.VIS-NIR reflection spectra of soil samples were obtained in the laboratory under nine moisture conditions,and soil texture prediction models for each moisture conditions were built using PLSR,and soil texture under other eight moisture conditions as validation set were predicted by air-dried soil cross prediction model.And quantitatively studied the effect of soil moisture on VIS-NIR spectrum to predict soil texture.On this basis we then discuss two methods to improve soil texture prediction accuracy under different soil moisture condition.One way is based on the theory of spectral calibration by External Parameters Orthogonal(EPO)for spectral preprocessing,so that we can eliminate the influence of soil moisture and improve forecast accuracy.The other way is based on the idea of classification,introducing perpendicular drought index as soil moisture indicator(PDI),and the soil sampled were re-grouped according to PDI,and then the modeling and validation set were established.The main results are as follows:(1)Analysis models for each moisture condition performance well under the same moisture conditions.The accuracy decreased dramatically with the increase of difference on soil moisture conditions between the built data set and validation data.(2)After correction by the EPO method,the spectral bands of different water conditions in the same soil sample were significantly smaller.Compared to the original spectrum with uncorrected,clay content results can be used in practical.Sand content prediction results also perform better.However,this method needs the spectral data of dry soil and its standard moisture contentto establish a calibration set.(3)The PDI index can be used to replace the soil moisture index to carry on the grouping model with unknown soil moisture state and large difference range.For the application,soil moisture conditions can obtained using satellite remote sensing and,separate models were build based on the soil moisture conditions to predict soil texture,which is an applicable method.