Prediction Of Soil PH And Organic Matter Based On Soil Color And Environmental Auxiliary Variables

Soil color is one of the most important soil characteristics,and reflects the reflection spectrum of soil in visible light band.With the development of science and technology,soil color could be accurately measured by CM600d spectrophotometer to predict soil pH value and organic matter content.The paper was based on 187 soil profiles collected and measured in the field.The spatial distribution patterns of CIE L*a*b*(L*is brightness,a*is redness,b*is yellowness)and RGB(red,green and blue)were revealed accurately by variance analysis,Kriging interpolation and true color realization system in the core area of Chengdu Plain.Using correlation analysis and multiple regression analysis,the prediction models were established by using environmental auxiliary variables as input and soil color and environmental auxiliary variables as input to predict soil pH value and organic matter content of each layer.The results were as follows:(1)Soil color L*,a*,b*,R and G showed a significant increasing trend in profile,and there was a significant difference among different soil layers(P<0.05),while the B value did not change significantly with the increase of soil depth(P>0.05).The values of L*,a*,b*,R and G of soil color in the study area showed a significant increasing trend in profile,and there were significant differences among different layers,while the values of B did not change significantly with the increase of soil depth.0-20 cm,20-40 cm,40-60 cm,60-100 cm,the mean values of a*were-0.12,0.13,0.42 and 0.59.Relative to surface soil(0-20 cm),the average value of soil color L*increased by 3.17%,5.97%and 5.93%,and in 20-40,40-60 and60-100 cm soil layers,the average value of soil color b*increased by 19.27%,29.30%and31.14%,the average value of soil color R increased by 4.69%,842%and 8.42%,and the average value of soil color G increased by 3.12%,5,94%and 5.81%.(2)The L*value of soil color was mainly affected by random factors in 0-20 cm soil layer,by both structural and random factors in 20-60 cm soil layer,and in,and by structural factors in 60-100 cm soil layer.The a*value of soil color was mainly affected by structural factors in 0-60 cm soil layers,and by both structural factors and random factors in 60-100 cm soil layers.The value of soil color b*had strong spatial autocorrelation in the surface layer(0-20 cm),while the other soil layers had moderate spatial autocorrelation.Spatial variability of soil color R and B values was affected by both structural and random factors.The nugget effect of soil color G value in 0-60 cm soil layers ranged from 25%to 75%,which belonged to moderate spatial autocorrelation.However,it was less than 25%and had strong autocorrelation in 60-100 cm soil layers.The spatial distribution of soil color L,a*,b*and RGB had obvious changes,and they all had good spatial structure.(3)Soil true color spatial distribution system was developed by Visual Studio 2010 and ArcGS Engine 10.1 software.The system could realize the distribution of soil true color spatial pattern.In different soil layers,the spatial distribution pattern of true color was similar in the core area of Chengdu Plain.The main color was grey in the middle area,and the color was dark,while the main color was brown and yellow in the northeast and southwest,and the color was light.The spatial distribution pattern of soil true color was consistent with that of parent material in the study area,which indicated that the spatial distribution pattern of soil color could reveal the distribution characteristics of parent material to a certain extent.(4)There were significant relationship among soil pH and organic matter content,soil color parameters,environmental auxiliary variables.Soil pH value was significantly correlated with soil color a*value,parent material type,crop rotation pattern and water system distance in 0-100 cm(P<0.05).Soil organic matter content was significantly correlated with soil color L*,a*,b*,R and G values(P<0.05),and positively correlated with soil type and parent material type as environmental auxiliary variables in 0-100cm(P<0.01).(5)Both models could be used to predict soil pH and organic matter content,and the introduction of soil color were significantly improve the prediction accuracy of the model.In different soil layers,the determinant coefficients R~2 of soil color prediction model increased by 108.00%,13.55%,14.55%and 10.46%,the average absolute error MAE,root mean square error RMSE and average relative error MRE decreased by 9.84%,15.19%and 12.79%,9.52%,8.93%and 10.15%;9.30%,1.96%and 3.98%,9.09%,13.79%and 12.65%.Compared with the predicted values,the determinant coefficient R~2 increased by 44.03%,34.16%,99.12%and 83.78%.The determinant coefficients R2 of the prediction model with soil color and environmental auxiliary variables were 13.37%,47.58%,58.49%and 27.04%higher than those with only environmental auxiliary variables;MAE,RMSE and MRE of the validated samples decreased by 7.93%,7.06%and 9.72%;11.4%,10.31%and 11.02%;0.40%,5.32%and 20.61%;6.03%,4.36%and 9.61%;and the actual samples were validated.Compared with the predicted value,the determinant coefficient R2 increased by 29.92%,21.84%,0.26%and 43.66%.