Construction Of Soil Bulk Density Prediction Model For Main Grasslands In Xinjiang

Soil bulk density is an important basic soil attribute,and the data lack in many areas is serious.The estimation and precise regulation of soil carbon and nitrogen can not be separated from the index of unit weight.Based on the data of sample points,the relationship model of soil bulk density and organic carbon in Xinjiang and six districts is established.The prediction model of soil bulk density was established by BP neural network modeling method.By analyzing the data of unit weight and organic carbon obtained from field sampling,the paper has mastered the change range of soil bulk density and organic carbon in Xinjiang grassland,the change range of soil bulk density and organic carbon of different grassland types and different soil types The difference,the distribution characteristics of the whole and local soil bulk density of Xinjiang grassland were analyzed.The following main conclusions ware obtained:(1)The variation range,mean value and coefficient of variation of soil bulk density and organic carbon in Xinjiang grassland were 0.24?1.99 g·cm³and 0.01%?24.51%,1.24 g·cm³and 1.86%,23.15%and 129.96%respectively.The classical statistical results show that the soil bulk density of 7 soil layers is of moderate variation,and the variation degree of 50-70cm and 70-100cm is the smallest,and the soil bulk density increases with the increase of depth.(2)The results showed that the soil bulk density of 11 grassland types was significantly different in soil layers of 0-5cm and 5-10cm,and the bulk density of alpine meadow and swamp soil was significantly lower than that of other grassland types(P<0.05).In 50-70 cm and 70-100cm soil layers,the soil bulk density of each grass type was not significant.Similarly,the vertical difference of soil bulk density of different soil types of grassland is also significant.The soil bulk density of meadow soil,felt soil,swamp soil,alpine grassland soil,black calcium soil,felted soil,gray calcium soil,cold calcium soil and salt soil increases with the increase of depth,and the relationship can be described by power function.(3)The existing 10 transfer function models are applied to the study area,and the applicability of these models is evaluated.The results show that models 1 and 5 are most suitable for estimating soil bulk density of main grassland in Xinjiang.At the same time,compared with the volume weight prediction model established by the field sampling data,the prediction error of the model is smaller,but the model decision coefficient R²is not significantly improved.(4)The optimal prediction model of soil bulk density in Xinjiang(SBD=1.3861e^-0.0704soc,R²=0.35,P<0.0001)and the optimal prediction models of soil bulk density in six areas were established:Altai Mountain(y=1.5251e^-0.0803x),Tacheng basin(y=-0.15lnx+1.3839),Ili Valley(y=1.2463e^-0.0531x),Northern Tianshan Slope(y=1.4107e^-0.0513x),Tianshan south slope(y=1.3438e^-0.0778x)Kunlun Mountain(y=-0.1537lnx+1.2263).The results can provide important parameters for the study of soil carbon reserves of Xinjiang grassland.(5)The prediction model of soil bulk density is established by BP neural network,and the soil organic carbon content,longitude,latitude,altitude and slope are selected as input parameters.The model is tested.The data shows that the correlation between the predicted value and the measured value of the verification set data is0.59,which shows that the parameters selected for establishing BP neural network model are reasonable The model can be used to predict the soil bulk density.