| Phosphorus and potassium are essential elements for plant growth and the biological cycle in the agroecosystem.It is the basis of improving soil quality and fertilization efficiency to clarify the spatiotemporal variability of available phosphorus and available potassium in topsoil and their relationship with grain yield.It is also of great significance to study the relationship between soil nutrient content and soil environmental capacity,soil environmental pollution.This work was conducted in Xuchang city,Henan province.Three representative digital soil mapping（DSM）methods,namely Ordinary Kriging（OK）,random forest（RF）,random forest combined with Kriging（RFK）,were employed to predict the spatial distribution of available Phosphorus and potassium in topsoil of cropland.The three models’performance and prediction accuracy were evaluated,and the relative importance of covariates for model prediction was analyzed to determine the main factors affecting the content and spatial variability of available phosphorus and available potassium in topsoil of cultivated land in the study area.According to the precision comparison results,the optimal model was determined for spatial prediction and digital mapping of available phosphorus and available potassium in topsoil of cultivated land in 2018,2008,and 1980.Based on the optimal model outputs at different three time points in the study area since the 1980s,the temporal and spatial variation characteristics of available phosphorus and available potassium intopsoil of cultivated land in Xuchang City in recent 40 years were analyzed.Ultimately,the Boruta feature selection algorithm was adopted to quantitatively evaluate the importance of all factors prepared in this study for winter wheat yield,summer maize yield,and total grain yield（The sum of yield of winter wheat and summer maize）and analyze the relationship between soil available phosphorus,available potassium content and grain yield.The main conclusions are as follows:（1）The spatial prediction accuracy is relatively low compared with other soil attributes because the effective nutrient content in topsoil is greatly affected by fertilization,sampling time,and other factors.In comparison,the three models’performance is quite different,among which the prediction accuracy of the OK model is lower,and that of the RFK model is higher.The latter has an explanation ratio of about 24%for available phosphorus and 26%for available potassium.The prediction accuracy fluctuates slightly in different periods.The r relative importance of covariates showed that the fertilization factor,climate factor,and altitude of terrain factor were the main factors affecting the content and spatial variation of available phosphorus and available potassium,especially fertilization rates.The spatial distribution characteristics of available nutrients in the surface layer of cultivated land in different periods were quite different.IIn 2018 and 2008,the high-value area of available phosphorus in topsoil was mainly concentrated in the distribution area of Shajiang Black Soil in the southeast and the cinnamon soil distribution area of Yuzhou City in the northwest and Xiangcheng County in the southwest.The high-value area in 1980 was mainly distributed in Xiangcheng County in the southwest.In 2018 and 2008,the high-value area of available potassium in the topsoil of cultivated land was mainly distributed in Yanling County in the southeast,while in1980,it was distributed in Yanling County and Xiangcheng County in the southwest.According to the sample data analysis,the average content of available phosphorus increased by 6.93mg·kg-1 from 1980 to 2018,and the average content of available potassium increased by 20.10mg·kg-1 from 1980 to 2018.（2）Based on the best model output’s spatial prediction results,according to the comparison of the national soil nutrient classification standards,the level of available phosphorus content in the surface layer of cultivated land showed an upward trend in the past 40 years.In the 1980s,it was chiefly distributed in four and five levels,and in 2018,it was mainly distributed in three levels.The potassium content in the cultivated land surface layer was mainly in grade 3 and grade 4 in the 1980s and grade 3 in 2018.In terms of absolute increment,the content of available phosphorus in the surface layer of cultivated land in the study area increased by 8.60mg·kg-1 from1980s to 2018,and the areas with 0-5mg·kg-1,5-10mg·kg-1 and above 10mg·kg-1 of available phosphorus increment accounted for 26.13%,30.45%and35.2%respectively.From the 1980 to 2018,the average increment of available potassium was29.27mg·kg-1.The proportion of cultivated land with 0-25mg·kg-1,25-50mg·kg-1 and above 50mg·kg-1 increment of available potassium in surface soil was 21.39%,41.98%and 20.33%respectively.From 1980s to 2018,the contents of available phosphorus and available potassium in the surface soil of the main cultivated soil types and different agricultural use types in the study area were significantly increased,and the differences of available phosphorus and available potassium in the surface soil of different soil types and different use types were gradually reduced.（3）The contents of available phosphorus and available potassium in the surface layer of cultivated land and their spatiotemporal changes are affected by multiple factors,including climate,parent material,topography,soil-forming environment,and soil genetic process,cultivated land use mode and cultivation model,fertilization and cultivated land management measures.Among them,fertilization,straw returning,and stubble management were the main driving factors for the dynamic changes of available phosphorus and available potassium in the surface layer of cultivated land on a small time scale.The relative importance analysis results indicated that the contents of available phosphorus and available potassium were important covariables in the spatial prediction of wheat,summer maize,and total grain yield,suggesting that they had significant grain yield effects.Conversely,this relationship was not significant in the direct correlation analysis,possibly because their importance was masked by covariates such as seasonal fertilizer application. |
PDF Full Text Request