| Ecological vulnerability is increasing due to human activities and climate change,and there is an urgent need to identify ecological degradation in time for early intervention.Soil microbial biomass carbon(MBC)is highly sensitive to ecological changes and is a recognized ecological early warning indicator.However,the dominant drivers of MBC are debated in existing studies,and the MBC inversion models are not yet clear.The Yellow River source area has various types of grassland,unique geographical environment,and is highly sensitive to climate change.It is an ideal sample site for studying the driving factors and inversion models of MBC.The study of MBC in the Yellow River source area will verify the feasibility of MBC as an early ecological warning indicator,and it is also important for predicting the response patterns of soil carbon pools at regional scale and the maintenance and restoration of ecological functions of degraded grasslands.In this study,the sample sites were deployed in the Yellow River source area using the conditional Latin hypercube sampling(c LHS)method,and 358 MBC measured data were systematically collected.The measured data were correlated with the corresponding 25 environmental variables to explore the potential dominant drivers of MBC changes;the feature selection is carried out through stepwise regression,ridge regression,and Lasso regression,and then combined with the four models of partial least squares regression,random forest,support vector machine,and back propagation neural network,after inputting the measured data and environmental variables into the models,the optimal model adapted to the area was screened according to the fitting effect of the test set;finally,the optimal model was applied to the inversion and digital mapping of MBC in the Yellow River source area to analyze the spatial and temporal trends of MBC from 2010 to 2021.Based on this,the spatial and temporal changes of MBC in different grassland types were discussed.The final main conclusions are as follows:(1)The values of MBC from the measured data ranged from 76.74 mg/kg to1663.65 mg/kg,with a wide range of values;the median value was 365.46 mg/kg and the mean value was 441.69 mg/kg,with the values concentrated around 400 mg/kg.(2)Among the soil physicochemical factors,MBC showed the strongest correlation with soil organic carbon and p H;among the vegetation factors,MBC showed significant positive correlation with all vegetation factors;among the meteorological factors,MBC showed significant correlation with precipitation and light duration;among the topographic factors,longitude showed significant positive correlation with MBC.The environmental variables that showed significant correlations with soil microbial carbon and correlation coefficients above 0.50 were:soil organic carbon(SOC),soil p H(p H),normalized difference vegetation index(NDVI),the fraction of absorbed photosynthetically active radiation(FPAR),precipitation(Prep),light duration(ILL),and longitude(Lon).(3)In the Yellow River source area,the best fit(R~2=0.64)and the smallest prediction error(RMSE=174.66)were obtained after feature selection based on stepwise regression followed by parameter training combined with a random forest model,which was the best model among all models for inversion.(4)The spatial distribution of MBC in the Yellow River source area is characterized by decreasing from southeast to northwest,and the temporal changes from 2010 to 2021 are characterized by decreasing and then increasing and exceeding the initial decline starting point,and the spatial and temporal changes of MBC in each major county and each grassland type in the region are significantly different.Based on the findings of this study,the following policy recommendations are proposed:1)apply MBC for ecological monitoring and timely early warning;2)treat it in a graded manner and take intervention measures according to local conditions;3)continue to insist on ecological priority and strict policy implementation. |
PDF Full Text Request