Identification Of Influencing Factors And Mechanisms Of Soil Inorganic Carbon Dynamic Changes Based On Different Study Scales

Soil inorganic carbon（SIC）is an important component of soil carbon pool,and its dynamic changes are of great significance to global climate change.Human activities,soil properties,and climatic factors are important factors affecting dynamic change of SIC content.However,the response results of SIC content to human activities in different studies are quite different,owing to the complex interaction of the above factors.As a result,the general effects and internal mechanisms of human activities on SIC content are still unclear.At the same time,there are few studies on quantitatively exploring the importance of the influencing factors of SIC content.In addition,the coastal estuary area,as an intersection of terrestrial ecosystems and marine ecosystems,is not only the area with a high intensity of material and energy exchange,but also the important SIC sink in terrestrial ecosystems.Currently,many studies have been conducted in relation to SIC in this area,but the variation rules of SIC content under different vegetation coverage and soil depth conditions and quantitative identification of key influencing factors are rarely reported.Therefore,it is necessary to establish an in-depth understanding of the relationship between the SIC dynamic changes and influencing factors,the relative importance of influencing factors,and the underlying mechanisms at the global and regional scales.This can provide important scientific theoretical basis for enhancing soil carbon sink function and evaluating the impact of SIC dynamic changes on climate change.To address the above problems,the main content and results are listed as follows:（1）Based on a data set compiled from the articles published globally,the effects of human activities and their interactions with soil properties and climate factors on SIC content at the global scale were quantitatively explored using the meta-analysis,random forest（RF）model,and structural equation model.Furthermore,the relative importance and internal mechanisms of influencing factors were revealed.The main results for this content are as follows:（1）SIC content significantly increased by 6.55%and 9.25%for cultivation and land use change,respectively,compared with that of the control,regardless of any influencing factors.Whereas,SIC content was insignificantly affected by the fertilization and irrigation.The response of SIC content to human activities can be significantly affected by soil physicochemical properties and climatic factors,but varied according to the specific type of human activity.For the land use change,the conversion of wetland to cultivated land caused the largest positive response（70.1%）of SIC content,followed by the conversion of sandy land to wood land（35.2%）,while the largest negative response of SIC content was found to the conversion of cropland to wood land（–17.0%）.（2）The soil properties were found to significantly affect SIC content.SIC content in alkaline,low bulk density,or deep soil groups were significantly higher than those in other corresponding soil groups,indicating that SIC is easy to accumulate in soils with high p H alkalinity,low bulk density,or/and deep layer.Moreover,the land use type showed a significant effect on the SIC content,and the SIC content of the grassland was significantly higher than that of other land use types.The SIC content in the mean annual temperature<0℃group was the highest among all the mean annual temperature groups,and the SIC content in the mean annual precipitation>800 mm group was the lowest among all the mean annual precipitation groups.（3）The RF model was constructed using the relevant data obtained from the literature.The fitting coefficient between the observed and predicted values of SIC content in the test set could reach 0.86,meaning a good model performance.The relative importance of each factor examined was ordered as follows:soil p H（18.2%）>soil type（16.4%）>mean annual precipitation（16.3%）>soil bulk density（15.2%）>soil depth（13.4%）>mean annual temperature（13.0%）>land use type（7.52%）.Combining the results of RF model univariate partial dependence plot and structural equation model,it was clear seen that SIC content had a strong positive correlation with soil p H,but a strong negative correlation with soil bulk density.（2）Taking the Jiaozhou Bay Yang Estuary area as the study object,the content and distribution characteristics of SIC under different vegetation coverage and soil depth conditions at the regional scale were deeply explored through field sampling and laboratory analytical testing experiments.The influencing factors of SIC content were selected based on the conclusions obtained from the literature data and the special geographical location of coastal wetlands.Further,the key influencing factors and internal mechanisms were quantitatively examined by combining use of correlation analysis,RF model and structural equation model.And the research content of the previous part was verified regionally.The main results for this content are as follows:（1）Overall,SIC content in different sample plots were ordered as sand plot(3.49g kg^–1)>Spartina alterniflora plot(2.37 g kg^–1)>Suaeda salsa plot(2.27 g kg^–1);and SIC content in different soil layers were ordered as 0–15 cm(2.80 g kg^–1)>15–30 cm(2.54 g kg^–1)>30–45 cm(2.36 g kg^–1).The orders of soil dissolved inorganic carbon content in different sample plots and soil layers were Suaeda salsa plot>Spartina alterniflora plot>sand plot and 0–15 cm>30–45 cm>15–30 cm,respectively.（2）Using field sampling experimental data,three machine learning models,including RF,support vector machine regression,and artificial neural networks,were constructed,and the results demonstrated that the performance of the RF model was better than other models.The fitting coefficient between the measured and predicted SIC content was 0.73 for RF model.Furthermore,the results of sensitivity analysis of RF model indicated that soil p H was the most important factor for the prediction accuracy of RF model.（3）The relative importance of each factor examined by RF model was ordered as follows soil p H（38.1%）>soil salinity（18.1%）>soil organic carbon（14.4%）>soil bulk density（8.89%）>Mg²⁺（7.13%）>soil water content（6.77%）>Ca²⁺（6.54%）.