全球和生物群系尺度土壤有机碳分布控制因子及其分异规律

Distinct controlling factors on the distribution of soil organic carbon stocks at global and biomes scales

土壤有机碳(SOC)是陆地生态系统碳库的核心组成部分,其动态平衡受气候、土壤、植被、地形及人类活动等的影响,但在不同的空间尺度上,这些影响因素的相对重要性和差异还不明确。为阐明不同尺度和不同土层深度土壤有机碳密度(SOCd,kg/m^(3)土壤)的环境影响因子差异,选用全球113571个土壤剖面SOCd测量数据以及38个环境协变量数据,利用数据挖掘方法,分析了全球尺度和生物群系尺度不同土层深度SOCd的控制因子,并量化了空间自相关对相关结果的影响。研究结果表明:仅空间自相关就能解释全球尺度不同土壤深度13%—20%的SOCd空间变异,但是随土壤深度的增加,空间自相关的解释率降低。在剔除空间自相关的影响后,分析结果表明:全球尺度上,气候因素对SOCd空间变异的解释率最高,但只能解释17%—20%,这种解释率在不同土层之间没有显著差异。在生物群系尺度上,除北方森林地区,气候因素能够解释SOCd空间变异的24%—37%;而在北方森林地区,地形是影响SOCd空间变异的重要因素,对SOCd的解释率为21%—43%。这些结果表明,SOCd的控制因子在不同的尺度上明显不同。无论是在全球尺度上,还是生物群系尺度上,如果不考虑空间自相关,地形的影响会被低估,其他环境因素的影响被严重高估。为了准确计算全球与生物群系尺度上各土层SOCd分布的控制因子及其分异情况,空间自相关必须被考虑。

Soil organic carbon(SOC)is a core component of terrestrial carbon pool and its turnover process can be strongly influenced by global environmental changes,making it an important part of the global carbon cycle.Soil carbon sequestration is an effective way of mitigating climate change and has significant implications for maintaining ecosystem function and services.The input and outputof SOC are affected by various factors including climate,topography,soil properties,vegetation,and human activities,resulting in high spatial heterogeneity of SOC storage at different scales.Elucidating environmental factors influencing SOC storage at different scales is vital for increasing SOC sequestration or reducing SOC loesses under climate change.This study used machine learning models to assess environmental factors controlling SOC density(SOCd,kg/m^(3)soil)in different soil layers at the global and biome scales,using SOC measurements from 113,571 soil profiles across the global together with 38 environmental covariates.Especially,we quantified the effect of spatial autocorrelation of SOC measurements on the interpretation of spatial variabiblity of SOC at different spatial scales.The results indicated that spatial autocorrelation alone could explain 13%—20%of the global variance of SOCd in different layers,particularly in the upper layers,and thus spatial autocorrelation had to be considered.Considering spatial autocorrelation can improve the precision of the model(represented by the variability of 10⁃cross⁃validation results)and more accurate(represented by R2)in terms of predicting SOCd.After excluding the influence of spatial autocorrelation,the results revealed that climate was consistently the most important factor in all soil layer depths,and explained 17%—20%of the variance of SOCd across the globe.At the biome level after excluding the influence of spatial autocorrelation,climate explained 24%—37%variance of SOCd depending on biome type,while terrain attributes contributed most(21%—43%)in boreal forest.The results revealed that at the global scale,the impact of terrain could be underestimated owing to spatial autocorrelation,while the influence of other environmental factors could be overestimated.In general,spatial autocorrelation should be considered carefully in order to accurately calculate the control factors and its difference of SOCd distribution in each layer depth at the global and biomes scales.Our results demonstrate distinct controls on SOC distribution at the global and biome scale.This study provides a scientific basis for the rational utilization of soil resources,promoting soil carbon sequestration,and reducing soil carbon emissions.This is essential for developing relevant management and protection policies for SOC,and ensuring the sustainable utilization of soil resources.