人为与环境因子对农田土壤有机质影响的比较研究--以典型黑土区和水稻土区为例

摘要

目前关于土壤有机质（SOM）影响因子的研究多涉及单一或少量因素，缺少环境因子与农田管理相结合的详细分析。为综合分析环境和人为因素对农田SOM的影响，利用2011年采集于典型黑土样区的281个样点和水稻土样区的193个样点，结合相应农田管理调查数据，采用双样本Kolmogorov-Smirnov检验、Kendall τ相关分析、随机森林模型进行对比研究。结果表明，通过综合环境与农田管理措施数据，随机森林方法可以较好地预测农田SOM含量，对其变异的解释度达到80%以上。人为与环境因素对SOM变异的解释能力在两样区间存在差异。SOM含量水平在黑土区受环境因子影响为主，在水稻土区受人为因子主导。相对重要性分析显示黑土区水热条件、黏粒含量影响显著，水稻土区田间管理作用明显。%Abstract[Objective]Soil organic matter(SOM)retains and recycles nutrients,improves soil structure and sustains soil microbes. Its content is not only an indicator of soil fertility,but also a direct reflection of soil organic carbon stocks. Keeping SOM content at a relatively high level can mitigate soil degradation,ensure food security and alleviate greenhouse gas emissions. Therefore it is important to dig out what are the factors that can exert influence on SOM content and what are the leading ones. Previous studies in this field used to focus on one or a few factors. Climate,terrain,and land-use data were frequently used in those researches. However,these factors were not adequate enough to reflect impacts of agricultural production on SOM.[Method]To comprehensively analyze influences of environmental and anthropogenic factors on SOM content in cropland in typical black soil region and paddy soil region,two grain-producing areas,281 and 193 soil samples were collected,respectively,in the two regions in the autumn of 2011. Field management data,such as cropping system,yield of grains,fertilizer amount and history of residue incorporation of the sampling plots were collected through consulting related farmers. Two-sample Kolmogorov-Smirnov tests were employed to compare the soil samples from the two regions in SOM content and in impact of the affecting factors on SOM. Kendallτ correlation analysis was conducted to screen out factors that were significantly related to SOM content for comparative analysis with random forest models established based on environment factors alone,anthropogenic factors alone or both,and the influence strength of each factor on SOM content was evaluated.[Result]Results show that all the factors expect for elevation and fertilizer amount,vary sharply in impact on SOM content between the two regions and so does SOM content. Among the environmental factors,mean annual precipitation(MAP)is the one SOM is closely related to in both regions,and mean annual temperature(MAT)and clay content in the Black Soil Region and parent rock and terrain in the Paddy Soil Region are the ones SOM is closely related to,while among the anthropogenic factors,fertilization,tillage and residue incorporation are in both regions and irrigation is not. Random forests models using both environment factors and field management in these two regions perform well in fitting,explaining over 80% of the variances of SOM content in croplands of the two regions. Environment and anthropogenic factors vary between the two regions in explanation of the variation of SOM content. Environment factors explain 84% in the Black Soil Region and 52% in the Paddy Soil Region of the variation of SOM, and field management practices do 62% and 72%,respectively. After ruling out the variance which could be explained by both environment factors and field management,environmental factors explain 4.7 times as much as anthropogenic factors the variation of SOM in the Black Soil Region,and 2/7 times as much in the Paddy Soil Region. Therefore SOM content is mainly affected by environmental factors in the Black Soil Region and by anthropogenic factors in the Paddy Soil Region. Relative importance analysis shows that key impact factors are MAT,MAP and clay content in the Black Soil Region and MAP,elevation and fertilizer in the Paddy Soil Region.[Conclusion]The findings of this study also demonstrate that although SOM content in topsoil could change rapidly as affected by human activities,it is still feasible to predict SOM content quite accurately by using Random forest models with the key lying in the integration of environment factors and field management. Therefore it is of important significance to collect field management data year by year. These data can be used not only to improve performance of the model,but also to analyze trend of spatio-temporal variation of SOM content. The key factors identified in this study affecting SOM may be used to guide field management in study area and soil sampling design for future studies.