区域土壤容重转换函数构建与预测结果评价——以河南省封丘县为例

Fitting and Precision Evaluation of Regional PTFs for Soil Bulk Density——A Case Study from Fengqiu County,Henan Province

以河南省封丘县为研究区,应用现有的代表性土壤转换函数(PTFs)对40个样点、169个样品的土壤容重实施预测并对其较差的预测精度进行了分析与评价,指出对于不同地区、不同类型、不同剖面垂直深度和不同发生层次的土壤,其容重PTFs的预测变量及相应参数各不相同。鉴于此,本研究基于获得的研究区基本土壤属性信息,应用回归分析手段分别构建了表层及心土层容重PTFs。预测结果显示,以土壤砂粒和有机质含量为预测变量的普通回归PTFs可以解释90%的土壤容重变异,且具有很高的预测精度;与心土层相比,表层土壤普通回归PTF的解释能力、预测精度均明显较低,表明表层土壤容重受其他随机因素尤其是人为活动因素的影响较强。进一步研究显示,通过逐步回归手段增加预测变量可以有效提高PTFs的解释能力和预测结果的合理性,但不一定能有效提高预测精度,PTFs预测精度主要取决于预测变量的遴选范围与数据质量。

Taking Fengqiu County, Henan Province as the study area, three representative pedo-transfer functions （PTFs） fitted in the different regions were employed to predicted bulk density of 169 soil samples collected at 40 sites, the results with rather low precision demonstrated that the PTFs should be one of the kind of functions characterized by regions, soil types and profile horizons. In view of the mentioned above, PTFs were fitted respectively for topsoil and subsoil on the basis of basic soil information obtained using the methods of regression analysis. The results indicated that subsoil bulk density could be accurately predicted by a classic regression PTF, in which sand and OM contents were best predictors and could explain above 90% of bulk density variance. Compared to that for subsoil, PTFs for topsoil had an evidently low predictive capability, perhaps because topsoil was always suffered much stronger influences of some stochastic factors, especially human activities. Further studies showed that a stepwise multiple regression （SMR） procedure could obviously increase the explained variance of the PTFs by adding predictors, but could not necessarily improve the predictive precision, which mainly depended on the numbers of candidate predictors and their data quality.