基于不同模型不同指纹因子的东北黑土区小流域泥沙来源分析

基于指纹识别技术计算了东北黑土区典型小流域不同侵蚀产沙源地的泥沙贡献比。通过分析农地、林地、草地表层土以及侵蚀沟样品中的33种物质,使用非参数检验和多元判别分析筛选出包括P、Ce、Ga、Rb和137Cs组成的最优复合指纹因子,并将放射性核素137Cs和210Pbex作为第2组指纹因子,将最优复合指纹因子中的单个因子分别作为单因子,作为第3组指纹因子,分别利用多元混合线性模型、Bayesian模型和单因子解析解等泥沙来源指纹分析方法计算了表层土和侵蚀沟的相对泥沙贡献比。结果表明:基于不同模型不同指纹因子的泥沙来源贡献比结果虽不尽相同,但无重大差别。利用多元混合线性模型计算时,由放射性元素137Cs和210Pbex作为指纹因子计算的泥沙来源(表层土47.5%,侵蚀沟52.5%),与最优复合指纹因子计算的泥沙来源(表层土44.6%,侵蚀沟55.4%)基本一致;利用Bayesian模型计算时,由放射性元素137Cs和210Pbex作为指纹因子计算的表层土和侵蚀沟的泥沙贡献比约各占1/2,而利用最优复合指纹因子计算得到的泥沙贡献比中,表层土(58.8%)多于侵蚀沟(41.2%);以复合指纹因子中单个因子为指纹因子计算解析解,P、Ga、Ce、137Cs 4个因子的判别能力较强,能有效判别泥沙物源区;为保证泥沙贡献比计算结果的精确性,有必要确定各模型的计算精度,并挖掘具体的影响因素,调整参数或算法,为模型改进提供依据。研究发现,面积占比不足1%的侵蚀沟贡献了流域近1/2的泥沙,表明侵蚀沟发育引起的土壤流失不容小觑,应加强对该区侵蚀沟道的治理。

Sediment Sources in a Small Watershed Located in the Black Soil Region of Northeast China Based on Different Models and Various Fingerprints

Fingerprint technique is effective in relative sediment contribution estimation. As recently developed methods, Bayesian model and analytical solutions for single factor received more and more attentions. However, compared with most often used multivariate mixing models, their performance and stability in calculating the proportion of sediment sources kept unknow. In order to make clear the contribution of topsoil and subsoil in a typical small watershed in the black soil region of Northeast China, where distributed vast farmlands experiencing serious soil erosion and a number of gullies developing very fast, these three methods were introduced to provide estimation. In this studied watershed with 27.60 km² area, totally 69 samples from the sediment sources area and 30 samples from the sedimentation area were collected. Sediment sources covered 45 topsoil samples in farmland, woodland and grassland, and 24 subsoil samples from gullies. Based on the analysis for 33 properties in these samples, the optimal composite fingerprints including P, Ce, Ga, Rb and ¹³⁷Cs were screened by non-parametric test and multiple discriminant analysis. Taken these 5 fingerprints as group I, radionuclides ¹³⁷Cs and ²¹⁰Pb_ex as group II, Walling-Collins model, the representative of multivariate mixing models, and Bayesian model were used to calculate the sediment contribution for the two potential sources, respectively. Taken each property in the optimal composite fingerprints as group III, analytical solutions for single factor were also used to provide such estimation. The results showed that the contribution ratios of sediment sources based on different fingerprint group and various methods were similar, for example, the ratio provided by group II (topsoil 47.5% and subsoil 52.5%) kept consistent with that calculated by group I (topsoil 44.6% and subsoil 55.4%), while the multivariate mixing model was adopted, the contribution estimated by group II (about half to half) was slightly different with that based on group I (topsoil 58.8% and subsoil 41.2%) in Bayesian model. However, the results provided by these two models were relative great-up to 14.2%, while taken group I as tracers. In group III, P, Ce, Ga, and ¹³⁷Cs could differentiate the sources while applying for the single factor in analytical solutions, the results that half contribution from topsoil and half from gully, were not completely same, but very closed. The differences in sediment contribution might be caused by different principles of models. There were still some spaces to make improvement for different models in order to obtain more reliable results. Attention should also be paid to the gully, as it has caused severe soil erosion and contributed about 50% sediments with less than 1% area ratio in the whole watershed. To reduce sediment derived from the watershed, it is necessary to strengthen prevention and treatment for the gullies to the future land management and soil erosion controls.