基于复合指纹法解析西南岩溶小流域泥沙来源  

Analysis of sediment sources in karst small watersheds in Southwest China using composite fingerprinting technique

作　　者：戴涛 蒋勇军 田兴 刘芳 韩莎 罗淑娥 DAI Tao;JIANG Yong-jun;TIAN Xing;LIU Fang;HAN Sha;LUO Shu-e(Chongqing Key Laboratory of Karst Environment,School of Geographic Sciences,Southwest University,Chongqing 400715,China)

机构地区：[1]西南大学地理科学学院,岩溶环境重庆市重点实验室,重庆400715

出　　处：《中国环境科学》2025年第2期954-965,共12页China Environmental Science

基　　金：重庆市科技局-院士专项(2022YSZX-JCX0008CSTB);西南大学创新研究2035先导计划项目(SWU-XDZD22003);国家自然科学基金资助项目(42202108)。

摘　　要：以重庆市中梁山凤凰村岩溶洼地为研究区,基于复合指纹识别技术,选取最佳指纹因子组合,利用多元线性混合模型(IsoSource)与贝叶斯混合模型(MixSIAR、SIMMR、SIAR)量化潜在泥沙源地对洼地沉积物的贡献率,并使用均方根误差(RMSE),结合前人的观测结果,评估模型的适用性.研究结果表明:沉积物总碳(TC)含量、砂粒含量、70%频度下土壤粒径(D70)、土壤有机碳(SOC)含量、硫元素(S)含量5种指纹因子的累积判别正确率为89.5%,可以作为最佳指纹因子组合.4个模型的RMSE为:MixSIAR(2.05)、SIMMR(2.05)、SIAR(2.07)、IsoSource(2.34),其中所有贝叶斯混合模型的RMSE均小于IsoSource模型,说明使用贝叶斯混合模型量化洼地沉积物泥沙来源比例的适用性高于多元线性混合模型.其中MixSIAR模型与SIMMR模型的准确性最高,两者的计算结果类似,均表明耕地是洼地沉积物主要的泥沙来源地,沟壁是仅次于耕地的第二大泥沙来源地,而林草地的侵蚀产沙量最低.本研究利用复合指纹识别技术揭示了西南典型岩溶槽谷区洼地沉积物的泥沙来源,以期为类似泥沙来源研究提供模型选择的参考.This study focuses on the karst depression in Fenghuang Village,Zhongliang Mountain,Chongqing,using the composite fingerprinting technique.An optimal combination of fingerprint properties was selected to quantify the contributions of potential sources to the depression deposits by using multivariate linear mixed model(IsoSource)and Bayesian mixed models(MixSIAR,SIMMR,and SIAR).In addition,the applicability of these models was further assessed by using their root mean square error(RMSE)and in combination with previous reports.Results indicated that the cumulative identification accuracy of the five fingerprint properties,i.e.,the total carbon(TC)content,sand content,grain size at 70%frequency(D70),soil organic carbon(SOC)content,and sulfur(S)content,reached 89.5%,and therefore these properties constituted an optimal combination.The RMSE values for the four models were:MixSIAR(2.05),SIMMR(2.05),SIAR(2.07),IsoSource(2.34).Since the RMSEs of the three Bayesian mixed models were lower than that of the IsoSource model,the applicability of the Bayesian mixed models for quantifying the contributions of sediment sources to the depression deposits was higher than that of the multivariate linear mixed model.Among the three Bayesian mixed models,the MixSIAR and SIMMR models had the highest accuracy.Results from the MixSIAR and SIMMR models indicated that arable land was the primary source of the depression deposits,followed by ditch walls,with forest and grassland contributing the least.The composite fingerprinting technique could effectively quantify the sediment sources in the small watersheds in the depression.In this study,the composite fingerprinting technique was employed to unveil the sediment source of depression deposits in a typical karst trough valley in Southwest China,aiming to provide a reference for model selection in similar studies.

关 键 词：土壤侵蚀 复合指纹 岩溶槽谷区 泥沙来源 模型 

分 类 号：X53[环境科学与工程—环境工程]