基于线性-非线性水库出流耦合模型的基流自动分割方法及应用

Baseflow separation is the key to watershed hydrological process study, water resource optimal allocation, and agricultural non-point source pollution research. However, the existing automatic baseflow separation algorithms are commonly based on a simple assumption that the relationship between the sub-surface water storage and outflow is linear or nonlinear, which might not be in accordance with the the actual situation. Besides, the rising limb baseflow separation using those existing automatic baseflow separation algorithms is usually criticised for the lack of physical basis, which may pose some uncertainty to baseflow separation results. In this study, the upriver watershed of South Tiaoxi River was selected as study area. Based on the linear/nonlinear reservoir-outflow model theory, numerical optimization algorithm (such as Genetic Algorithm, Bayes Method, and some other optimized functions in Matlab), and stable isotope baseflow separation technique, the major aims of this project are 1) to set up an automatic linear-nonlinear reservoir-outflow coupling model baseflow separation algorithm to separate (linear reservoir and nonlinear reservoir) baseflow of the rising limb, surface flow-basflow mixed falling limb, and pure basflow falling limb more accurately, reliably, and efficiently; 2) to make a quantitative analysis on the change law of baseflow contributed by the linear and nonlinear reservoirs under different conditions to reveal the mechanism of watershed baseflow hydrological processes. And, it is expected that the results would be beneficial to the development of baseflow separation methodology and baseflow non-point source pollution research in the future.

基流分割是流域水文过程机理研究、水资源优化配置以及农业非点源污染研究的关键。然而，现有的基流自动分割算法通常将地下水储量-出流量关系假定为单一线性或非线性关系，与实际地下水出流情况不符；加之此类方法对涨水段基流的分割缺乏物理基础，可能导致基流分割结果较大的不确定性。本项目以南苕溪上游流域为研究对象，根据线性/非线性水库-出流模型理论，综合运用遗传算法、贝叶斯法和Matlab内嵌的其他数值优化算法，结合稳定同位素基流分割技术，建立一种更贴近地下水实际出流情况的基于线性-非线性水库出流耦合模型的基流自动分割算法，实现涨水段、存在地表径流的退水段以及纯基流退水段(线性库、非线性库)基流准确可靠的自动分割。在此基础上，定量分析本研究区不同条件下的线性库和非线性库基流的变化规律，揭示流域基流水文过程的机理，以期为今后基流分割方法的发展和基流非点源污染研究提供理论探索和技术支撑。

基流的准确定量是当前乃至今后水资源调查评价、非点源污染定量研究、水资源优化配置必须解决的一个关键问题。然而，现有的基流分割算法或者模型通常基于单个线性或者非线性水库出流模型，这种简单的假设可能与流域地下水实际的出流情况存在较大的出入，从而可能导致基流分割结果中存在较大的不确定性。本项目在充分研究基于单个线性或非线性水库出流模型的分割算法在我国南方典型多雨地区基流及其非点源污染负荷定量方面适用性的基础上，综合运用遗传算法（Genetic Algorithm）和粒子群优化算法（Particle swarm optimization），结合稳定同位素基流分割技术，建立了一种基于线性-非线性水库出流耦合模型的自适应基流分割算法 (ALNA)，实现涨水段、存在地表径流的退水段以及纯基流退水段基流准确可靠的自动分割。对比分析ALNA与传统基于单个（线性或非线性）水库出流模型的基流分割算法的优缺点，为今后基流分割方法的发展和基流非点源污染研究提供理论和技术支撑。