基于贝叶斯理论的水文多模型预报

水文预报对于防洪、抗旱以及水资源调度等具有重要意义。水文预报通常依靠水文模型来完成,由于受到不同流域特点、产汇流机制等的限制,每个水文模型都具有各自的特点及适用区域。单一模型具有非常大的水文预报不确定性,为了解决单一模型局限性的问题,多模型水文预报常作为降低水文预报不确定性有效方法之一。选用三种常见的水文模型:时变增益水文模型、新安江模型和萨克拉门托模型,在珠江飞来峡流域进行分布式建模,采用相同的输入与初始场,三个模型独立进行模拟,然后对比三个模型的结果,并进行贝叶斯多模型加权平均和简单平均得到多模型平均结果,研究结果表明,贝叶斯模型处理后的结果要比单个模型模拟结果和简单平均处理后的结果准确率高。

Hydrological multi-model forecast based on Bayesian Theory

Hydrological forecast is very important for flood control,drought resistance and water resources regulation.It is usually based on model simulation.Each hydrological model has its own characteristics and feasible basin.Multi-model hydrological forecast is one of the effective methods to reduce the forecasting uncertainty.This paper chose three commonly used hydrological models:time—variant gain hydrological model,Xin'anjiang model,and Sacramento model.The case study was on Feilaixia basin of Pearl River.The three hydrological models were used for independent simulation with the same input and initial value.Then,BMA and SMA were run on the three models'results.The research results show that the BMA results are better than the results from single model simulation.