尹家波,郭生练,王俊,朱青,曾青松,刘汉武.基于贝叶斯模式平均方法融合多源数据的水文模拟研究[J].水利学报,2020,51(11):1335-1346 |
基于贝叶斯模式平均方法融合多源数据的水文模拟研究 |
Blending multi-source data in hydrological simulations based on BMA method |
投稿时间:2020-03-13 |
DOI:10.13243/j.cnki.slxb.20200154 |
中文关键词: 卫星降水 再分析数据 贝叶斯模式平均 偏差校正 水文模拟 |
英文关键词: satellite precipitation reanalysis data Bayesian model averaging bias correction hydrological simulation |
基金项目:国家自然科学基金项目(51879192,52009091,51539009);博士后创新人才支持计划项目(BX20200257) |
|
摘要点击次数: 990 |
全文下载次数: 679 |
中文摘要: |
可靠的长系列气象数据是开展流域水文模拟、水旱灾害防治和水资源综合管理的基本依据,但是我国气象站网布设不均、地面观测资料系列相对较短,难以满足工程应用需要。本文融合有限的地面气象观测数据,长系列高精度MSWEP-V2卫星集成降水数据集和欧洲中期天气预报中心的ERA5气温数据,首先通过基于分位数映射的日偏差校正(DBC)、基于月尺度的回归校正(LRBC)和等率校正(RBC)等3种方法,对遥测栅格降水和再分析气温日系列进行偏差校正,再采用季节性贝叶斯模式平均(BMA)方法描述各偏差校正系列的后验分布优选相应权重,从而得到融合多种偏差校正模式的长系列日降水、气温过程。以巢湖流域为例,采用174个自动气象站2015—2019年的观测数据和7个国家基本气象台站1979—2019年的长系列资料检验校正效果,并在2个子流域分别驱动新安江、GR4J和HMETS水文模型验证水文模拟的适用性。结果表明:BMA方法能够综合考虑各偏差校正方法的优势,校正后的日降水和气温数据偏差较小,与实测系列的相关性系数接近0.8;水文模型率定期及检验期的KGE系数超过0.67,校正后的气象数据满足水文模拟要求。 |
英文摘要: |
Reliable meteorological series with long records are important basis for hydrological simulation, flood and drought mitigation, and integrated water resources management, but the gauge-based networks in China are uneven distributed and have short records and thus inhibiting their engineering application. This study blends in-situ observations, a lengthy series of MSWEP-V2 precipitation products and ECMWF ERA5 temperature datasets; the gridded and reanalysis data are firstly corrected by Daily Bias Correction (DBC) method based on quantile mapping, monthly Linear Regression Bias Correction (LRBC) and Ratio Bias Correction (RBC) methods, and then the seasonal Bayesian Model Averaging (BMA) approach is employed to optimize the weight of those bias-corrected series based on their posterior distributions and to derive integrated precipitation and temperature series. Chao Lake is selected as a case study,and the 174 automatic stations with 2015-2019 records and seven national meteorological stations with 1979-2019 records are used to validate the performance of bias correction methods,and are then employed to validate their hydrological applications by driving Xinanjiang, GR4J and HMETS hydrological model in two sub-basins. The results demonstrate that the BMA method enables considering all bias correction methods' advantages, because the bias after correction is relatively low and the correlation coefficients approach 0.8; and the KGE values during both calibration and validation periods outweigh 0.67, implying that the bias-corrected meteorological series have a good capacity in hydrological simulations. |
查看全文
查看/发表评论 下载PDF阅读器 |
关闭 |
|
|
|