期刊
JOURNAL OF HYDROLOGY
卷 558, 期 -, 页码 275-289出版社
ELSEVIER
DOI: 10.1016/j.jhydrol.2018.01.042
关键词
Satellite precipitation; Precipitation data fusion; Precipitation downscaling; Geographically weighted regression; Mixed geographically weighted regression
资金
- National Key Research and Development Program of China [2016YFC0402701, 2016YFC0402705, 2016YFC0402710]
- National Natural Science Foundation of China [51539003]
- Open Foundation of State Key Laboratory of Hydrology-Water Resources and Hydraulic Engineering [20165042212, 2017490311]
- MWR of China [201501022]
- Fundamental Research Funds for the Central Universities of China [2015B28514]
- Priority Academic Program Development of Jiangsu Higher Education Institutions
Real-time precipitation data with high spatiotemporal resolutions are crucial for accurate hydrological forecasting. To improve the spatial resolution and quality of satellite precipitation, a three-step satellite and gauge precipitation merging method was formulated in this study: (1) bilinear interpolation is first applied to downscale coarser satellite precipitation to a finer resolution (P-s); (2) the (mixed) geographically weighted regression methods coupled with a weighting function are then used to estimate biases of Ps as functions of gauge observations (P-o) and P-s; and (3) biases of P-s are finally corrected to produce a merged precipitation product. Based on the above framework, eight algorithms, a combination of two geographically weighted regression methods and four weighting functions, are developed to merge CMORPH (CPC MORPHing technique) precipitation with station observations on a daily scale in the Ziwuhe Basin of China. The geographical variables (elevation, slope, aspect, surface roughness, and distance to the coastline) and a meteorological variable (wind speed) were used for merging precipitation to avoid the artificial spatial autocorrelation resulting from traditional interpolation methods. The results show that the combination of the MGWR and BI-square function (MGWR-BI) has the best performance (R = 0.863 and RMSE = 7.273 mm/day) among the eight algorithms. The MGWR-BI algorithm was then applied to produce hourly merged precipitation product. Compared to the original CMORPH product (R = 0.208 and RMSE = 1.208 mm/hr), the quality of the merged data is significantly higher (R = 0.724 and RMSE = 0.706 mmihr). The developed merging method not only improves the spatial resolution and quality of the satellite product but also is easy to implement, which is valuable for hydrological modeling and other applications. (C) 2018 Elsevier B.V. All rights reserved.
作者
我是这篇论文的作者
点击您的名字以认领此论文并将其添加到您的个人资料中。
推荐
暂无数据