Thông tin chung
  English
  Đề tài NC khoa học
  Bài báo, báo cáo khoa học
  Hướng dẫn Sau đại học
  Sách và giáo trình
  Các học phần và môn giảng dạy
  Giải thưởng khoa học, Phát minh, sáng chế
  Khen thưởng
  Thông tin khác
  Tài liệu tham khảo
  Hiệu chỉnh
 
Số người truy cập: 107,734,752

 Validating satellite precipitation products through input to a commonly-used hydrologic model
Tác giả hoặc Nhóm tác giả: C. Stephens, H. T. Pham, L. Marshall
Nơi đăng: The 23rd International Congress on Modelling and Simulation (MODSIM2019); Số: 2019;Từ->đến trang: 874;Năm: 2019
Lĩnh vực: Khoa học công nghệ; Loại: Báo cáo; Thể loại: Quốc tế
TÓM TẮT
In many large catchments, the spatial distribution of rain-gauge networks is important for accurately capturing average rainfall depths for input in hydrologic models. Where gauge density is not adequate, point measurements may misrepresent the overall rainfall amount potentially leading to poor model performance in simulating streamflow. As an alternative to gauge data, a range of satellite precipitation products (SPPs) are available for hydrologists to use when modelling large catchments. These products are either uncorrected products using only satellite data or corrected products based on gauge and/or (re)analysis data. It is common practice to validate satellite precipitation through comparison with rain gauge data. However, in many cases the rainfall statistics that are important for accurately simulating streamflow are not intuitive. For this reason, rather than comparing rainfall statistics directly, we propose assessment of satellite rainfall as a hydrologic model input. This involves calibrating a hydrologic model with various products to see which is able to give the best streamflow simulation. The two uncorrected SPPs compared in this study are the TRMM Multi-satellite Precipitation Analysis Real Time (TMPA-3B42RT) and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN). Two corrected SPPs based on satellite and gauge data (TMPA-3B42 and PERSIANN-CDR) and the Multi-Source Weighted-Ensemble Precipitation (MSWEP) are also included. These products combine satellite, gauge and reanalysis data. The study catchment for this work is Reedy Creek in Victoria (Hydrologic Reference Station 403209A), which has an area of 5500 km2 and contains 23 gauges covering the study period 01/06/2003 to 31/05/2013. Thiessen polygons were delineated to derive area-weighted gauged rainfall over the catchment. This was used as an input to the conceptual hydrologic model GR4J. Daily satellite precipitation from the five aforementioned SPPs was weighted based on grid cell area inside the catchment and also applied in the GR4J model. Potential evapotranspiration was the same for all simulations, estimated based on temperature data from the Australian Water Availability Project and the McGuiness-Bordne formula. Calibration over the ten-year period against streamflow data gave the following results in terms of Nash-Sutcliffe Efficiency (NSE): Data source Gauges (23) MSWEP TMPA-3B42 PERSIANN-CDR TMPA-3B42RT PERSIANN Calibration NSE 0.86 0.74 0.61 0.47 0.19 0.09 These results give an indication of how well GR4J is able to simulate streamflow with each input rainfall series. High calibration NSE indicates the dataset represents the rainfall statistics such that the model is able to reproduce the observed streamflow series well. The rain gauge data clearly outperformed all of the SPPs when 23 gauges were available. The uncorrected SPPs performed worse than the corrected SPPs. Among the three corrected SPPs, the model using MSWEP data had the best performance. The TMPA-3B42RT based on passive microwave data slightly outperformed the PERSIANN using infrared-based dataset. It is also important for hydrologists to understand the gauge density required for point measurements to outperform gridded satellite products. This aids in selection of the best available dataset for each application. As a next step, we repeatedly removed a given number of gauges at random and recalculated the rainfall series with reduced gauge density. For the Reedy Creek catchment, we found that gauge data generally outperformed even the most reliable SPP until the number of gauges was reduced to two. This will be investigated further, and for more catchments, in later work
ABSTRACT
In many large catchments, the spatial distribution of rain-gauge networks is important for accurately capturing average rainfall depths for input in hydrologic models. Where gauge density is not adequate, point measurements may misrepresent the overall rainfall amount potentially leading to poor model performance in simulating streamflow. As an alternative to gauge data, a range of satellite precipitation products (SPPs) are available for hydrologists to use when modelling large catchments. These products are either uncorrected products using only satellite data or corrected products based on gauge and/or (re)analysis data. It is common practice to validate satellite precipitation through comparison with rain gauge data. However, in many cases the rainfall statistics that are important for accurately simulating streamflow are not intuitive. For this reason, rather than comparing rainfall statistics directly, we propose assessment of satellite rainfall as a hydrologic model input. This involves calibrating a hydrologic model with various products to see which is able to give the best streamflow simulation. The two uncorrected SPPs compared in this study are the TRMM Multi-satellite Precipitation Analysis Real Time (TMPA-3B42RT) and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN). Two corrected SPPs based on satellite and gauge data (TMPA-3B42 and PERSIANN-CDR) and the Multi-Source Weighted-Ensemble Precipitation (MSWEP) are also included. These products combine satellite, gauge and reanalysis data. The study catchment for this work is Reedy Creek in Victoria (Hydrologic Reference Station 403209A), which has an area of 5500 km2 and contains 23 gauges covering the study period 01/06/2003 to 31/05/2013. Thiessen polygons were delineated to derive area-weighted gauged rainfall over the catchment. This was used as an input to the conceptual hydrologic model GR4J. Daily satellite precipitation from the five aforementioned SPPs was weighted based on grid cell area inside the catchment and also applied in the GR4J model. Potential evapotranspiration was the same for all simulations, estimated based on temperature data from the Australian Water Availability Project and the McGuiness-Bordne formula. Calibration over the ten-year period against streamflow data gave the following results in terms of Nash-Sutcliffe Efficiency (NSE): Data source Gauges (23) MSWEP TMPA-3B42 PERSIANN-CDR TMPA-3B42RT PERSIANN Calibration NSE 0.86 0.74 0.61 0.47 0.19 0.09 These results give an indication of how well GR4J is able to simulate streamflow with each input rainfall series. High calibration NSE indicates the dataset represents the rainfall statistics such that the model is able to reproduce the observed streamflow series well. The rain gauge data clearly outperformed all of the SPPs when 23 gauges were available. The uncorrected SPPs performed worse than the corrected SPPs. Among the three corrected SPPs, the model using MSWEP data had the best performance. The TMPA-3B42RT based on passive microwave data slightly outperformed the PERSIANN using infrared-based dataset. It is also important for hydrologists to understand the gauge density required for point measurements to outperform gridded satellite products. This aids in selection of the best available dataset for each application. As a next step, we repeatedly removed a given number of gauges at random and recalculated the rainfall series with reduced gauge density. For the Reedy Creek catchment, we found that gauge data generally outperformed even the most reliable SPP until the number of gauges was reduced to two. This will be investigated further, and for more catchments, in later work
© Đại học Đà Nẵng
 
 
Địa chỉ: 41 Lê Duẩn Thành phố Đà Nẵng
Điện thoại: (84) 0236 3822 041 ; Email: dhdn@ac.udn.vn