Statewide real-time quantitative precipitation estimation using weather radar and NWP model analysis: Algorithm description and product evaluation
Introduction
Using data from the U.S. Weather Surveillance Radar-1988 Doppler (WSR-88D) network, the Iowa Flood Center (IFC) has provided a statewide real-time rainfall product since the IFC's establishment in 2009. This product generation was motivated by the need for real-time flood prediction in Iowa, which has repeatedly experienced devastating floods at various scales in recent decades (e.g., Smith et al., 2013; Vennapusa and White, 2015; Seo et al., 2018). Our goal to forecast “everywhere” and cover all Iowa communities (e.g., regardless of catchment scale) has led to the use of a distributed hydrologic model, which requires spatially variable rainfall inputs (Krajewski et al., 2017).
The IFC quantitative precipitation estimation (QPE) framework was initially built on the real-time Hydro-NEXRAD application (Krajewski et al., 2011a, 2013; Kruger et al., 2011; Seo et al., 2011). Most scientific algorithms in our QPE system have evolved according to the WSR-88D's hardware and polarimetric upgrades (e.g., Istok et al., 2009). The IFC QPE system acquires real-time data from seven WSR-88Ds (KARX in La Crosse, Wisconsin; KDMX in Des Moines, Iowa; KDVN in Davenport, Iowa; KEAX in Kansas City, Missouri; KFSD in Sioux Falls, South Dakota; KMPX in Minneapolis, Minnesota; and KOAX in Omaha, Nebraska), as shown in Fig. 1. The system then generates a composite rain rate map covering the entire domain (Fig. 1), with temporal and spatial resolutions of 5 min and 0.5 km, respectively, using a variety of processing algorithms, as documented in Seo et al. (2011, 2015) and Seo and Krajewski (2015).
As of early 2019, we had added a state-of-the-art polarimetric algorithm known as “specific attenuation” (e.g., Ryzhkov et al., 2014; Wang et al., 2019) to our QPE procedures to fully benefit from the WSR-88D's dual-polarization (DP) capability. This new algorithm required several new elements; for example, one that retrieves temperature soundings from the numerical weather prediction (NWP) model analyses to identify the melting layer (ML) location. Before this new implementation, the use of DP in the IFC system was limited to basic data quality control (e.g., removal of non-meteorological returns), and the system's main estimator was a single polarization–based algorithm using a reflectivity-rain rate (Z-R) relation. The new implementation using the specific attenuation method promises to be the most significant milestone in our system's 10-year history, as the method has demonstrated meaningful improvements in QPE accuracy (Seo et al., 2020b).
Therefore, we take this opportunity to document the architecture and capabilities of our fully automated QPE system, including algorithm updates and new developments as well as the way it complements the outdated descriptions presented in Seo et al. (2011). To validate the attainable improvement in QPE and subsequent hydrologic prediction, we generated the statewide QPE products using the latest and prior algorithms for a three-year period (2016–2018) and evaluate the performance of each one using rain and stream gauge observations. We also compared the performance of our QPE products with that of U.S. national QPE products (e.g., Zhang et al., 2016; Cunha et al., 2013) that have been widely used for meteorological and hydrological applications.
In Section 2, we describe the architecture of our QPE system by specifying the three main subsystems associated with real-time data acquisition, NWP analysis and radar data processing, and final product visualization. Section 3 provides the algorithm details of module elements in the NWP, individual radar, and composite data processing. Section 4 evaluates the QPE products generated by our algorithms using rain gauge data and hydrologic simulations. In Section 5, we summarize the algorithm features and main findings from the product evaluation. Finally, we discuss the improvements gained by implementing the new polarimetric algorithm and potential future developments.
Section snippets
System architecture
A real-time QPE system requires several modular elements, ranging from data acquisition to final product generation. These modules include real-time radar and NWP data retrieval, algorithms for data quality control, volume scan data processing and precipitation estimation, and digital product and statewide map generation. Our QPE system comprises three main subsystems: (1) the Local Data Manager (LDM; e.g., Fulker et al., 1997) system for real-time NWP analysis and radar data acquisition; (2)
Algorithms
In this section, we provide details on scientific algorithms implemented in the data processing subsystem described in Section 2. The processing algorithms are categorized into three groups: (1) NWP analysis processing; (2) individual radar data processing; and (3) composite data processing. Because the system structure and algorithm elements have continuously evolved over time since their initial deployment, we also discuss the changes and advancements of the algorithm components that aim to
Product evaluation
We assess the QPE products generated via the combination of algorithms described in Section 3 for the three-year period of 2016–2018 in two ways: (1) quantitative comparison with ground reference (i.e., rain gauge): observations; and (2) evaluation of hydrologic simulations driven by the QPE products using streamflow observations. We generated the products using a QPE reproduction system (Seo et al., 2019) that can readily retrieve the radar Level II data from Amazon's cloud archive (e.g.,
Summary and discussion
In this paper, we document the architecture, algorithm configuration, and estimation performance of our real-time QPE system used primarily for flood prediction in Iowa. The IFC system is the only academia-based large-scale (statewide) real-time radar-rainfall monitoring system in the United States. The system retrieves data from operational radars and NWP models and uses a variety of scientific algorithms to account for precipitation microphysics and uncertainties in radar remote sensing
Declaration of competing interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.
Acknowledgments
The Iowa Flood Center at the University of Iowa supported this work. The evaluation of various QPE products was partially supported by the NOAA Hydrometeorology Testbed (HMT) Program within NOAA/OAR Office of Weather and Air Quality under Grant NA17OAR4590131. Thanks to Dr. Felipe Quintero at the Iowa Flood Center for providing the HLM simulation results driven by the different QPE products for the hydrologic evaluations in this project.
References (63)
- et al.
Towards an integrated flood information system: centralized data access, analysis, and visualization
Environ. Model. Software
(2013) - et al.
High resolution rainfall measurements by radar for very small basins: the sampling problem re-examined
J. Hydrol
(1994) - et al.
Decomposition of the mean squared error and NSE performance criteria: implications for improving hydrological modelling
J. Hydrol
(2009) - et al.
Inter-comparison of reflectivity measurements between GPM DPR and NEXRAD Radars
Atmos. Res.
(2019) - et al.
Statistical model of the range-dependent error in radar-rainfall estimates due to the vertical profile of reflectivity
J. Hydrol
(2011) - et al.
River flow forecasting through conceptual models. part I–A discussion of principles
J. Hydrol
(1970) - et al.
Improvement and evaluation of the Iowa flood center hillslope link model (HLM) by calibration-free approach
J. Hydrol
(2020) - et al.
A pilot infrastructure for searching rainfall metadata and generating rainfall product using the big data of NEXRAD
Environ. Model. Software
(2019) - et al.
Correcting temporal sampling error in radar-rainfall: effect of advection parameters and rain storm characteristics on the correction accuracy
J. Hydrol
(2015) - et al.
Using the new dual-polarimetric capability of WSR-88D to eliminate anomalous propagation and wind turbine effects in radar-rainfall
Atmos. Res.
(2015)
Utility of vertically integrated liquid water content for radar-rainfall estimation: quality control and rain type classification
Atmos. Res.
An asynchronous solver for systems of ODEs linked by a directed tree structure
Adv. Water Resour.
Seeded region growing
IEEE T. Pattern Anal.
Unlocking the potential of NEXRAD data through NOAA's big data partnership
Bull. Am. Meteorol. Soc.
Identification of the melting layer through dual-polarization radar measurements at vertical incidence
J. Atmos. Ocean. Technol.
Error statistics of VPR corrections in stratiform precipitation
J. Appl. Meteorol.
A North American hourly assimilation and model forecast cycle: the Rapid Refresh
Mon. Weather Rev.
Recording, archiving, and using WSR-88D data
Bull. Am. Meteorol. Soc.
An early performance evaluation of the NEXRAD dual-polarization radar rainfall estimates for urban flood applications
Weather Forecast.
Long term observations of the melting layer of precipitation and their interpretation
J. Atmos. Sci.
Unidata: a virtual community sharing resources via technological infrastructure
Bull. Am. Meteorol. Soc.
The WSR-88D rainfall algorithm
Weather Forecast.
Automatic designation of the melting layer with a polarimetric prototype of the WSR-88D radar
J. Appl. Meteor. Climatol.
Snow versus rain: looking beyond the “magic” numbers
Weather Forecast.
Sampling interval and area effects on radar-derived rainfall estimates
WSR-88D dual-polarization initial operational capabilities. Preprints
The objective use of observed and forecast thickness values to predict precipitation type in North Carolina
Weather Forecast.
A real-time delivery system for NEXRAD Level II data via the internet
Bull. Am. Meteorol. Soc.
Characteristics of reprocessed hydrometeorological automated data system (HADS) hourly precipitation data
Weather Forecast.
Real-time flood forecasting and information system for the State of Iowa
Bull. Am. Meteorol. Soc.
Hydro-NEXRAD-2: real-time access to customized radar-rainfall for hydrologic applications
J. Hydroinf.
Cited by (18)
EfficientRainNet: Leveraging EfficientNetV2 for memory-efficient rainfall nowcasting
2024, Environmental Modelling and SoftwareTowerpy: An open-source toolbox for processing polarimetric weather radar data
2023, Environmental Modelling and SoftwareA rebalanced performance criterion for hydrological model calibration
2022, Journal of HydrologyCitation Excerpt :To drive standalone simulations of the CoLM, this study has constructed the meteorological forcing data from the meteorological point data directly measured at the Automated Synoptic Observing System (ASOS) in the 77 meteorological stations managed by the Korea Meteorological Administration (KMA), as shown in Fig. 3. In the ASOS meteorological data network (https://data.kma.go.kr/cmmn/main.do) widely used for hydrological modeling and climate change analysis in the Republic of Korea (Ahn and Kim, 2019; Jung et al., 2020; Seo and Krajewski, 2020), precipitation (mm), snow (cm), air pressure (hPa), temperature (oC), specific humidity (%), zonal/meridional wind speeds (m/s), and downward long/short wave radiation (MJ/m2) are available for the past thirty years from 1990 to 2019. For the grid-based meteorological forcing data in the CoLM, the ASOS point data were spatially interpolated by the Inverse Distance Weight (IDW) method onto the 30-km computational domain covering the four study watersheds.
Multi-scale investigation of conditional errors in radar-rainfall estimates
2021, Advances in Water ResourcesCitation Excerpt :The IFC product is a composite of data from seven NEXRAD radars that illuminate the Iowa domain as shown in Fig. 1. Its current algorithm (Seo et al., 2020; Seo and Krajewski, 2020) uses the specific attenuation method (e.g., Ryzhkov et al. 2014, Wang et al. 2019) based on polarimetric observations and environmental variables that help identify the altitude of the melting layer. While the IFC product is generated in real time with 500 m and 5 min resolutions, we reproduced an hourly-basis product for the period of 2016 to 2018 to avoid data missed in real-time operation.