Special Issue on Radar Rainfall Data Analyses and Applications

The use of technology to estimate radar-based precipitation (rainfall) began in the early 1960s. During the early 1990s, the use of this technology proliferated as the National Oceanic and Atmospheric Administration (NOAA) installed many radar stations across the United States as part of the Weather Surveillance Radar–88-Doppler (WSR-88D) and the Next-Generation Weather Radar (NEXRAD) programs initiated by the National Weather Service (NWS). Currently, more than 160 WSR-88D radar stations are in operation, providing nearly contiguous coverage across most of the United States and a way of estimating the intensity of rain or snowfall.

WSR-88D radars operate by emitting short (250 m) pulses of coherent microwave energy. When a target is encountered, such as a building, airplane, bird, or precipitation droplet, the emitted energy is scattered in all directions. Small amounts of energy, known as backscatter, are returned to the radar where they are detected and recorded. The intensity of the returned signal is then related to the size of the object and analyzed according to the time required for the pulse to reach the target and return. This provides information regarding the range and Doppler velocity of the target relative to the radar. The WSR-88D radar is a 10-cm wavelength (S-band) radar. It is designed for long-distance surveillance because its wavelength penetrates rainfall with little attenuation. However, European radar systems predominantly use lower (less than 5 cm) wavelength (C-band) radar. Shorter wavelength radars suffer attenuation caused by the absorption and scattering of the electromagnetic radiation that degrades performance as the distance from the radar increases, thus requiring an increased density of radar systems. Under most conditions, the useful range of an S-band radar is considered to be approximately 180 km, although the WSR-88D system produces precipitation estimates up to 230 km. As distance increases, the beam becomes increasingly higher above average ground level because the radar’s lowest elevation angle is 0.5 degree, the Earth is curved, and there is atmospheric refraction.

Radar data provide rainfall amount products with two primary spatial resolutions: $4\times 4\mathrm{km}$ Cartesian, and $2\mathrm{km}\times 1°$ azimuth-range. The radar rainfall data are limited by relying on the measurement of raindrop reflectivity, which can be affected by factors such as raindrop size and signal reflection by other objects. Because the reflected signal measured by the radar is proportional to the sum of the sixth power of the diameter of the raindrops in a given radar volume, small changes in the size of raindrops can have a dramatic effect on the radar’s estimate of rainfall. For this reason, and because of overshooting beams and radar miscalibration, radar-derived rainfall data generally are scaled to match the volume measured at coincident rain gauges by using bias adjustment techniques.

There is a statistical tradeoff between rainfall data collected by rain gauges and by radars. Rain gauges can provide point values of rainfall depth and intensity but may not cost effectively provide the spatial distribution of rainfall on a regional scale. Whereas rain gauges might suffice for widespread rain events, a gauge network can miss smaller-scale localized convective rainfall events altogether. For this reason, radar rainfall estimates incorporate coincident rain gauge data in several ways.

Radar rainfall outputs are generally in the native polar coordinate system of the radar or resampled to a grid in Cartesian coordinates. Aggregation or disaggregation of gridded radar rainfall data is often necessary with grid-based distributed models because the model grid is not at the same resolution as the radar rainfall data input or has a different geographic projection. In addition to file-format manipulation, the link between radar rainfall data and hydrologic models requires spatial aggregation from one grid to another or to subbasin areas. Aggregation from polar coordinates to basins or to rectangular grids is usually accomplished with the help of GIS or special-purpose spatial analysis tools that can handle spatial data in geographic projections.

High-quality radar rainfall records in the United States have only been kept since the mid-1990s. Moreover, some significant changes in computation algorithms have been implemented during that period. Fifteen years of records are still insufficient to use radar rainfall data for long-term statistical studies. However, short-term storm analyses can be performed using radar rainfall data for a specific, recent historical event coupled with continuous hydrologic modeling.

Although radar rainfall data are increasingly available over the Internet, data processing, quality assurance/quality control, and radar rainfall data calibration using rain gauge data, all require specialized skills not widely available in the engineering community. At present, those skill sets are limited to a small number of firms or organizations specializing in radar rainfall estimation for hydrologic applications. The use of governmental unadjusted radar data without proper understanding, or the application of tools for quality control and correcting for bias, could lead to serious errors in hydrologic analyses. The methods of estimating precipitation from radar have changed appreciably since the original deployment of the weather radar network, and these methods will continue to evolve. For example, statistical biases can change over time as algorithms are refined. Thus, a major challenge is the proper interpretation of radar-derived rainfall estimates in long time series of 10 years or more.

A primary advantage of radar-derived rainfall data is the density of measurements that is not obtainable by rain gauges alone. Combining these two sensor systems produces better rainfall estimates that more accurately characterize rainfall over a watershed. By using radar rainfall data, hydrologists benefit from having more information about rainfall rates at high resolution in space and time over large areas. How radar measures rainfall rates depends on assumptions about the number and sizes of raindrops in a representative volume of the atmosphere. Various reflectivity ($\mathit{Z}$) and rainfall rate ($R$) relationships have been derived from a theoretical or empirical basis. Depending on storm type and the power of the radar, a range of $\mathit{Z}$-$R$ relationships is possible. Once an appropriate $\mathit{Z}$-$R$ relationship is selected, comparison with rain gauge accumulations is carried out to remove any systematic error, known as bias. After bias removal, differences between radar rainfall and gauge rainfall estimates remain as random error. Dramatic progress has been made in the recent decade toward achieving representative rainfall measurement through a variety of methods. More promising methods include dual-polarization and postprocessing algorithms to account for attenuation and other artifacts in short-wavelength radars such as X- and C-band radars.

Distributed models are particularly well-suited to using radar rainfall. Understanding the importance of radar rainfall estimation errors within a modeling context has particular advantages. Some errors tend to average out over a watershed area, whereas others may dramatically increase prediction errors. Understanding real-time and postanalysis quality control and postprocessing will yield improved radar rainfall data for hydrologic applications. Distributed models designed from the outset to use high-resolution rainfall rates from multiple sensors (radar, satellite, and rain gauge) allow detailed predictions at almost any location in a watershed characterized with geospatial data relating to topography, soils, and land cover. In addition to hydrologic design, radar rainfall can also provide timely inputs to operational decisions for water control structures and flood warning systems and can also guide emergency management personnel in taking proactive steps to protect life and property from flooding. Considering the need for precipitation measurements at different spatial and temporal scales for hydrologic analysis and prediction, radar-derived rainfall estimation is clearly one of the most significant advances in hydrologic engineering and practice.

Within the previously discussed framework, the papers in this special issue can be grouped into four categories: (1) two papers are about radar rainfall data development methods; (2) six papers are about radar rainfall data improvement and validation; (3) four papers are about the application of radar rainfall data; and (4) and five papers are case studies.

The papers titled “Radar and Multisensor Precipitation Estimation Techniques in National Weather Service Hydrologic Operations” and “Independent Assessment of Incremental Complexity in NWS Multisensor Precipitation Estimator Algorithms” fall into the category of radar rainfall data development methods.

The papers titled “Validation of the NEXRAD DSP Product with a Dense Rain Gauge Network,” “Use of Storm Life Cycle Information and Lightning Data in Radar-Rainfall Estimation,” “Continuous Forecasting and Evaluation of Derived $\mathit{Z}$-$R$ Relationships in a Sparse Rain Gauge Network Using NEXRAD,” “Rainfall Space-Time Organization and Orographic Control on Flash Flood Response: The Weisseritz Event of August 13, 2002,” “Deriving Spatially Distributed Precipitation Data Using the Artificial Neural Network and Multilinear Regression Models,” and “Self-Learning Cellular Automata for Forecasting Precipitation from Radar Images” are in the radar rainfall data improvement and validation category.

The papers about the application of radar rainfall data category are “Spatial Assessment of Five Years of WSR-88D Data over the Mississippi River Basin and Its Estimation Bias around Rain Gauge Sites,” “Radar Rainfall Application in Distributed Hydrologic Modeling for Cypress Creek Watershed, Texas,” “Physically Based Hydrological Modeling of the 2002 Floods in San Antonio, Texas,” and “Estimation of Spatio-Temporally Variable Groundwater Recharge Using a Rainfall-Runoff Model.”

The five case studies in this issue include “Hydrologic Analyses of the July 17–18, 1996, Flood in Chicago and the Role of Urbanization,” “Distributed Hydrologic Forecast Reliability Using Next-Generation Radar,” “Using Local Weather Radar Data for Sewer System Modeling: Case Study in Flanders, Belgium,” “Preparation and Evaluation of a Dutch-German Radar Composite to Enhance Precipitation Information in Border Areas,” and “Prediction of Rainfall-Runoff in an Ungauged Basin: Case Study in the Mountainous Region of Northern Thailand.”

As the guest editor, I sincerely hope that this special issue on the topic of radar rainfall data analyses and applications is useful to researchers and practicing engineers alike as this discipline of hydrologic engineering progresses in the future.