Uncertainty in radar‐rainfall composite and its impact on hydrologic prediction for the eastern Iowa flood of 2008

Abstract

[1] This study addresses a significant potential source of error that exists in radar-rainfall maps that are combined using data from multiple WSR-88D radars of the Next Generation Radar (NEXRAD) national network in the United States. This error stems from different radar calibration offsets that create a border within discontinuous rainfall fields at the equidistance zone among radars. The discontinuity in rainfall fields could lead to misestimation of rainfall over basins and subsequently, to significant errors in streamflow predictions through a hydrologic model. In this study, we produce enhanced radar-rainfall estimates (HN3) based on a novel approach that allows us to reduce the effects of the relative radar calibration bias. We use the relative bias information previously presented in a radar reflectivity comparison study. To investigate the effects of the relative bias adjustment, we evaluate the HN3 and Stage IV radar-rainfall by comparing them with rain gauge data and analyzing their ability to simulate streamflow for an extreme flood case. While the HN3 estimates are statistically comparable to the Stage IV estimates in the rain gauge data comparison, the borderline that identifies discontinuous rainfall fields disappears in the HN3 estimates. We performed hydrological simulations using a physically based, data-intensive, calibration-free, hillslope-link hydrologic model called CUENCAS and demonstrated CUENCAS's ability to accurately simulate flows by comparing results with observed and predicted streamflow generated by the Sacramento (SAC) model. SAC is the operational flood forecast model that has been used by the National Weather Service since 1969, and it was extensively calibrated based on historical data. The simulation results show that the adjustment improves streamflow predictions in the regions where the misestimation of rainfall quantity is considerable. We conclude that systematic error arising from different calibration offsets in rainfall fields can significantly affect hydrologic predictions.

Evaluation of NU-WRF Rainfall Forecasts for IFloodS

“…Stage IV data are available at 4-km grid spacing with 1-h frequency. It is a widely used rainfall product for both hydrological and meteorological communities because of its national coverage, high spatial and temporal resolutions, and overall low biases (e.g., Tang et al 2014;Seo et al 2013). Its good performance in mean-square error (MSE) and total bias results from the effectiveness of bias correction and the manual QC procedures (Cunha et al 2015).…”

Data-Enabled Field Experiment Planning, Management, and Research Using Cyberinfrastructure

“…The browser covers the period from 2002 to 2012 and enables users to identify significant rainfall events for their own purposes (e.g., flooding). We ingested Stage IV (Lin and Mitchell 2005), IFC (HUC 0708 and Flood 2008; Krajewski et al 2011Krajewski et al , 2013Seo et al 2013), and Q2 (Zhang et al 2011) radar-rainfall and CMORPH (Joyce et al 2004), PERSIANN (Hsu et al 1997), and TMPA 3B42, version 7 (Xue et al 2013), satelliterainfall products into a relational database that provides faster access to the data at users' requests. Since the use of a static image overlay for the display of rainfall maps often leads to a map distortion or projection error (especially at higher zoom levels), we utilize JavaScript, WebGL, and GPU to directly draw individual gridded rainfall cells on a map environment.…”

“…The Iowa legislature established the IFC in 2009, following a devastating flood in 2008 that caused multibillion dollar losses to the state, with the charge of conducting applied research to help the state mitigate future flood disasters. The Iowa group has experience (Demir and Krajewski 2013) and the Hydro-NEXRAD system Kruger et al 2011;Seo et al 2011). These two groups exchanged and developed new tools and ideas, many of which we describe in this paper.…”

A diagnostic study of spectral multiscaling on spatio-temporal accumulations of rainfall fields based on radar measurements over Iowa

“…However, this single metric is not sufficient to characterize the quality of the product. For example, [36] reported inter-radar calibration differences impacting hydrologic simulations of stream-flow in the Iowa River basin. Overall, uncertainty of radar-rainfall represents a structure in terms of systematic and random components of the involved errors.…”

“…This is possible by forcing spatially downscaled SARR/STARR fields, at such fine scales, into highly sophisticated hydraulic models generating flows through the river network of a given basin [9]. Moreover, spatial downscaling of SARR/STARR may potentially facilitate comparisons between remotely sensed measurements versus direct measurements made by networks of rain-gauges on the ground, thus contributing to the assimilation of different types of measurement accompanied by different types of error [3,4,7,36,37,49]. A body of literature on different methods, deterministic and stochastic, for temporal, spatial and spatiotemporal disaggregation (downscaling) of rain rate processes, is summarized in the introductory section of [35].…”