While TC damage from wind hazards has been the focus of research for many years, TC induced flood vulnerability is an emerging and an ill posed problem. As a result, communities and infrastructure are arguably better prepared to withstand wind hazards than those from TC flooding. Next generation efforts now underway are focused on understanding feedback between flooding and localized rain from TCs. These hydrology-based studies are important as communities develop regional infrastructure and vulnerability assessments. Since all flood events and associated socioeconomic impacts are not equal, there is a need to assess location-based flooding and rainfall impacts using globally available datasets. However, it is still not clear which rainfall products should be used or what that choice means for the flood impact assessment. In this presentation, we summarize our experience, building on a series of studies to investigate answers to the questions (i) Are satellite quantitative precipitation estimates (QPEs) reliable in the TCs scenarios? (ii) How variable are the gridded precipitation data products from each other? and (iii) What is the implication of the difference in gridded precipitation data products on surface hydrology? Is satellite informed streamflow reliable enough to act in extreme weather event? Each of these are outlined below.
Multiple storms, namely, Hurricane Charley 2004, Hurricane Frances 2004, Hurricane Jeanne 2004, Tropical Storm Fay 2008, Tropical Storm Beryl 2012, Hurricane Irma 2017, Hurricane Michael 2018 have been selected carefully to represent the diversity of strength and its impact on land to observe how well they are captured using various gridded precipitation estimates. Each of these storms have unique characteristics in terms of its origin, formation, accumulated cyclone energy, track over land and ocean, the atmospheric and land conditions that drive the spatial and quantitative distribution of precipitation from such a system of rain. For example, the maximum rain observed by NCEP during these storms were 9.88, 23.57, 11.97, 27.65, 15, 21.66, 13.01 inches respectively. The only common aspect is the study region – South Atlantic Gulf region which is impacted by these storms. Gridded precipitation products used for this study are NCEP Stage-IV, GPM IMERG and interpolated gauge observation from NWS Observer Program (COOP network data) over CONUS (G-COOP).
Our previous study involving Hurricane Florence 2018, that caused widespread flooding, noted the different quantitative precipitation estimates (QPEs) highlighted disparity in the assessments. The comparison analysis shows that GPM IMERG (RMSE = 23.97 mm/day, r2 = ~0.8) followed by StageIV (RMSE = 26.46 mm/day, r2 = ~0.71) were in close agreement with the in-situ Automated Surface Observing Sites (ASOS) observations.
The analysis in the case of TC Beryl 2012 highlighted the differences between GPM IMERG, Stage IV and G-COOP precipitation estimates. The spatiotemporal distribution of these products were analyzed using the MODE analysis that quantifies more nuanced characteristics of rainfall between data products. The MMI score derived from MODE analysis established that IMERG has good match with other gridded products (Stage IV and G-COOP) for rainfall accumulation < 65 mm which is consistent with the findings of qualitative analysis of spatial patterns. The area ratio of the objects formed by the convolution precipitation and radius threshold is ~0.86, while the 90th and 50th percentile intensity ratios are 0.77 and 0.76, between IMERG and G-COOP (meaning 77 percent area has similar 90th percentile IMERG derived precipitation accumulation as G-COOP).
To quantify how differences in rainfall intensity and location derived from these gridded precipitation datasets impact surface hydrology simulations, the Variable Infiltration Capacity (VIC) model has been used to assess the propagation of variability in daily rainfall rate to the time of peak streamflow as well as the total volume of surface runoff leading to flooding. The VIC hydrology model is supplemented with GIS Routing model to observe the streamflow amounts. Results from hydrologic modeling help study the hypothesis that surface runoff and time of peak of streamflow will be suitable for flood warning despite the differences in total precipitation flux from the storm. Testing this hypothesis in multiple storm cases, this will help attest the value of IMERG and other globally available precipitation datasets as a valuable resource for estimating hydrologic effects of TC storms worldwide.
Finally, we place our findings in the context of studies from other groups that have investigated the biases among rainfall products. Evaluating these studies, we can make several important conclusions that highlight (i) spatial and temporal variability in cumulative rainfall estimates due to the ‘choice of dataset used’, (ii) the need for more metrics to compare the difference in rainfall products - over and beyond the gauge-based time series and bias comparison of rainfall estimates, and (iii) impacts of the choice of the product on surface runoff and flood potential.
To the last point, our study highlights that rainfall point comparisons are useful but can lead to incorrect conclusions about the timing and amount of rainfall during a TC event. Thus, point or grid QPEs need to be supplemented with precipitation flux estimates rather than rainfall accumulation comparisons. More conclusions related to hydrology impacts from different gridded precipitations are evolving while the cases of multiple storms are being analysed, that will be presented during the meeting.
- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner