This project looked at the challenges and advances in design of infrastructure for floods under non-stationarity. Based on more than 300 references, this review covers i) the potential sources of non-stationarity in time series of floods, ii) the methods for estimating design floods that rely on the stationary assumption, iii) the methods for estimating design floods that assume non-stationarity resulting from climate change and i) discussion on the current design methodologies in view of the pervasive uncertainties and strategies to manage the consequences of those uncertainties.
This project explored the use of climate-informed parameter models for assessing future precipitation and streamflow extremes. Rather than using projections of local weather variables, such as precipitation variable, climate-informed approaches use large-scale climate variables (e.g., El Nino-Southern Oscillation [ENSO]) to condition the parameters of the models. This project analyses the performance of Global Circulation Models (GCMs) in reproducing large-scale climate variables leading to precipitation extremes. The project team proposes a general methodology to set up climate-informed regional models for streamflow extremes and drive climate projections projected large-scale climate variables. The Ohio River basin was the main focus.
The role snow variable plays in flooding in the Upper Missouri basin was looked at during this project. More specifically, assessing the quality of the available datasets for temperature, precipitation and snow water equivalent variables across the Upper Missouri Basin was explored. Climate projections for the region are then discussed across the basin. The consequences of the changing snowpack across the Upper Missouri Basin on floods were investigated. A comparison between hydrological simulations and data-driven models (i.e., artificial neural networks) is conducted in order to better understand the role of snowpack in flooding in the region. Results show that uncertainty in snow water equivalent may affect significantly detection of change in streamflow extremes.
The drivers of precipitation in the East-South-Central U.S during the cool season to advance understanding of the conditions that could lead to major flooding in the area were examined. The study especially focuses on the different sources of bias in precipitation that are introduced by the regional climate models (RCMs). The main sources of bias in this region are linked to moisture flux into the region, transient, synoptic-scale low-pressure systems, Gulf of Mexico and Caribbean seas surface temperatures (SSTs), and ENSO-related teleconnections. Following these results, caution should be taken when using RCM projections to analyze hydro meteorological extremes.
The project then extended the analysis of precipitation and streamflow extremes to the scale of the continental U.S. The same set of nine catchments that covers rather well the range of climate conditions in continental U.S. A straightforward approach to define climate factors that account for climate change uncertainties is proposed. The suggested approach combines stationary statistical model for precipitation extremes with a sensitivity analyses regarding the model parameters of the precipitation distribution. Climate projections from GCMs and RCMs are used ex-post to infer the potential changes in precipitation distribution parameters. Climate factors can be chosen to provide satisficing levels of protection for a chosen range of uncertainty from climate projections. The project team then provides a comparison study across different climate areas and different models to predict change in streamflow extremes. The test-bed of models includes stationary approaches, climate-informed and trend-informed statistical models and hydrological simulations. Results show that no method performs better than others perform and thus suggests using several approaches for flood design.
The project also investigates the predictability of short- and long-term horizons for climate extremes in Ohio and Mississippi River Basins. Results show that both scales provide two distinct pieces of information with crucial implications in the management of water and crucial infrastructure systems in the region.
The project focuses on decision-analysis and robust adaptation regarding risk from hydrological extremes. The project looked at an application of the Decision Scaling approach for which the climate stress test is conducted using large-scale variable such as soil moisture across the Ohio River Basin and the seas surface temperature in the Pacific (e.g., ENSO). The considered case study for the application is Louisville, Kentucky. The study discusses some of the benefits and limitations of climate-informed stress test highlights areas of future research. The project presents a set of stylized experiments to assess the uncertainties and biases involved in estimating future climate risk over a finite future period, given a limited observational record. Results suggest that shorter design lives are preferred for situations where inter-annual to decadal variability can be successfully identified and predicted, suggesting the importance of sequential investment strategies for adaptation.