Evolving Adaptive Hydrologic Design and Water Resources Management in A Changing Climate: Experiences from the U.S
Ramesh S. V. Teegavarapu
Florida Atlantic University
The design of hydrologic or hydraulic infrastructure required for the management of available water resources is based on precipitation as the major input to simulation models. A hydrologic design built on precipitation frequency analysis (PFA) under stationary climate is no longer valid considering the observed changes in the frequency and magnitude of precipitation resulting from different storm events in many regions of the world. Recent studies (Cheng and AghaKouchak, 2014; De Paola et al., 2014; Wright et al., 2019; Teegavarapu, 2013; ball et al., 2018; Yadanfar and Sharma, 2015) reinforce this sentiment that the existing infrastructure in many parts of the world including the U.S. is inadequate to cope with extreme hydroclimatic events mainly attributed to changing climate. Wright et al. (2019) confirm that rainfalls exceeding commonly used Intensity Design Frequency (IDF)-based design standards are becoming more frequent over the conterminous United States (CONUS) since 1950, especially in the eastern region of the U.S. Precipitation extremes both in terms of magnitude and frequency that contribute to catastrophic floods are evaluated by many studies (e.g., Madsen and Wilcox, 2012) around the world to understand the influence of climate change and variability on these events.
The Intergovernmental Panel on Climate Change (IPCC) has developed the Climate Change 2014 Synthesis Report (fifth assessment report). This report indicates that
There are likely more land regions where the number of heavy precipitation events has increased than where it has decreased. The frequency and intensity of heavy precipitation events have likely increased in North America and Europe. In other continents, confidence in trends is at most medium. It is very likely that global near-surface and tropospheric air specific humidity has increased since the 1970s. Inland regions where observational coverage is sufficient for assessment, there is medium confidence that anthropogenic forcing has contributed to a global-scale intensification of heavy precipitation over the second half of the 20th century.
Evidence of changes in the global hydrologic cycle as suggested by Huntington (2006) was attributed to the ongoing intensification of the water cycle and increased precipitation in the past few decades (Wentz et al., 2007). Lohmann (2008) indicates that anthropogenic influences have been shown to have both increasing and decreasing effects on convective precipitation. Climate-induced changes in precipitation intensity and rain-snow partitioning are two critical issues affecting the future water resources management decision-making process (Nicholls et ah, 2008). The IPCC has concluded that increases in the amount of precipitation are very likely in the high latitudes, while decreases are likely in most subtropical land regions (IPCC, 2007). Climate change models suggest an increase in global average annual precipitation during the 21st century, although changes in precipitation may vary from one region to another (USEPA, 2008). Min et al. (2011) showed that human-induced increases in greenhouse gases have contributed to the observed intensification of heavy precipitation events. They indicate changes in extreme precipitation projected by models, and thus the impacts of future changes in extreme precipitation are underestimated as models seem to underestimate the observed increase in heavy precipitation with warming.
Precipitation Extremes and Stormwater Infrastructure Design
Precipitation is one of the critical inputs to a single or a continuous event simulation model that is used for hydrologic design. While traditional hydrologic design approaches continue to rely on the assumption of stationarity (Milly et al., 2008), new approaches that address the issue of considering uncertainty in the projected future precipitation extremes have been developed in the past two decades. Natural climate variability and future long-term changes in climate may alter precipitation intensities or durations of the storm events, and the gap between the events would ultimately influence the design of stormwater infrastructure under strict stationarity assumption (Teegavarapu, 2013). The influence of the Atlantic multidecadal oscillation (AMO), a major coupled oceanic- atmospheric oscillation on extreme precipitation depths in the state of Florida for a given return period, was recently reported by Teegavarapu et al. (2013) and Teegavarapu (2018).
The inabilities of climate change models in reproducing precipitation extremes accurately and the limitations of downscaling models in replicating the spatial and temporal variability of the same are discussed in different studies (Benestad et al., 2008; Goly and Teegavarapu 2012; Teegavarapu, 2013; Teegavarapu and Goly, 2018; Goly and Teegavarapu, 2020). Highlighting one of the issues related to downscaling, Fowler et al. (2007), Teegavarapu (2013), and Goly et al. (2014) have concluded that temperature can be downscaled with more skill than precipitation. The emission scenarios and limited skills of multi-model general circulation model (GCM)-based projections of the future considered primary sources of uncertainty, respectively (Jacob and Hurk, 2009). Stainforth et al. (2007) point to many sources of uncertainties in the models, including forcing uncertainty, initial condition uncertainty, and climate modeling uncertainties.
Stormwater Infrastructure Design
Most of the stormwater control and management facilities are designed based on single-event design-storms of a given frequency (ASCE, 1996; Haan et al., 1994; Liew et al., 2013; Teegavarapu, 2013; Jacob, 2013). In all single event-based design methods, peak discharges are sought for design and losses or abstractions due to evaporation and evapotranspiration are often considered negligible (ASCE, 1996; Gironas et al., 2009) and are not incorporated in the calculation of runoff. This is mainly due to the short duration of the events (e.g., one day or three days). Future temperature changes are expected to influence runoff volumes derived from continuous event-based hydrologic simulation models that consider all abstractions including evaporation and evapotran- spiration. The starting point of any method should assume that GCM-based climate change scenario-driven simulations and downscaled precipitation data are available. Storm runoff also depends on temperature changes. However, as the design of stormwater management systems is based on short-duration single events (design rainfall events) and peak discharges, influences of temperature changes on evaporation and evapotran- spiration are often considered negligible during these events. Single-event simulations are usually insensitive to the evaporation rates, and therefore evaporation is typically neglected when a single rainfall event or a synthetic storm is simulated (Gironas et al., 2009). Design storms derived using intensity-duration-frequency (IDF) relationships (Adams and Howard, 1986) are widely employed for the design of stormwater conveyance systems in engineering practice. Watt and Marsalek (2013) and Teegavarapu (2019) suggest that design practices need to be revised by adopting a comprehensive approach considering all design storm event characteristics and their sensitivity to climate change and inherent uncertainties in the existing IDF relationships. Changes in IDF relationships based on natural cycles of climate variability were recently documented by Teegavarapu et al. (2013), thus confirming the influence of climate variability on regional precipitation extremes with critical implications for hydrologic design.
Evaluations of revised urban drainage design practices are carried out by incorporating climate change factors (Arnbjerg-Nielsen, 2012), analyzing trends in precipitation extremes and their frequencies (De Toffol et ah, 2009), evaluating impacts of changing extremes using downscaled precipitation data from GCMs (Grum et ah, 2006), and designing frameworks for risk and uncertainty management (Arnbjerg-Nielsen, 2011). Traditionally, the IDF curve development for hydrologic design relied on the available historical precipitation data under the assumption of stationarity (Teegavarapu et ah, 2019). However, methodologies for the development of IDF relationships under changing climate were provided in several recent studies (De Paola et ah, 2014; Peck et ah, 2012; Liew et ah, 2013; Solaiman and Simonovic, 2011; Simonovic and Peck, 2009; Zhu et ah, 2012). Peck et ah (2012) used a non-parametric weather generator (WG) with shuffling and perturbation mechanisms to generate synthetic rainfall records similar (but not identical) to the observed historical record. In their study, two climate scenarios were selected with lower and upper bounds, the former scenario uses the observed rainfall record as WG input, and the latter modifies the observed record using the results of a selected global climate model and then uses the modified rainfall datasets as WG inputs. Liew et ah (2013) used a downscaling-comparison-derivation (DCD) approach to derive IDF curves for present and future climates using the extracted dynamically downscaled data of an un-gauged site with simulation data from a regional climate model (RCM). Solaiman and Simonovic (2011) used possible realizations of future climate from 29 scenarios derived from downscaled daily precipitation data and a non-parametric kernel estimation approach to develop IDF curves for different durations. The daily precipitation data are disaggregated to sub-daily durations using a non-parametric nearest-neighbor based disaggregation scheme. As evident from the review of recent literature, IDF curves for future climate can be developed along with upper and lower bounds on the future precipitation intensities. These revised IDF curves can be used for hydrologic design.
Adaptive Hydrologic Design
One adaptive approach for stormwater management is to upgrade the existing infrastructure with revisions of hydrologic designs considering changing IDF relationships over time and evolving temporal precipitation distributions and characteristics. Alternatively, a sustainable climate change-sensitive hydrologic design that identifies a compromise between current and future climates based on future climate projections can be devised. Preferences of hydrologists, water managers, or design personnel towards uncertain future climate change projections can be included in such a design process (Teegavarapu, 2013). Groves et al. (2008) explored different hypotheses about how different characterizations of climate change uncertainty may affect water managers’ beliefs, opinions, choices, and actions. Acknowledging the limitations of the climate change models, hydrologists may develop perceptions towards the accuracy of results from these models and assign importance to a specific model for a region for hydrologic design. These perceptions attached to future climate based on different model results can be translated to preferences using fuzzy membership functions (Teegavarapu and Simonovic, 1999) towards predicted future changes in the main hydrological inputs (Teegavarapu, 2013). Several studies in the recent past have provided methodologies to develop IDF curves based on GCM simulations and upper and lower limits of precipitation extremes obtained from downscaled GCM simulations considering multi-model multiple scenario uncertainties. The starting point of any study is to use these bounds on the precipitation intensities and solve a mathematical programming formulation with an objective of modeling the preferences of decision-makers towards future precipitation intensities.