Defining Robustness, Vulnerabilities, and Consequential Scenarios for Diverse Stakeholder Interests in Institutionally Complex River Basins

Abstract

The Upper Basin of the Colorado River in the southwestern United States supports municipal, industrial, agricultural, and recreational activities worth an estimated $300 billion/year within the state of Colorado alone. The allocation of water to these activities is fundamentally shaped by water rights that in turn distribute risks among a diverse suite of sectors and stakeholders. In this study, we assess the vulnerabilities faced by the hierarchy of hundreds of water users in the portion of the Upper Basin within the state of Colorado, as a result of changes in hydrologic extremes, demand growth, and institutional and physical infrastructure in the basin. We also determine how robust the different users are to these potential changes, examining how their sensitivity to these factors depends on the magnitude and frequency of shortage they are unwilling to exceed. This advances previous robustness evaluation methods by formalizing an a posteriori exploration of alternative definitions of robustness tailored to each user's unique context. Our analysis reveals that robustness varies significantly not only across users but also across different definitions of acceptable performance for each user. We further show the importance of using scenario discovery to evaluate how the factors that drive consequential scenarios differ depending on the definition of robustness. Our results highlight the need for robustness and vulnerability frameworks to avoid broad categorical aggregations of stakeholder groups, to carefully capture complex institutional dependencies (e.g., water rights), and to facilitate a more transparent illustration of the implications of alternative definitions of robustness for specific stakeholders.

Can Exploratory Modeling of Water Scarcity Vulnerabilities and Robustness Be Scenario Neutral?

“…This user is analyzed in the SI. As highlighted previously by Hadjimichael, Quinn, Wilson, et al (2020), despite modest changes in right seniority and no change in sector, two users in the same basin can experience starkly different shortage distributions across historical and alternative hydrologic conditions. Here we see that these differences also depend on the experimental design.…”

“…All rights reserved. manuscript submitted to Earth's Future or exceed climatic influences on vulnerability (Herman et al, 2014;Hadjimichael, Quinn, Wilson, et al, 2020). Yet such assessments rarely consider the correlation between human decisions and the climate, when correlations were shown in this study to significantly influence which factors and interactions were most important.…”

“…In this study, we explore how vulnerability assessments performed over competing hypotheses of how future hydrology might evolve dictate which uncertainties are found to most control water shortages for different users in an institutionally complex, multiactor system, and subsequently, which users are found to be most robust. Several studies have compared how robustness ranks of alternative management strategies or multiple water users (i.e., policies and objectives) differ under alternative definitions of robustness (Herman et al, 2015;Giuliani & Castelletti, 2016;Spence & Brown, 2018;McPhail et al, 2018;Hadjimichael, Quinn, Wilson, et al, 2020), or under alternative assumptions about the range and joint distribution of uncertain factors (i.e., the experimental design) (Moody & Brown, 2013;Taner et al, 2019;Reis & Shortridge, 2019). Yet none of these studies has explored if and how the importance of uncertain factors differs under alternative experimental designs.…”

“…This shortage level is displayed on the y-axis as a fraction of the user's demand.Fig. 9(e) displays a satisficing surface introduced byHadjimichael, Quinn, Wilson, et al (2020) illustrating the percent of realizations in the All-Encompassing experiment meeting a range of such possible success definitions (see SIFig. S19-S20 for how this surface differs depending on the experimental design and number of Latin hypercube samples).…”

Author response to referee comment #1

“…In the revised version, we will shorten the material and method section by focusing on fuzzy thresholds combined with one generating method: the logistic regression. This choice is motivated by the fact that this approach has received a lot of attention recently in the literature (in addition to Kim et al, 2019; the paper will cite Quinn et al, 2018, Lamontagne et al, 2019, Hadjimichael et al, 2020Marcos-Garcia et al, 2020).…”