Making multi-stakeholder water resources decisions with limited streamflow information
streamflow information makes water resource decisions more challenging. Moreover,
modeling choices for addressing streamflow uncertainty may benefit some stakeholders
more than others. This dissertation offers three relatively parsimonious approaches for
screening multi-stakeholder water resources decisions, all of which account for
streamflow uncertainty. While these approa... read moreches are applicable to many water resources
problems, I present applications that (i) characterize changes in flood variability,
(ii) identify long-term floodplain management plans robust to uncertain climate change,
and (iii) evaluate tradeoffs between reservoir storage benefits and riverine ecosystem
conservation. First, I present a parsimonious two-stage ordinary least squares (OLS)
regression model, which detects trends in both the central tendency and variability of
annual floods. Neglecting increasing (decreasing) trends in the coefficient of variation
at stations with concurrent increases in the mean annual flood can lead to the
substantial under-design (over-design) of flood protection strategies. This OLS
regression modeling approach also provides trend uncertainty estimates (type I and II
errors) useful for decision-making. Second, in some cases, there may be insufficient
information for modeling changes in floods probabilistically, especially when
characterizing impacts of future climate change using different General Circulation
Models (GCMs). Under these circumstances, minimax regret optimization models can still
identify adaptation plans that minimize both the over- and under-design consequences of
climate change uncertainty. I introduce a mixed integer programming (MIP) optimization
model identifies sequences for implementing river reach- and property-scale adaptation
measures that are robust to uncertain changes in floods during the 21st century. Third,
attenuating hydrologic extremes beyond critical thresholds can also harm ecosystems and
communities adapted to them. In settings with short pre- and post-dam records, it is
often difficult to rule out that these changes are, in fact, due to dam operations and
not random variability. Hypothesis tests offer information about (i) the likelihood of
losing hydropower production by unnecessarily changing dam operations and (ii) the
likelihood of adverse ecological impacts by not changing them when necessary. Moreover,
this probabilistic information can be incorporated into a Bayesian decision-making
framework that evaluates the potential hydropower and ecosystem consequences of
different operating rules.
Thesis (Ph.D.)--Tufts University, 2017.
Submitted to the Dept. of Civil Engineering.
Advisor: Richard Vogel.
Committee: Paul Kirshen, James Limbrunner, and Charles Kroll.
Keywords: Water resources management, Hydrologic sciences, and Civil engineering.read less