| Περίληψη: | The EU Water Framework Directive 2000/60/EC (WFD) has set as necessity the
formulation and implementation of Integrated River Basin Management (IRBM) plans for
all EU member states. The backbone of river basin management is the monitoring of
qualitative and quantitative river characteristics. A common difficulty in many rivers is the
absence of permanent measurement equipment combined with low financial means and
time restrictions for implementing monitoring programs. Alternatively, quick measurement
methods of low cost and reliability (e.g. floats, air bubbles release) could be employed to
estimate river discharges. Moreover, river basins are exposed to a plethora of
environmental stresses, resulting in degradation of their quantitative and qualitative status.
This led to the reduction of the availability of clean water as well as to increasing
competition among water users. It has given rise to the need for optimal water allocation
for each river unit. In most countries water resources management is scourged by high
uncertainty and by imprecise and limited data, which may be easier approximated through
estimates of intervals. Due to these difficulties and complexities the need to adapt and
apply optimal water allocation methodologies under uncertainty has arisen.
The present PhD research is focused on two main challenges of river basin
management. The first part aims to propose and to develop the conceptual, mathematical
and computational framework of an original correction technique of quick river discharge
measurements in ungauged rivers. A methodological framework is developed based on the
principles of volume and pollutant mass conservation, considering intermittent nonmeasurable
latent quantities. Parallel measurements of discharge and natural tracers for
representative cross-sections of a river and its tributaries are required. The water volume
conservation is combined with pollutant/tracers mass balance expressed synchronously not
only for each single node of a river, but also for all possible multiple-node combinations
covering the entire river. This “divide and conquer” process relies on linear optimization.
According to the WFD the river monitoring programs should determine apart from the
level of predefined pollutants also their mass load. Discharge data are essential for the
estimation of loads of sediments or chemical pollutants of a river or stream. Therefore, the
proposed methodology enables the estimation of river discharges with higher accuracy and
reliability compared to the initial discharge estimates. Subsequently, it enables the
estimation of more reliable pollution loads. It intends to decrease duration, work force and
xlii
expense of river monitoring programs along with their management plans.
The suggested methodology was successfully implemented to the Alfeios river in
Greece including tributaries, where only limited short-term quantitative and qualitative
measurement data are available. It enabled the estimation of: (a) corrected discharges,
pollutant concentrations and pollution loads for eight combinations of initial values as
estimated from the qualitative analysis of the river basin, (b) a best/worst case (Min/Max)
interval and the corresponding error of the computed/optimized river discharges, pollutant
concentrations and pollution loads for the cross-sections of the main river and its
tributaries, where tracer concentrations were measured, and (c) the unknown not directly
measured parameters, including latent flow rate, the correspondin pollutant concentrations
and pollution loads of each river node.
Based on these results the methodology succeeded in restricting the errors of the
corrected mean discharge values of all measured cross-sections. The resulting error of the
corrected latent discharges is much wider compared to the corresponding error for the
measured cross-sections. However, it is of note that the determination of a hypothetical
unknown latent discharge and subsequently the correction of its estimation, even if it is
relatively inaccurate, is very important and useful, since the direct measurement of the
latent discharge, and generally of the assumed latent terms, is impossible. Besides, it is
worth underscoring that the combination of the single-node balances together with all
possible multiple-node balances based on the previous findings resulted in a considerable
reduction of the river discharge interval of the ensemble of the cross-sections for the
Alfeios river.
The direct confirmation of the corrected river discharges with simultaneous accurate
measurements is hampered by the lack of such precise measurements. Thus, the
consistency of the proposed methodology based on the linear form of the optimization
problem was compared with the results from the nonlinear model and the following
conclusion can be extracted. Firstly, the value ranges of the nonlinear model lie into similar
but not exactly the same value region as the ranges of the linear model. Both ranges have a
wide common value region, whereas the ranges from the nonlinear model are wider.
Secondly, by comparing each type of model (linear and nonlinear) with the measurements,
for both of them the measured discharge values and the corrected ones are linearly
connected. More precisely, through the t-test statistics it is proven that the results from the
linear as well as from the nonlinear model are not overestimated or underestimated based
on the measurements. This result confirms the consistency of the resulting solutions from
the optimization process with the measurements.
The second part of the present PhD thesis aims at proposing a decision support (DS)
framework for optimal water allocation under uncertain system conditions in a real and
complex multi-tributary and multi-period water resources system, and more precisely in
the Alfeios River Basin. Firstly, an inexact two-stage stochastic programming technique
(ITSP) with deterministic-boundary intervals (Huang and Loucks, 2000) and secondly, a
similar in terms of concept but more sophisticated and advanced methodology (FBISP) (Li
et al., 2009) embody the core of the proposed DS frame. Both hybrid methods are based on
the concept that in real-world problems, some uncertainties may indeed exist as ambiguous
intervals, since planners and engineers may not have enough information and data to
specify probability distributions, and therefore, find it much easier and realistic to define
fluctuation ranges for these uncertainties.
The ITSP method combines ordinary two-stage stochastic programming with
uncertainties expressed as deterministic boundary intervals and is simpler and easier to
follow up compared to FBISP. Stable intervals for optimized water allocation targets and
probabilistic water allocation volumes and shortages are estimated under a baseline
scenario and four water and agricultural policy future scenarios. On the other hand, the
FSBIP methodology combines an ordinary multi-stage stochastic programming with
uncertainties expressed as fuzzy-boundary intervals.
Upper- and lower-bound solution intervals for optimized water allocation targets and
probabilistic water allocation volumes and shortages are also estimated under the same
baseline scenario and future scenarios for an optimistic and a pessimistic attitude of the
decision makers. In both methods the uncertainty of the random water inflows is
incorporated through the simultaneous generation of stochastic equal-probability
hydrologic scenarios at various inflow positions, instead of using the scenario-tree
approach which is commonly used in most applications of these methodologies. The
comparison of the corresponding results of the FBISP method with that of ITSP revealed
that the results are consistent and compatible. In addition, the incorporation of the fuzzy
nature of the uncertainties in the FBISP results in a more analytic and fine approximation
of the effect of the uncertainties on the minimum and maximum values of the boundaries
of the results, providing also a more complicated structure of the results.
|
|---|
