Economic implications of climate change for

Economic implications of climate change for infrastructure planning in transboundary water systems: an example from the Blue Nile

Marc Jeuland1

1

Sanford School of Public Policy; Box 90239 Duke University Durham, NC, 27708, USA Email: marc.jeuland@duke.edu

Abstract This research develops a hydro-economic modeling framework for integrating climate change impacts into the standard water resources planning problem. It then illustrates use of that framework through the evaluation of a potential hydropower project along the Blue Nile in Ethiopia. Storing water in a Blue Nile reservoir provide an interesting case for testing this integrated approach because such a project would induce a number of physical and economic changes – both transboundary and climate-dependent. The proposed framework makes two contributions to the existing literature on water resources project appraisal. First, it demonstrates how routinely-used hydrological modeling techniques can be supplemented with Monte Carlo simulation to include the economic risks inherent in the planning problem, in addition to its more commonly considered physical dimensions. Second, it demonstrates how analysts can include a number of linkages between climate change, hydrology and economic production in conventional planning models to develop better understanding of the complexities and important uncertainties associated with future conditions. While the results for any given application of the framework will be highly context-specific, the general approach developed here can be implemented in a variety of settings, to evaluate changes in both design and management of water resources systems. Key words: Water resources planning, cost-benefit analysis, dams, climate change, transboundary waters, Nile river basin

Climate change implications for water resources infrastructure planning in transboundary water systems: an example from the Blue Nile

1. Introduction According to the Intergovernmental Panel on Climate Change, climate change is likely to have a complex set of impacts on water resources throughout the world [IPCC, 2007b]. Warming attributable to rising atmospheric concentrations of greenhouse gases will affect ocean and surface temperatures, precipitation patterns, evapotranspiration rates and the demand for water in agriculture, the frequency and intensity of storms, the timing and magnitude of runoff, and sea level in coastal communities [Frederick and Major, 1997; IPCC, 2007a]. There has been fairly extensive research and sophisticated modeling aimed at assessing the magnitude of such impacts under different emissions scenarios in the global context and in specific regions [Arnell, 2004; Leavesley, 1999; Milly et al., 2005; Vicuna and Dracup, 2007]. Still, there is considerable debate and uncertainty about regional impacts. To date, little practical research and guidance exists on how the range of possible effects of global warming should be integrated into planning for new capital-intensive investments in water resources, as well as renewal of old infrastructures. This research develops a hydro-economic modeling framework for integrating climate change into the traditional water resources planning problem. The framework allows consideration of a number of the effects of climate change on water resources systems. Impacts can enter the framework at two levels – hydrological and economic. As such, they can better be considered simultaneously. The first level applies to the traditional river basin planning tool: the hydrological routing model. At this level, the framework includes explicit linkages between

climatic factors (such as temperature and precipitation) – specified based on General Circulation Model, Regional Climate Model, or other relevant future projections – and many of the hydrological model components that these factors influence. Such components include relationships with runoff, net evaporation from storage structures and lakes, water demands, flood and drought risks throughout the system, and basic streamflow routing equations,. The proposed framework thus allows for comparison of system outputs in physical terms under historical and changed conditions. The second, economic level of the framework adds emphasis to economic uncertainty, and aims to tackle the issue of relative changes in the real value and productivity of the goods and services generated by hydrological systems. Like the physical outputs of the system, their economic value is also likely to be influenced by changing climate, a fact that has not been widely acknowledged in the water resources planning literature on climate change. Large-scale changes in temperature, precipitation, and sea-level rise, for instance, are likely to influence the demand for water and hydropower, via at least two types of changes. First, the economics of the production processes using these inputs may be altered. For example, farmers may increase or decrease their use of irrigation based on changes in temperature and rainfall. In supply-constrained systems, this could lead to increases in the economic value of water. Also, energy suppliers and society may favor hydroelectric power generation to conventional fossil sources if carbon offsets gain value or if such sources become relatively more costly. Second, consumption patterns of resources will be altered. For example, warmer temperatures may increase the demand for cooling energy while decreasing heating requirements. Sea level rise may increase the demand for environmental releases in river basins threatened with salt water intrusion. Indeed, relative changes in the value derived from resources affected by climate

change play an important role in determining its economic consequences [Sterner and Persson, 2008]. While it is very difficult to predict relative price changes over the course of a long project lifespan, it is important to acknowledge that these effects could compensate for or exacerbate the physical effects of climate change. They could also alter the trajectory of efficient, or even socially beneficial, policies and investments in infrastructure. The high level of uncertainty associated with these types of physical and economic changes argues for a new approach for integrating the economic valuation problem with typical approaches used to inform water resources project appraisal. Current planning methods, which have proven extremely useful in water resources project appraisal, were largely developed and extended from the early contributions of the Harvard Water Program (see for example Maass et al. [1962] and Hufschmidt and Fiering [1966]). The analyses usually rely on some form of economic optimization to allocate water to its highest value uses or to maximize net benefits resulting from investment in new infrastructures [Harou et al., 2009]. Planners using such tools often devote substantial attention to hydrological risks and uncertainties, but tend to treat economic factors (or cost and benefit curves) as fixed parameters (or functions) within the valuation equations for costs and benefits that are used to calculate net benefits of new projects. This paper is not the first to suggest some need for rethinking these planning methods. Critiques of the traditional approach have largely been motivated by a heightened concern over climate change, although it should be admitted that changing conditions defined more broadly have long presented problems for water resources specialists [Lettenmaier, 2008]. In effect, it is no coincidence that there is a large body of research demonstrating new modeling approaches for integrating climate change aspects into hydrological planning applications [Frederick et al., 1997; Lettenmaier et al., 1999; Stakhiv, 1998]. To date, however, this research has generally

been limited to the physical aspects of water resources systems, and provides only limited insights into the economic implications of changed conditions. Analyses limited to physical impacts may miss the possibility of important economic feedbacks and lead to erroneous conclusions of project viability. Though it may seem counterintuitive, a hydropower dam may actually become more valuable to society if river flows are reduced under climate change, if the value of that power increases or if the value of carbon offsets exceeds the power production that is lost. Similarly, the value of an irrigation dam may decrease when river flows increase if the value of storage decreases due to reduced water scarcity. The paper is organized as follows. Section 2 describes in basic terms the elements of the framework that is typically used in systems planning applications, explains some of its key assumptions, and offers a proposal for extending that assessment framework to better consider the uncertainty over future conditions. Section 3 specifies how this modified approach can be made operational, using the example of a specific water resources system (the Nile Basin). Section 4 describes the project used to illustrate the modified framework. Section 5 presents results pertaining to the influence of different components of climate change and project uncertainties on the economic outcomes of the example reservoir project described in Section 5. Section 6 concludes.

2. Comparing current water resources planning models with the proposed framework The traditional systems planning model We begin with a basic review of the textbook systems planning model that is commonly taught in graduate-level courses on water resources project evaluation and design [Loucks et al., 1981;

Loucks et al., 2005; Maass et al., 1962; ReVelle and McGarity, 1997]. 1 In its simplest form, the planning problem is set up as a deterministic optimization problem that treats hydrologic or economic parameters (or functions) as fixed and known quantities. Though the form of the objective function that is maximized (or minimized) varies, the most appropriate economic criterion is to maximize total economic net benefits derived from the infrastructure(s) in question. Other common formulations are to minimize investment costs subject to constraints on water supply reliability or firm hydropower generation capacity, or to maximize total benefits subject to cost or other constraints [ReVelle and McGarity, 1997]. A variety of studies using different objective functions can be found in the literature [Loucks et al., 1981]; a detailed summary is not presented here. Many of today’s water resources modeling packages – for example Riverware [Zagona et al., 2001] or WEAP [Yates et al., 2005] – can be parameterized to allow some form of economic optimization. Water resources planning textbooks acknowledge that economic optimization at best constitutes a partial analysis. The key shortcoming associated with optimization is the assumption of perfect foresight; water resources managers are allowed to know river flows with certainty, and to operate the system efficiently based on that knowledge, allocating water to its highest value uses [Harou et al., 2009]. More complete project planning (Figure 1) usually relies on studies that use operating rules developed using optimization to conduct repeated simulations that test the sensitivity of results to natural variability in river flows and future demands, or to other model and infrastructure design parameters. These planning studies generally assume that historical climate conditions provide sufficient information for reliably predicting future system behavior.

1

In reality there are a very large number of different variations of this basic planning model; the stylized version presented here is used for illustrative purposes.

The constraint structure of water resources optimization and simulation models is largely provided by system continuity equations, though other policy-relevant constraints may enter into the problem depending on the application. For storage points in a system (i.e. lakes and reservoirs), the continuity equations take the form of Equation 1 below (adapted from Loucks et al. [1981]). These equations are typically written for a time step of length t; all the variables in equation 1 are therefore water volumes that correspond to the same length of time t. Ss,t+1 = Ss,t + Qs,t - Rs,t - Es,t - Ls,t - Ds,t, where

(1)

Ss,t

Storage in the reservoir at site s and time t, [L3].

Qs,t

Inflow to the reservoir (or node) at site s and time t, [L3].

Rs,t

Outflow from site s at time t, [L3].

Es,t

Net evaporation losses from site s at time t, [L3].

Ls,t

Seepage (storage in groundwater) losses at site s at time t, [L3].

Ds,t

Water withdrawal (for consumptive use, or partial consumption and return to the system further downstream) from site s at time t, [L3].

For flow at modeled river nodes without storage infrastructures, the storage terms Ss,t and Ss,t-1 and the evaporation and seepage terms Es,t and Ls,t are usually set equal to zero, such that the constraint in equation 1 simplifies to: Qs,t = Rs,t + Ds,t.

(2)

Integrated water resources system models rely on a collection of such reservoir and flow nodes, which correspond to the configuration of the system in question and to the availability (for

calibration) of flow gauge and water demand data. The flows of water into and out of those nodes are calculated based on the general relationships presented in equations 1 and 2. Furthermore, inflows Qs,t are often represented, as in Cohon [1978], as the combination of outflows from the upstream site or node s-1 and the local increment to natural streamflow ΔFs,t (or local runoff) between sites s-1 and s: Qs,t = Rs-1,t + ΔFs,t = Rs-1,t + Fs,t - Fs-1,t , where Fs,t

(3)

Volume of flow measured at gauging station at site s over time increment t, [L3].

In this formulation, the flow increment ΔFs,t is determined by calculating the difference in flow (Fs,t - Fs-1,t) at adjacent nodes in the historical series. 2 This increment is not strictly equal to the local runoff into the river between nodes s-1 and s. Unless losses are explicitly considered, ΔFs,t includes evaporation and seepage losses between the nodes, and may include unmeasured consumptive use by people or industries located along the river. In practice, due to data limitations, it can be difficult to measure the various components that make up ΔFs,t. When the contribution of local runoff is important (from tributaries that join the river system between two nodes), modelers often back-calculate ΔFs,t from flow records at gauging stations in the system, an approach that leads to systematic bias in the models if these factors vary over time. When there are no important sources of local runoff, flow calibration is instead conducted with various fitting techniques applied to historical gauge records at nodes s-1 and s, for example lag regression. The observed flow in the historical record Fs,t at node s is then expressed as some combination of n lagged flow terms at nodes s and s-1 (where αj,t = the regression coefficient on flow at node j and time t) plus a constant term ks, for example: 2

It is also possible to use physical or statistical methods for predicting local inflows into a river system, using rainfall-runoff models, but these models must also be calibrated from historical flow data.

Fs,t = α s-1,t·Fs-1,t +…+ αs-1,t-n·Fs-1,t-n + α s,t-1·Fs,t-1 +…+ α s,t-n·Fs,t-n + ks.

(4)

The planning problem thus usually relies on analysis using models composed of nodes for which flows are determined using Equations 1-4. In the simplest application of these models, the planner relies on historical flow sequences to explore the value of system impacts (based on assumptions needed to monetize hydropower, water demand, and other such outputs) that result from adding new infrastructures or withdrawals to the system, and/or shifting water allocations or operating strategies within it. The analyst can use optimization to determine the values of storage, inflow, outflow and demand variables that maximize net economic benefits. In many cases, planners may have additional objectives to the net benefits criterion, which can be reflected by additional constraints. Most systems modelers then relax the assumption of perfect foresight, and conduct simulation using the historical record combined with realistic reservoir operating rules (often developed using optimization procedures, and modified in iterative fashion to respond to results in the sensitivity analysis). At best, the planner will supplement simulations based on the historical record with ones that use stochastic simulation, to quantify the risks posed by natural flow variability and associated with the various favored system designs and/or operating rules, and to refine those aspects. The literature is rich with application of such tools [Harou et al., 2009; Loucks et al., 1981; Yeh, 1985]. In addition, stochastic linear programming or dynamic programming techniques can be used to help guide design of operating rules, though the objective function of these models is however generally limited to physical objectives [Yeh, 1985]. For testing the sensitivity of economic outcomes, modelers tend to analyze “worst” and “best” case scenarios that correspond to pessimistic and optimistic assumptions about the future; these are considered to generate reasonable upper and lower bound indicators of the performance

of alternative investments. Probabilities are sometimes assigned to future scenarios, such that analysts can compute and consider the expected net benefits obtained from different project options. Challenges to use of the traditional planning model It has been argued that the textbook planning model depicted in Figure 1 is very infrequently used in real-world systems design [Rogers and Fiering, 1986] and that reservoir operating rules have only rarely been developed based on results obtained from systems analysis techniques [Yeh, 1985]. In spite of the emergence of the integrated water resource management paradigm [Bouwer, 2002], this reality does not appear to have changed considerably in the past twenty years, as evidenced Harou et al.’s recent review [2009]. More precisely, it is clear that hydrological routing models and general water resources modeling tools are widely used by engineers and planners around the world, but that systems optimization applications – particularly economic optimization – remain rare outside the academic world. This is somewhat surprising considering the number of such applications in the water resources literature. A noteworthy exception to the lack of use of systems techniques is the application of the CALVIN model for water resources planning applications in California [Jenkins et al., 2004; Tanaka et al., 2006]. Scholars offer a number of explanations for why systems optimization models have not been used more widely. Rogers and Fiering [1986] emphasize institutional resistance and contextual factors, whether in the US or developing countries. In the US, economic analysis of public investment projects really began in the field of water resources [Hanemann, 1992; Inter-Agency River Basin Committee, 1950], but systems planning techniques developed too late to influence

the construction of large infrastructures, such that the payoff from adopting them may now be considered small. Infrastructures were built piecemeal, integrated river basin plans were rarely developed, and the economic studies ostensibly used to guide implementation decisions only played a marginal role in the ultimate choices that were made [Eckstein, 1958; Krutilla and Eckstein, 1958]. For developing countries, key constraints to wider use appear to include insufficient high-quality data to inspire confidence in results, a lack of validation tools (because few models are complete enough to test the robustness of ‘optimal’ solutions), and a lack of stakeholder involvement in model development. Rogers and Fiering [1986] and Harou et al. [2009] discuss other technical and interpretation problems related to optimization models. In the context of this paper, three interrelated issues stand out: a) the challenge of how to deal with general planning uncertainties, some of which are very large, b) the insensitivity of many systems to wide variations in design alternatives, c) the existence of multiple near optimal alternatives, and d) the inability to easily represent social, political and/or environmental objectives and risk aversion preferences in the mathematical expressions of optimization models. These issues are related by the fact that uncertainty in optimization model parameters or the constraints considered relevant for planning can lead to different optimal infrastructure bundles depending on their assumed values. Multiple designs can often be justified, and optimization models, which are costly to develop, may have little to say about the relative strengths of different options. My own view is that optimization tools can provide extremely valuable information to the planning process for new water resources infrastructures. These models are particularly useful in narrowing the choice set of project alternatives, to allow a focus on specific regions or types of infrastructures that are likely to generate significant economic benefits. Applied to the example

considered further in this paper, the Eastern Nile, optimization tools have in fact been used in precisely this way [Whittington et al., 2005; Wu and Whittington, 2006]. That work succeeded in focusing the policy debate on possibilities for investment in the Ethiopian Blue Nile that might not otherwise have received as much consideration. Like other water resources modeling tools, however, optimization models cannot effectively be used in isolation, because of their considerable limitations in dealing with the large uncertainties inherent in the economic evaluation of long-lived, capital-intensive projects. Towards an integrated simulation-based approach: Assumptions and uncertainties in the systems planning model Besides the specific issues highlighted in the literature, I believe there are a number of additional important difficulties which create challenges for water resources planners, particularly in developing countries. Some of these limitations are: 1. The data used to simulate the continuity equations (especially evaporation, seepage, and water demand) tend to be approximate average values, due to measurement difficulties; 2. There may be a lack of confidence in the historical record of inflows or concerns over the long-term stability of model (“black box”) calibrations developed using the instrumental flow record; 3 3. The specification of Ds,t (future water demands) for planning purposes is very difficult, especially when considering long-lived infrastructures and the potential for large-scale economic changes; 3

A notable example is the Nile Basin system modeled later in this paper (see Hassan, 1981; Shahin, 1985; Davies and Walsh, 1997; Nicholson, 2001; Nicholson and Yin, 2001; Marchant and Hooghiemstra, 2004).

4. There could be systematic changes in the quantity, variability, and timing of runoff due to climate or land use changes, and these remain difficult to model despite progress in modeling (see Wood et al. [1997] or IPCC [2007a] for further discussion); 5. In addition to affecting runoff, climatic perturbations will interact with other parts of the system, altering physical properties such as net evaporation from reservoirs Es,t, demands for irrigation water Ds,t and reservoir storage Ss,t (via changes in siltation rates); 6. Existing management structures and regimes may also change in response to long-term perturbations in climate, with system-wide consequences; 7. The traditional appraisal model is not easily adaptable to account for the effect of economic uncertainties in project appraisal calculations, even though the net benefits of capital-intensive water resources projects are known to be highly sensitive to economic parameters such as the discount rate, estimates of infrastructure cost, the value of irrigation water, etc.; and 8. Some relevant costs and benefits may be difficult to monetize and be omitted from the objective function, such that analysts may find it difficult to interpret economic outcome (and may appeal to “intangible” or secondary benefits to justify projects ex ante). These various issues may be why ex post reviews of large water projects have often disagreed with the predictions of ex ante analyses. For example, the World Commission on Dams [2000] found that hydropower production from many large dams was often much lower than planners originally expected, and that assumptions about infrastructure costs were usually too low, both of which often led to overestimates of net benefits.

An illustration that shows how modeling assumptions enter into the system continuity equations may be helpful (Figure 2). This figure shows the terms from equation 1 that contribute to the flow balance calculations at a generic storage node. There are uncertainties associated with each of the terms Ss,t; Qs,t; Rs,t; Es,t; Ls,t; Ds,t,; and ΔFs,t. Only some of these uncertainties are of a physical nature; many others are socio-economic. Because complex surface water systems are composed of many such nodes, the uncertainties have the potential to build on each other and create substantial modeling biases’. The combined errors have the potential to undermine the economic analysis of new projects and/or management regimes. These effects are not unlike the “cascade of uncertainty” that has been associated more closely with climate change [Mearns et al., 2003]. The standard practice in economic evaluation of water resources projects is to make specific assumptions to parameterize the cost and benefit functions that appear in their objective functions, and to then conduct limited sensitivity analysis to explore how results change when these parameters are altered. One of the most peculiar features of this analytical approach is the fact that it only barely incorporates economic risks. Perturbation of individual model parameters, construction of a few best and worst case scenarios, and/or alteration of system constraints does not give sufficient attention to the combined effect that these uncertainties can have, and makes it difficult to choose between different sites and project design options. A proposed integrated hydro-economic simulation model This paper proposes to extend the standard project evaluation framework by using simulation methods to test the performance of different system configurations in various plausible states of the world. Such a simulation framework could be used to test the effect of natural variability and

climate change on the physical behavior of systems, as well as the economic uncertainties associated with changes in the physical system and the value of the physical outputs derived from it. It can also serve to identify the sources for the most important uncertainties in the planning problem, thereby informing the selection of more robust project and system designs. A schematized representation of such a framework is shown in Figure 3 (changes from the basic project evaluation framework are italicized). As shown, there are two levels of linked simulations – physical and economic. To use the framework, the analyst must construct future climate and demand scenarios, and choose the project(s) (or operational strategies) that are to be evaluated. 4 Next, a number of relationships between climate conditions and the system (henceforth referred to as ‘linkages’) must be specified, including changes in runoff, water demands, reservoir evaporation, etc., using theoretical or empirical relationships. 5 The physical behavior of the system, with or without new projects, can then be simulated under the changed conditions, and physical system outcomes can be determined. The second level of the evaluation framework combines the scenario-specific physical outcomes with cost-benefit calculations for the project(s) being studied, using Monte Carlo methods. These economic simulations also include theoretical or empirical linkages with climate factors, for instance alterations in the real economic value of the hydropower or irrigation water obtained 4

For example, climatic or other changes in a river basin may increase the need for flood control or other priorities. In practice, the ability to adapt will be constrained by the speed with which real changes are detected, and the degree

to which new investments have intrinsic flexibility – through design and/or operational features – to respond to shifting future conditions. 5

Net evaporation Es,t from reservoirs will be affected by temperature increases and altered precipitation patterns. Consumptive demands Ds,t will be affected by changes in crop-water requirements due to changed temperatures, precipitation, and carbon fertilization, and the extent to which farmers adapt to climate change (Mendelsohn et al., 1994; Schlenker et al., 2006). Seepage rates from storage reservoirs Ls,t, storage variables Ss,t and reservoir operating rules could change as a result of larger impacts on the system water balance or changes in sediment deposition (due to different timing and magnitude of runoff).

from the system, as well as the value of carbon emissions or offsets. Use of Monte Carlo methods to conduct these economic simulations allows the project analyst to simultaneously include both physical and economic uncertainties associated with climate change and the projects being evaluated, and to understand the factors that are most important to changing project outcomes under different plausible future scenarios.

3. Example Application: Using the Hydro-Economic Framework This section describes in more detail how the general framework described above is made operational, to evaluate different components of the effects of climate change on an infrastructure example in the Nile Basin. The analytical procedure. At a generic, non-basin specific level, the simulation framework can be organized into eight steps and three component models (Figure 4). Steps 1 through 5 occur within the hydrological portion of the framework; steps 6 through 8 make up its economic portion: 1. a) Define k future scenarios [1,…,K] and p+1 infrastructure experiments [0,…,P], where p=0 is the “no project” baseline configuration; b) Choose the first scenario (k=1) and the “no project” configuration (p=0); 2. Specify climatic and economic parameter values and linkages for scenario k (for example theoretical and/or empirical relationships governing net evaporation from reservoirs, water demands, changes in routing, economic value of system outputs, etc.);

3. Use a streamflow generation model to create n sequences of synthetic inflows [1,…,N] corresponding to the runoff characteristics for the chosen scenario k, each of which is at least as long as the potential lifespan for the project(s) being assessed; 4. Use a hydrological model to conduct simulation with each of the inflow sequences to generate N series’ of physical output measures for the system performance over the relevant time horizon (alternatively, use only a single historical series for analysis); 5. Store the physical outputs from the system; then, if p = 0, skip to step 7; otherwise calculate incremental changes in the outputs due to new projects relative to the “no project” baseline; 6. Determine economic impacts of project p in climate scenario k using the sequences of incremental measures from step 5 as inputs to a Monte Carlo simulation model for calculating project NPV; 7. If k < K and p < P, choose the next future scenario k + 1, or if k = K and p < P, set k = 1 and choose the next project p + 1 and repeat steps 1-6 (otherwise go to step 8); 8. When all projects P have been assessed in all scenarios K, analyze results and compare the performance of alternative plans across multiple scenarios with the aid of decision rules. Component models. Three models, described below, are needed to make the simulation framework operational for a given application. Details about the calibration and parameterization of these individual models for the Nile case are available from the authors upon request. A stochastic streamflow generator: This model is used in step 3 to generate inflow sequences with historical or perturbed characteristics for input to the hydrological simulation model of step

4. It allows the analyst to generate many years of synthetic data that have similar statistical properties (or properties perturbed according to climate-change projections) to the flows observed in the historical hydrological series for the system of interest. The techniques used, based on various types of autoregressive models with random error components of known standard error, are well known in the field of hydrology, see for example Fiering and Jackson [1971] for the basic theory, or Bras and Rodriguez-Iturbe [1993] for more up-to-date developments. Stakhiv [1998] provides an example of how these models can be used to assess scenarios of climate change. For the application described in this paper, 10,000 years of monthly inflows were generated at 11 inflow nodes to the Nile system (depicted by the inflow arrows in Figure 5) for the two climatic conditions that were considered. A hydrological simulation model: The second model describes how inflows and water enter and move through the water resources system, with and without new projects, and under changing climate conditions. The structure of this model and much of its calibration is similar to the NileDecision Support Tool, or Nile-DST, developed at the Georgia Water Resources Institute [Yao and Georgakakos, 2003]. The inputs to the model include historical or synthetically-generated inflows, water withdrawals corresponding to modeled demand scenarios, climate variables that influence the physical performance of the water system, and project-specific design and operational attributes. This model is thus composed of nodes (Figure 5), and water flows are described by continuity equations of the type described in equations 1-4. Climate linkages (description below) built into the model describe how the climatic characteristics of a given scenario affect these continuity equations. For the application explored in this paper, each of one hundred simulations for each climate condition is conducted using a single 100-year inflow series, taken from the stochastic flows generated in step 3. Each simulation thus results in a

unique 100-year set of system outputs, with and without the dam project being evaluated. The hundred output sequences can then be analyzed to obtain the physical reliability measures that are often of interest to water resources engineers. Incremental changes in water demands met, hydropower produced in the system, and peak flood flows due to the project are calculated, and these are then used in the economic analysis of the project. An economic appraisal model: The third model of the framework takes the sequences of incremental project outputs obtained from the hydrological model for each climate scenario to conduct simulations of the economic performance of the infrastructure project being assessed. Using Monte Carlo methods, the model samples randomly from these stored sequences of outputs (each of which is assumed equally likely) and from specified distributions of uncertain economic parameters in the valuation equations. In our application, 10,000 random draws are taken from the 100 sequences of incremental outputs and the economic parameter distributions. The model then aggregates costs and benefits to compute project net present value (NPV): NPV =

∑

T t =0

( Bt − C t ) ⋅ r (t ), where

NPV

Sum of discounted net benefits, in monetary units.

Bt

Sum of benefits in year t, in monetary units.

Ct

Sum of costs in year t, in monetary units.

r(t)

Discount factor in year t, assumed to be 1/(1+δ)t, representing a constant

(5)

exponential discount factor with discount rate δ The 10,000 economic simulations yield a distribution of economic outcomes for the infrastructure option being considered, conditional on the natural variability and economic

parameters specified in the scenario of interest. Because new projects are added to the system at the beginning of the time horizon, the calculations can readily incorporate the time needed for dam construction and reservoir filling if the infrastructure is empty at the beginning of the simulation. Additional details and interpretation issues. Before describing the specific climate linkages that are included in this study, two aspects of the simulation procedure deserve additional attention: a) the use of synthetic streamflow generation for climate change scenario analysis and b) the interpretation of results obtained for alternative system configurations under different future conditions. There are a number of well-known problems with using streamflow generation procedures for conducting climate change analysis. First, the most tractable statistical models assume that flow stationarity holds over the planning horizon (which will not be the case under climate change). While it is certainly possible to use non-stationary models, it may be simpler to explore infrastructure performance under a range of stationary climate conditions that are consistent with downscaled projections at different points in the future. The results obtained from such a procedure must then be interpreted carefully, because water flows and system outcomes pertaining to any climate experiment (with specified stationary conditions) will not represent real occurrences. Nonetheless, such specified runoff perturbations can be used to explore the general consequences of mean changes in flow that are projected to be realistic at some point in the future. Seen in this way, the results help to inform the planner of how robust the system is to mean changes in flow, temperature, etc. Thus, if the time horizon of interest is 50 years, the planner might consider projections of such changes at various points over the 50-year time horizon, say 25 or 50 years, to see how the system performance could change over the relevant

time horizon. Another way of thinking about this is that sensitivity analysis on mean changes in inflows provides valuable insights to planners and should probably be included in all water resources planning models. We will explore the sensitivity of the economic calculations to such changes in the example application described in this paper. Also, the choice of which statistical properties of flows (i.e. variance, skewness, spatial and temporal autocorrelation) to preserve is a matter of judgment. Under climate change, such properties are also likely to change as the intensity of rainfall events changes or regional circulation patterns shift. There has been very limited research into the hydrological effect of shifts in autocorrelation and in the higher moments of the statistical distribution of inflows. The application described here only considers mean changes. The second issue of importance for interpretation has to do with how results from different possible future scenarios (in terms of climate, demand, or any other projections) should be jointly considered in the planning process. While such issues are beyond the scope of this paper (we only explore one infrastructure and one climate change scenario), a few comments are warranted. Traditional economic appraisal methods for water projects seek to quantify as best possible all of their economic costs and benefits, assuming that historical hydrological and economic conditions will be roughly maintained over the planning horizon. Environmental economists such as this one typically spend considerable effort on parameterizing cost and benefit functions (i.e. demand and damage curves) based on the best available information at the time of planning, and use those functions to determine the expected net benefits of the investments. Sensitivity analysis on select model parameters then provides insight into the uncertainty associated with the calculation of net benefits. Unfortunately, the climate change problem does not readily lend itself to

calculation of expected net benefits, because the probabilities associated with different climate futures are unknown (making aggregation of outcomes across scenarios impossible), and the stability of the economic relationships may be highly uncertain. The conceptual problem with conducting such traditional economic analysis emerges from several aspects of climate change, most notably that: a) accurate prediction of future emissions levels is difficult, especially given uncertainty about future mitigation; b) the ranges of most changes caused by greenhouse gas emissions – both physical and economic – are highly uncertain; and c) the impacts are likely to vary regionally and temporally in ways that are not well understood and/or predicted using climate models available today. As a result, well-calibrated functions for expected costs and benefits may be overly precise for the valuation problem. Indeed, the type of uncertainty that results is not unlike the “unmeasurable” or “true uncertainty” of the “non-quantitative” type, that Frank Knight described in Risk, Uncertainty and Profit [1921]. The key point here is that application of the proposed simulation framework only allows the planner to begin to analyze the relative risks and upside potential of different infrastructure projects across a range of plausible climate and/or other future scenarios. Decision analytic tools can then be applied to help interpret the results across scenarios. This paper does not make recommendations for how such multi-scenario policy analysis should proceed; rather, it seeks to demonstrate the types of linkages and uncertainties that can be considered using the proposed modeling approach.

4. Description of the Modeling Example This section and the next describe an illustrative application of the hydro-economic modeling framework to a potential large infrastructure project on the Blue Nile. The application described here does not comprise a policy analysis of Blue Nile development opportunities or even of this specific Blue Nile project. The aim is instead to show how the various linkages between climate and infrastructure performance can be included in the economic appraisal of projects, and to test whether or not these are likely to be important. The analysis likewise does not focus on the state of the art for producing projections from climate models, and only uses a single climate change scenario for comparison with the historical condition. The importance to project NPV of uncertainty in different economic and hydrological factors is then discussed. The Nile Basin is an interesting case study for several reasons. First, much of the Blue Nile upstream of Egypt and Sudan remains unregulated. A series of studies conducted by the U.S. Bureau of Reclamation in the 1950s and 60s identified several attractive sites for large hydropower infrastructures in its upstream reaches (with regards to hydropower potential, surface-to-volume reservoir ratio, low potential for displacement of local populations, and low risk of earthquakes) [BCEOM et al., 1998; USBR, 1964; Whittington et al., 2009]. Second, there is an opportunity for collaborative planning of Blue Nile water resources investments among Nile riparians due to increasing participation in the Nile Basin Initiative and a growing understanding that upstream regulation has the potential for generating system-wide benefits. Third, initial research on climate change suggests that arid and semi-arid developing countries (such as those which make up the set of Nile riparians) are particularly vulnerable to the impacts of climate change [Abou-Hadid, 2006; Conway et al., 1996; Deressa, 2007; IPCC, 2007a; Strzepek and McCluskey, 2007]. New or existing infrastructures may play an important role in

adaptation to climate change, but little practical research exists to guide planners in which aspects of water resources projects provide such adaptation benefits in economic terms. Finally, there is substantial uncertainty concerning how climate change will impact the Nile Basin [Conway and Hulme, 1996; Gleick, 1991; Sayed and Nour, 2006]. The effect that such uncertainty could have on the economics of new projects has not been considered systematically in the literature, either in specific applications, or in general. Approach For simplicity, we consider the economics of one Blue Nile hydropower dam under historical conditions, and then compare results with a case pertaining to a single scenario of climate change. The latter scenario is constructed based on projections from the A2 emissions scenario from the Intergovernmental Panel on Climate Change (IPCC). 6 The two climate scenarios – historical and A2 – are characterized by temperature, precipitation and inflows into the river system. Average historical values for temperature and precipitation are used to define the historical condition, and these are perturbed according to the A2 projections for the climate change scenario. The stochastic inflow generator model is then used to predict inflows to the system (the term ΔFs,t in equation 3). For the historical condition, mean monthly inflows are set equal to those in the historical series; for the A2 scenario, these means are perturbed based on the A2 scenario projections of runoff downscaled and averaged over the modeled sub-catchments of the Nile. Other moments and temporal and spatial correlation statistics of the inflow series are not altered.

6

These projections were obtained from Alyssa McCluskey (2008). Because this project has a long time horizon, the year 2050 projections were used (roughly the midpoint of a 75-year project time horizon), rather than other available projections for 2030 or 2080. Only mean changes for 2050 were applied to historical conditions in constructing this scenario.

This paper reports on the individual and combined influence of the following climate change linkages on project NPV: 1. Physical linkages: a. The effect of temperature and precipitation on net evaporation from system lakes and reservoirs (changes in the term Es,t in equation 1 for all lakes and reservoirs); b. Temperature-induced changes in crop water requirements (changes in the term Ds,t in equations 1 and 2, for all nodes where water demands exist – see Figure 5); c. The effect of precipitation changes on irrigation requirements (again, changes in Ds,t); 2. Economic linkages: a. Increases in the value of energy due to reduction of supply of (plus increasing demand for) conventional sources; b. Changes in the value of water due to more or less constrained water supplies (due to inflow changes combined with growing demand for water in irrigation); and c. The value of carbon offsets obtained from hydropower. Because the first set of these linkages enter the modeling framework at the hydrological level, these are hereafter called the physical linkages, even though they have economic consequences. The second set enters the framework only at the economic level and are referred to as economic linkages. 7 The specific modeling assumptions for each of these linkages are explained in more

7

An obvious limitation of this approach is that there are no model feedbacks between physical outputs (for example whether irrigation water supply is sufficient) or economic factors (such as the value of water or energy), and the

detail in a technical appendix of this paper, which is available upon request. The important point to make here is that the changes caused by the climate linkages and the addition of the dam propagate through the full set of systems continuity equations. The analysis also tests the sensitivity of results to a wider range of system-wide inflow changes, spanning from a 15% overall decrease to a 6% increase, based on a review of runoff projections deemed most likely from a review of the literature for the Nile Basin. Both scenarios (historical and A2) assume the same level of irrigation and infrastructure development in the basin. Specifically, it is assumed that some additional irrigation development in the Eastern Nile will occur in Sudan and Ethiopia, consistent with aims expressed in country Master Plans [Unpublished, internal documents]. This development involves carrying out half of the Master Plan period projects in the Eastern Nile (Baro-Akobo, Blue Nile and Tekeze-Atbara catchments) plus some irrigation expansion along the Main Nile and Blue Nile in Sudan. Some of the expansion projects are currently underway, for example at Lake Tana in Ethiopia [World Bank, 2008] and around the Merowe Dam in Sudan [McDonald et al., 2008]. Total water withdrawals targets are summarized by country in Table 1. Model Parameters The costs and benefits of a large Blue Nile hydropower project have been described in other research [Whittington et al., 2009]. This case study application uses similar data, largely obtained from unpublished, internal project studies (Table 1). For the economic simulations, highly uncertain parameters are assumed to be distributed uniformly over their possible ranges, because irrigation “demands” assumed in the system. The quantity of water demanded is in reality simultaneously determined by supply and demand. As a result, demand deficits may be overstated by our analysis, if current users cut back on their water consumption in response to growing scarcity. Since water is rarely priced at it marginal value, however, this adaptation feedback may be lower than expected.

all values within the ranges are deemed equally plausible. When the data sources suggest that more confidence in expected parameter values is justified, a triangular frequency distribution is assumed. 8

5. Results This section presents the results of this illustrative application of the integrated hydro-economic assessment framework. We first explore the effect of the various climate change linkages on the physical and economic outcomes of the new Blue Nile hydropower project. The first set of results is not unlike those obtained using the standard project appraisal methodology, in the sense that the parameters in the valuation equations are assumed to be fixed (at their base case values; see Table 1) and known with certainty, and the average hydrological flow series is used. We then consider a wider range of possible NPV results, using the two stages of model simulations to incorporate uncertainty in both the economic and hydrological aspects of the planning problem. The section concludes with sensitivity analysis to determine which uncertain parameters are most important in affecting economic outcomes. The effect of the climate change linkages on the Nile system: the base case analysis Under historical climate conditions, the project NPV is 7.2 billion US$ (Internal Rate of Return

9

= 10.6%) (Table 2, Column 1). The new dam contributes an average of 10890 GW-hr/yr of hydropower, but also affects energy production from downstream reservoirs, such that power 8

In a triangular distribution, the lower and upper bounds of the specified ranges are assigned zero probability, and the triangular frequency distribution increases linearly from the lower bound up to the expected parameter value, and then decreases linearly back down to the upper bound. 9

The Internal Rate of Return (IRR) is the discount rate for which the present value or costs is just equal to the present value of benefits, such that project NPV is equal to zero.

output in Sudan increases from 6170 to 7250 GW-hr/yr and decreases at the High Aswan Dam (HAD) in Egypt from 8680 to 6940 GW-hr/yr. The increase in Sudan results primarily from more regular flow in the Blue Nile, while the reduction in Egypt is due to reduced long-term storage levels in Lake Nasser. This lower storage in Lake Nasser is itself due to two factors: a) increased irrigation withdrawals in Sudan as a result of flow regulation, and b) storage of water in the upstream reservoir. Indeed, without upstream regulation, Sudan is generally unable to reach its demand target because of the highly seasonal flow in the Blue Nile. Meanwhile, lower storage at Lake Nasser has a negligible effect on demand deficits in Egypt. The climate change linkages affect these calculations in several ways. The A2 scenario projections used here predict reduced runoff in the Blue Nile, and the resulting decreases in streamflow have an important negative effect on the project economics (Column 2). This decrease stems from reduced hydropower production and a diminished ability of the dam to reduce demand shortfalls in the downstream system. System deficits relative to target demands decrease from about 1.4 bcm/yr without the dam in Sudan and Egypt, to 0.8 bcm/yr (a reduction of 0.6 bcm/yr). This is similar to the 0.8 bcm/yr net reduction that is possible under historical flow conditions (from 0.9 bcm/yr to 0.1 bcm/yr). 10 Under the modeled conditions, changes in net evaporation from Nile Basin reservoirs and lakes have a slightly negative effect on the system (Column 3), because increased evaporation is mostly offset by greater rainfall over the large Equatorial lakes in the White Nile system. Also, even though most withdrawals are downstream of the new dam, temperature-induced increases in crop-water requirements result in higher irrigation demands in Sudan (from 16.1 bcm/yr to 16.7 bcm/yr) and Ethiopia (from 3.4 bcm/yr to 10

It should be noted that this analysis assumes that the Blue Nile Dam would be operated based on proposed operating rules designed to maximize hydropower. Coordination with downstream demands would allow greater reduction of downstream demand shortfalls under climate change, though this would come at the cost of reduced hydropower generation.

3.8 bcm/yr). This decreases project NPV because more water is removed from the system, such that incremental effects on hydropower production and irrigation deficits in Egypt are increased (Column 4). The effect of the final linkage – changes in precipitation over irrigated lands – on the calculations is very small (Column 5), because most irrigation in the Nile Basin occurs over arid lands, such that changes in precipitation have a minor effect on the calculations . When all physical linkages are combined (Table 2, Column 6), the NPV is reduced to 5.0 billion US$ (IRR = 9.4%). This decrease results from three factors mentioned above. The first is reduced hydropower production in the new dam (10100 GW-hr/yr, compared to the 10890 under historical conditions) (Figure 6). The second factor is the downstream reduction in hydropower production at the High Aswan Dam (which is reduced from 5650 GW-hr/yr without the Blue Nile dam to 3570 GW-hr/yr). The third important factor in the reduced NPV of the project is the increase in target water demands in the downstream system. Under historical conditions, the regulated flow from the Blue Nile Dam enables Sudan to meet target demands without affecting Egyptian irrigation. With A2-scenario temperature and crop-water demand increases, Blue Nile flow regulation due to the dam project still allows Sudan to achieve its target withdrawals. This increased level of upstream abstractions, however, somewhat increases deficits in Egypt. The important point here is not the precise location of these demand shortfalls (since these could be shared among riparians), but rather the fact that climate change could impact water availability such that existing plans for irrigation expansion require water abstractions that cannot be maintained over the long term. Upstream storage appears to have the potential to reduce total deficits in the system, but cooperation among riparians will be necessary to manage reservoir releases most effectively.

In contrast to the physical linkages, the economic linkages increase the value of the project in the base case. This is because of the positive potential link between climate change and the real value of hydropower, since a Blue Nile dam contributes easily enough energy to outweigh decreases in power generated downstream (Column 7). At this site, carbon offsets outweigh estimated construction and reservoir emissions and therefore also contribute to increasing net benefits (Column 9). The assumed increase in the value of water over time however does not change the value of the dam appreciably because the project has a very modest impact on system deficits (Column 8), even as it may affect their distribution in time and space. Due to the net positive effect of the economic linkages, the project NPV in the base case with all linkages is US$7.9 billion (IRR = 10.8%) (Column 10), which is slightly higher than the NPV for historical conditions. Results of the economic simulations: the effects of uncertainty Consistent with these results, the economic simulations show that the linkages with decreased inflows and temperature-related increases in crop-water requirements both shift the cumulative distribution of possible NPV outcomes to the left (Figure 7A). For the decrease in runoff alone, this shift is about 2.0 billion US$ at the median of the distribution (2.1 billion US$ at the mean), but very few (only 1.7%) simulations result in negative project NPV (Table 3). The increased crop-water requirements (from temperature) in the system shifts the distribution an additional US$340 million at the median (-280 at the mean), and this increases the percentage of simulations that yield negative NPV to 3.0%. The other physical linkages have a small negative impact on the NPV distribution. The cumulative effect of all physical linkages is thus a decrease of US$2.4 billion at the median of the NPV distribution (2.6 at the mean) relative to historical

conditions, and 3.4% of simulations yield a negative project NPV, compared with 0.2% in the historical case. Also similar to the analysis of the base case, the effect of the economic climate change linkages is mostly to increase the economic value of the Blue Nile dam (Figure 7B). Including increases in the real value of hydropower shifts the median of the distribution to the right by about US$1.8 billion (2.5 at the mean), and decreases the percentage of simulations with negative NPV outcomes to about 1.1%. The increased real value of water has a minor impact of +US$20 million at the median (-50 at the mean). Finally, including the value of carbon offsets in the economic analysis shifts the distribution to the right by an additional US$1.5 billion at the median (1.6 at the mean). Once all linkages are included, only 0.3% of simulations yield negative NPV outcomes, and the economic value of the dam is higher than under historical conditions in over 96% of the simulations (Figure 7B). These results suggest that a Blue Nile hydropower dam is likely to have value in adaptation to climate change, at least under these particular conditions. Negative economic outcomes here only occur with an unlikely convergence of negative factors: extended periods of reduced flows from the Blue Nile coupled with large increases in the relative cost of irrigation demand shortfalls, a low value of carbon offsets and hydropower, as well as capital costs that are much higher than expected. Additional sensitivity analysis Given the large uncertainties associated with Nile Basin precipitation and runoff predictions under climate change, additional sensitivity analyses were conducted over a wider range of inflow changes and withdrawal scenarios, including all climate linkages and the A2 scenario temperature projections. These experiments show that economic outcomes are highly sensitive to

changes in inflows (Figure 8). The span in the median NPV across the range of inflow sensitivity scenarios – from -15% to +6% – is about US$10 billion (-15% inflows median NPV = US$2.9 billion; +6% median NPV = US$12.9 billion). Also, the point at which the dam’s value decreases from the historical climate condition is somewhere between a 5 and 10% decrease in system inflows. If flow were to decrease by 15% in the Nile Basin, the simulations show that it is somewhat likely that the Blue Nile Dam project would not pass a cost-benefit test (21% of simulations yield a negative NPV outcome), though median and mean NPV outcomes remain positive. On the other hand, if flows increase by 6% or even stay the same, it is very difficult to imagine conditions that would result in the project failing a cost-benefit test. Thus, in this particular application, changes in inflows do more to change outcomes than the inclusion of the other physical climate change linkages, which sum to -0.1 billion US$ at the median (Table 3). The linkages with the strongest influence on economic outcomes are the negative effect of increasing crop-water demand (-0.3 billion US$) and the positive effects of the increased real value of hydropower (+1.5 billion US$) and carbon offsets (+1.6 billion US$). Other parameters in the economic model also have an important effect on project NPV, though no single parameter is sufficient to induce a negative project NPV on its own (Figure 9). These parameters are: the discount rate, the value of energy (both base hydropower value and % change in real value over time), the length of the time horizon, natural hydrological variability (determined by the stochastic sequence number that is randomly selected for the simulation), the value of downstream irrigation water, and construction delays. Comparing Figure 9A and B, we further note that many of the same parameters contribute to the variation in outcomes under historical and A2 scenario conditions. However, the uncertainty associated with climateperturbed inflows (discussed above, and spanning 10 billion US$) is second only to the discount

rate, which alone alters the dam’s economic value by 14.0 billion US$ in the climate change scenario). The inability to predict runoff precisely therefore represents a significant source of uncertainty in the economics of Blue Nile hydropower infrastructure. The other parameter that contributes to a large change in outcomes (third in importance; spanning 8.3 billion US$) is the value of energy.

6. Discussion This research has demonstrated the use of a hydro-economic framework for assessing the economic value of new infrastructures in the context of climate change. Development of this framework was motivated by the difficulty associated with using conventional water resources planning models when future climatic, hydrological and economic conditions are highly uncertain. The framework relies on integrated hydrological and Monte Carlo simulation methods to assess the influence of various climate linkages and uncertainties on the net present value of large water infrastructure projects. The framework was used to evaluate the economics of a potential hydropower project located in the Blue Nile in Ethiopia. The novel aspect of the research described in this paper was its exploration of the simultaneous influence of a number of physical and economic effects of climate change on project net present value: physical changes in runoff, net evaporation from reservoirs, and crop water requirements in irrigation, as well as economic changes in the value of water and energy, and the value of carbon offsets. One important goal of the research was to learn about the importance of including such linkages when conducting project assessments. In the Blue Nile application, four such linkages proved noteworthy: changes in runoff, crop-water requirements, the changing value of

energy, and the inclusion of carbon offset value. Another important result was the finding that the linkages may have compensating effects. Comparing the effects of the various modeled climate linkages for the Blue Nile example, it was shown that decreases associated with physical linkages and reduced runoff could be offset by increased benefits resulting from higher energy values and the inclusion of carbon offset value. The overall effect of modeling the linkages was to increase the percentage of outcomes at the upper end of the cumulative NPV distribution. It thus seems possible that a Blue Nile dam could provide greater benefits under climate change than in the historical condition. It should not be generally expected that these specific climate linkages will be most important in other river basins or even in planning development in other parts of the Nile. It is an empirical question where different linkages will be most important, and more research is needed to understand them. Data permitting, attempts should thus be made to build them into water resources models, to better understand where they are likely to be important, and how natural and economic feedbacks may influence outcomes. There should also be continued research on identifying and better modeling linkages that represent evolving conditions within hydrological systems. The roles of adaptation in the agriculture sector, population migration, and land degradation, for example, have scarcely been considered in water resources planning. In the Blue Nile Dam project application, it was also shown that the uncertainty associated with plausible and modest changes in mean inflows was second only to the discount rate in shifting NPV outcomes. Other parameters of importance in altering project NPV were related to the value of energy generated at the dam site and in the system, the length of the planning horizon, the natural hydrological variability in the Blue Nile, the value of downstream irrigation water, and risks related to delays in the start of dam operation (construction delays). The individual

physical climate change linkages were relatively less important than inflow changes, but an increase in crop-water requirements (demand for irrigation water) due to climate change could have important physical consequences in this system, via its effect on the downstream water balance. The analysis conducted in this paper was meant to illustrate a methodological approach for integrating climate change into the conventional reservoir planning problem, and should not be considered to be a full economic appraisal of this project. To be viewed as such, the analysis would need to satisfy additional requirements, including for example: a) more detailed analysis of the distributive impacts of the project, b) consideration of different designs and operating regimes for the infrastructure in question, c) the economics of alternative projects, and d) the performance of these projects in a variety of plausible future climate conditions. For example, it is likely that downstream irrigation water shortfalls in Egypt could be mitigated by coordinating operation of a Blue Nile dam with water levels at Lake Nasser, or that combinations of dams might provide synergies in (or increase risks associated with) hydropower-based development. The costs and benefits of such infrastructure coordination strategies have not been assessed in this research. Also, it is possible that no project in the choice set of alternatives will perform best in all plausible future scenarios; in such cases, decision-analytic tools will be required to select solutions that are more robust or that leave open the possibility of high economic gains.

Acknowledgements Thanks are due to the Eastern Nile Technical Regional Office in Addis Ababa, which generally facilitated this research. Dale Whittington, Donald Lauria, Gregory Characklis, Mohamed Abdel-

Aty Sayed, Nagaraja Harshadeep and Harvey Jeffries all provided useful comments on earlier versions of this work. Other colleagues who provided useful comments and support include Abdulkarim Seid, Ahmed Khalid Eldaw, Yohannes Daniel, Ken Strzepek, Alyssa McCluskey, Casey Brown, Declan Conway, and many others. Thanks are also due to the anonymous reviewers of this paper. The author is solely responsible for any errors that remain.

References Abou-Hadid, A. F. (2006), Assessment of impacts, adaptation and vulnerability to climate change in North Africa: food production and water resources., 127 p., Washington, D.C. Arnell, N. (2004), Climate change and global water resources: SRES emissions and socioeconomic scenarios, Global Environmental Change, 14(1), 31-52. BCEOM, et al. (1998), Part 1: Main Report - Water Resources Studies, in Abbay River Basin Integrated Development Phase 3 Master Plan Project, Ministry of Water Resources, Federal Democratic Republic of Ethiopia, Addis Ababa, Ethiopia. Bouwer, H. (2002), Integrated water management for the 21st century: Problems and solutions, Journal of Irrigation and Drainage Engineering, 128, 193. Bras, R. L., and I. Rodríguez-Iturbe (1993), Random Functions and Hydrology, 559 p., Courier Dover Publications. Cawood & Associates (2005), An initial rapid appraisal of flood damages along the Blue and Main Nile Rivers in Sudan: Draft report, 34 p., Michael Cawood & Associates Pty Ltd

Cohon, J. L. (1978), Multiobjective Programming and Planning, Academic Press, New York. Conway, D., and M. Hulme (1996), The Impacts of Climate Variability and Future Climate Change in the Nile Basin on Water Resources in Egypt, International Journal of Water Resources Development, 12(3), 277-296. Conway, D., et al. (1996), Future availability of water in Egypt: The interaction of global, regional, and basin scale driving forces in the Nile Basin, Ambio. Stockholm, 25(5), 336-342. Davies, H., and R. Walsh (1997), Historical Changes in the Flood Hazard at Khartoum, Sudan: Lessons and Warnings for the Future, Singapore Journal of Tropical Geography, 18(2), 123-140. Deressa, T. T. (2007), Measuring the economic impact of climate change on Ethiopian agriculture: a Ricardian approach, 32p., The World Bank, Washington, D.C. Eckstein, O. (1958), Water-resource development: the economics of project evaluation, Harvard University Press, Cambridge, MA. Fiering, M., and B. Jackson (1971), Synthetic Streamflows Water Resources Monograph no. 1, American Geophysical Union, Washington, DC. Frederick, K., and D. C. Major (1997), Climate Change and Water Resources, Climatic Change, 37, 7-23. Frederick, K. D., et al. (1997), Climate change and water resources planning criteria, 316 p., Kluwer Academic, Dordrecht; Boston and London. Gleick, P. H. (1991), The vulnerability of runoff in the Nile Basin to climate changes, The Environmental Professional, 13, 66-73.

Hanemann, W. M. (1992), Preface, in Pricing the European environment, edited by S. Navrud, Scandinavian University Press, Oslo. Harou, J. J., et al. (2009), Hydro-economic models: Concepts, design, applications, and future prospects, Journal of Hydrology, 375(3/4), 627-643. Hassan, F. (1981), Historical Nile Floods and Their Implications for Climatic Change, Science, 212(4499), 1142-1145. Hufschmidt, M. M., and M. B. Fiering (1966), Simulation techniques for design of waterresource systems, Harvard University Press, Cambridge, MA. Inter-Agency River Basin Committee (1950), Proposed practices for economic analysis of river basin projects, Sub-Committee on Costs and Budgets, Washington, DC. IPCC (2007a), Climate Change 2007: Impacts, Adaptations and Vulnerability: ScientificTechnical Analyses: Contribution of Working Group II to the Second Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK. IPCC (2007b), Climate Change 2007: The Physical Basis of Climate Change: Contribution of Working Group I to the Fourth Assessment Report of the Intergovernmental Panel on Climate Change, Cambridge University Press, Cambridge, UK. Jenkins, M. W., et al. (2004), Optimization of California’s Water Supply System: Results and Insights, Journal of Water Resources Planning and Management, 130, 271-280. Knight, F. L. (1921), Risk, Uncertainty and Profit, Houghton Mifflin Company, Boston.

Krutilla, J. V., and O. Eckstein (1958), Multiple purpose river development, Johns Hopkins Press, Baltimore, MD. Leavesley, G. H. (1999), Overview of models for use in the evaluation of the impacts of climate change on hydrology, in Impacts of Climate Change and Climate Variability on Hydrological Regimes, edited by J. C. van Dam, pp. 107-122, Cambridge University Press, Cambridge, U.K. Lettenmaier, D. (2008), Editorial: Have We Dropped the Ball on Water Resources Research?, Journal of Water Resources Planning & Management, 491-492. Lettenmaier, D. P., et al. (1999), Water resources implications of global warming: A U.S. regional perspective., Climatic Change, 43, 537-579. Loucks, D., et al. (1981), Water Resource Systems Planning and Analysis, Prentice-Hall, Inc., Englewood Cliffs, NJ. Loucks, D. P., et al. (2005), Water Resources Systems Planning and Management: An Introduction to Methods, Models and Applications, Paris: UNESCO. Maass, A., et al. (1962), Design of Water-Resource Systems: New Techniques for Relating Economic Objectives, Engineering Analysis, and Governmental Planning, Harvard University Press, Cambridge, MA. Marchant, R., and H. Hooghiemstra (2004), Rapid environmental change in African and South American tropics around 4000 years before present: a review, Earth-Science Reviews, 66(3), 217-260.

McDonald, K., et al. (2008), Exporting dams: China’s hydropower industry goes global, Journal of Environmental Management, 1-9. Mearns, L. O., et al. (2003), Guidelines for use of climate scenarios developed from Regional Climate Model experiments, Data Distribution Centre of the Intergovernmental Panel on Climate Change. Mendelsohn, R., et al. (1994), The Impact of Global Warming on Agriculture: A Ricardian Analysis, American Economic Review, 84(4), 753-771. Milly, P., et al. (2005), Global pattern of trends in streamflow and water availability in a changing climate, Nature, 438(7066), 347-350. Nicholson, S. (2001), Climatic and environmental change in Africa during the last two centuries, Climate Research, 17(2), 123-144. Nicholson, S., and X. Yin (2001), Rainfall Conditions in Equatorial East Africa during the Nineteenth Century as Inferred from the Record of Lake Victoria, Climatic Change, 48(2), 387398. ReVelle, C., and A. E. McGarity (1997), Design and Operation of Civil and Environmental Engineering Systems, Wiley-Interscience, New York. Rogers, P. P., and M. B. Fiering (1986), Use of systems analysis in water management, Water Resources Research, 22(9), 146-158. Sayed, M. A., and M. Nour (2006), Impacts of Climate Change on Nile Flows, 19p., Eastern Nile Technical Regional Office (ENTRO), Addis Ababa.

Schlenker, W., et al. (2006), The Impact of Global Warming on U.S. Agriculture: An Econometric Analysis of Optimal Growing Conditions, Review of Economics and Statistics, 88(1), 113-125. Shahin, M. (1985), Hydrology of the Nile Basin, Elsevier Science Publishers B.V., Amsterdam, the Netherlands. Stakhiv, E. (1998), Policy implications of climate change impacts on water resources management, Water Policy, 1(2), 159-175. Sterner, T., and U. M. Persson (2008), An even sterner review: Introducing relative prices into the discounting debate, Review of Environmental Economics and Policy, 2(1), 61-76. Strzepek, K., and A. McCluskey (2007), The Impacts of Climate Change on Regional Water Resources and Agriculture in Africa, 62p., The World Bank. Tanaka, S. K., et al. (2006), Climate Warming and Water Management Adaptation for California, Climatic Change, 76(3), 361-387. USBR (1964), Land and Water Resources of the Blue Nile Basin: Main Report and Appendices I-V, United States Department of Interior; US Government Printing Office, Washington, D.C., USA. Vicuna, S., and J. A. Dracup (2007), The evolution of climate change impact studies on hydrology and water resources in California, Climatic Change, 82(3), 327-350. Whittington, D., et al. (2005), Water Resources Management in the Nile Basin: The Economic Value of Cooperation, Water Policy, 7(3), 227-252.

Whittington, D., et al. (2009), The Challenge of Improving Water and Sanitation Services in Less Developed Countries, Foundations and Trends in Microeconomics, 4(6-7), 469-609. Wood, A. W., et al. (1997), Assessing climate change implications for water resources planning, Climatic Change, 37(1), 203-228. World Bank (2008), Project Appraisal Document on a Proposed Credit to the Federal Democratic Republic of Ethiopia for a Tana & Beles Integrated Water Resources Development Project, 131 p., The World Bank, Washington, DC. World Commission on Dams (2000), Dams and Development: A New Framework for DecisionMaking: The Report of the World Commission on Dams, 356p., Earthscan Publications Ltd., London, UK. Wu, X., and D. Whittington (2006), Incentive compatibility and conflict resolution in international river basins: A case study of the Nile Basin, Water Resources Research, 42(2417), doi:10.1029/2005WR004238. Yao, H., and A. Georgakakos (2003), Nile DST User Manual: River and Reservoir Simulation, 44 p., Georgia Water Resources Institute; Georgia Tech, Atlanta, Georgia. Yates, D., et al. (2005), WEAP21–A Demand-, Priority-, and Preference-Driven Water Planning Model, Water International, 30(4), 487-500. Yeh, W. W. G. (1985), Reservoir Management and Operations Models: A State-of-the-Art Review, Water Resources Research WRERAQ, 21(12), 1797-1818.

Zagona, E. A., et al. (2001), Riverware: A generalized tool for complex reservoir system modeling, Journal of the American Water Resources Association, 37(4), 913-929.

Tables and Figures

Figure Captions

Figure 1. The traditional framework for economic appraisal of water resources investments

Figure 2. Illustrative representation of nodes modeled in a water resources system, with description of typical assumptions

Figure 3. A modified simulation framework for economic appraisal of water resources investments, showing the two levels (hydrological and economic), with additions to the traditional framework identified by shading and italics.

Figure 4. Flow chart showing the connections between the models of the proposed framework. Solid arrows show the eight steps of the operational framework, and dotted lines represent functional linkages among model components.

Figure 5. Nile model schematic. Inflows for the eleven nodes can be generated using the stochastic simulator, which preserves cross-node correlations.

Figure 6. The effect of the climate change linkages on hydropower production in Sudan, at the High Aswan Dam in Egypt, and at Mendaya (base case outcomes)

Figure 7. The cumulative effect of the A) physical and B) economic climate change linkages on the NPV outcomes of the Blue Nile Dam project (results of 10,000 Monte Carlo trials)

Figure 8. The effect of inflow variation on the NPV outcomes of the Blue Nile Dam project, assuming moderate upstream irrigation development (results of 10,000 Monte Carlo trials)

Figure 9. The effect of individual parameters on the net present value of the project under A) Historical and B) A2 climate conditions; with median outcome indicated by the solid black line.

1 2 3 4

5 6 7 8 9 10

Tables Table 1. Key parameters for the economic assessment of a Blue Nile dam

General model parameters Target demand in Egypt (bcm/yr)a Target demand in Sudan (bcm/yr, baseline temperatures)a Target demand in Ethiopia ((bcm/yr, baseline temperatures)a Dam construction time (yrs) Dam project duration (yrs) Discount rate Cost components Capital cost of dam (millions of US$) Capital cost of transmission to Egypt (millions of US$) Shadow value of capital Construction delay (yrs) Renewal of electrical infrastructures (yrs) O&M expenditures (As % of annual capital cost) Cost of deficits (multiple of the value of downstream water) # Households displaced Economic loss per displaced household (US$) Area of downstream production lost: grazing + agriculture (hA) Economic loss per hectare (US$) Risk of catastrophic failure (%) Project emissions (millions of tons of CO2 ) Benefit components Hydropower generated at dam (GW-hr/yr) Installed capacity (MW) Net gain in hydropower in Sudan and Egypt (GW-hr/yr) Value of hydropower (US cents/kW-hr) Change in timely irrigation water downstream (bcm/yr) Net value of timely water downstream ($US/cm) Expected flood damage in Sudan (millions of US$/yr) b Decrease in probability of flood (%) Price of offsets (US$/ton CO2) Carbon offset factor Change parameters: Historical (in %/yr, net of inflation) Annual change in value of hydropower Annual change in value of timely water Change parameters: A2 Climate (in %/yr, net of inflation) Annual change in value of hydropower Annual change in value of timely water Annual change in value of offsets a b c

Base Case Value

Sensitivity Range

Probability Distribution

55.5 16.1 3.4 10 75 0.04

N/A N/A N/A N/A 30 – 100 0.02 – 0.06

N/A N/A N/A N/A Triangular Uniform

2750 800 1 0 20 50 2 120 3500 25000 20 0.01 6.4

2200 – 3300 Uniform 640 – 960 Triangular 1 N/A (-2) – 2 Uniform, integer None N/A 35 – 65 Triangular 1–3 Triangular 60 – 340 Uniform 1750 – 5250 Triangular 12500-37500 Triangular 10 – 100 Triangular 0.002 – 0.02 Uniform 4.8 – 8.6 Triangular

Simmodelc N/A 2000 MW N/A Simmodelc N/A 6.5 4–9 Simmodelc N/A 0.075 0.025 – 0.15 8.8 4.4 – 17.6 Simmodelc N/A 20 10 – 30 0.52 0.3 – 0.6

N/A N/A N/A Uniform N/A Uniform Triangular N/A Triangular Triangular

0 0.5

(-0.5) – 0.5 0–1

Uniform Uniform

0.5 1.0 0.5

0 – 1.5 0.5 – 1.5 0 – 1.5

Uniform Uniform Uniform

With A2 scenario temperature increases, these demands increase to 16.7 bcm/yr and 3.8 bcm/yr in Sudan and Ethiopia. Egyptian target demands remain at 55.5 as specified in the 1959 Nile Waters Agreement. Estimate from report done by Cawood and Associates [2005]. Each 100-yr series from the simmodel was assigned an integer label. A random draw from the uniform distribution of integers then determines which series is used in each of the 10,000 Monte Carlo trials.

Table 2. Summary of individual linkage experiments No Project Hydropower (HP): Sudan (GW-hr/yr) HP: High Aswan Dam (GW-hr/yr) Deficits relative to target withdrawal: Sudan (bcm/yr) Deficits relative to target withdrawal: Egypt (bcm/yr) Largest annual system deficit (bcm) % of years Sudan cannot meet target withdrawals % of years Egypt cannot meet target withdrawals Average monthly peak flow (bcm) Net change in hydropower (GW-hr/yr) Change with Project HP from project (GW-hr/yr) 99% Firm HP (GW-hr/yr) HP: Sudan (GW-hr/yr) HP: High Aswan Dam (GW-hr/yr) Deficits relative to target withdrawal: Sudan (bcm/yr) Deficits relative to target withdrawal: Egypt (bcm/yr) Largest annual system deficit (bcm) % of years Sudan cannot meet target withdrawals % of years Egypt cannot meet target withdrawals Average monthly peak flow (bcm) Net change in hydropower (GW-hr/yr) Base Case Economic Metrics Present Value Costs Present Value Benefits Net Present Value Internal Rate of Return Total effect of linkage on NPV

Key:

H R NE CWT

(billions US$) (billions US$) (billions US$) (%) (billions US$)

1. H

2. R

3. R+NE 4. R+CWT 5. R+CWP

6170 8680 0.9 0.0 14.7 99.7% 0.4% 13.6 -

5970 6170 1.1 0.3 20.7 100.0% 5.3% 12.8 -2720

5980 6170 1.1 0.3 20.5 100.0% 5.2% 12.8 11

5890 5650 1.6 0.4 22.2 100.0% 7.6% 12.7 -590

5970 6170 1.1 0.3 20.5 100.0% 5.2% 12.8 7

H

R

R+NE

R+CWT

R+CWP

10890 7560 7250 6940 0 0.1 18.2 1.0% 2.0% 7.6

10170 7190 6970 4190 0 0.8 22.8 1.0% 13.4% 7.0 -3750

10110 7180 6970 4150 0 0.8 21.7 1.0% 13.2% 7.0 -90

10160 7190 6890 3600 0 1.2 26.0 1.0% 19.7% 7.0 -690

10170 7190 6970 4210 0 0.8 22.7 1.0% 13.3% 7.0 10

H

R

R+NE

R+CWT

R+CWP

3.6 10.8 7.2 10.6% N/A

3.6 9.0 5.4 9.5% -1830

3.6 8.9 5.3 9.4% -90

3.6 8.7 5.1 9.4% -300

3.6 9.0 5.4 9.5% 5

= Historical conditions = Runoff only (A2 Scenario) = Net Evaporation (A2 Scenario) = Temperature-linked Crop Water Requirement (A2 Scenario)

CWP VH VW O

6. All Physical 5900 5650 1.6 0.4 22.3 100.0% 7.4% 12.7 -3300 All Physical 10100 7180 6890 3570 0 1.2 23.8 1.0% 19.5% 7.0 -4510 All Physical 3.6 8.6 5.0 9.3% -2210

7. R+VH 8. R+VW

9. R+O

10. All

R+VH

R+VW

R+O

All

3.6 10.5 6.9 10.2% 1500

3.6 8.9 5.4 9.5% -20

3.7 10.6 6.9 10.4% 1550

3.7 11.6 7.9 10.8% 700

= Precipitation changes over irrigated zones (A2 Scenario) = Increasing value of hydropower = Increasing value of water = Carbon offsets

Table 3. The effect of the climate change linkages on the NPV of the Blue Nile dama Linkage Experiment

H

R

R+NE

R+NE+ CWT

All Physical

Physical+ VH

Physical+ VH+VW

All

Mean (billions of US$)

7.9

5.7

5.7

5.4

5.3

7.7

7.7

9.3

Median (billions of US$) 2.5% Outcome (billions of US$) 97.5% Outcome (billions of US$) % of simulations with NPV N

5. Catalog physical / incremental simulation outputs

If p = 0

Economic parameters:

If p > 1

Costs, Benefits, Discount rate, planning horizon, etc.

6a. Run Model 3:

Economic model Simulate economic outcomes

Economic Linkages:

Offsets, Value of water, Value of HP, etc.

6b. Store Economic Outcomes:

NPV Probability distribution, Sensitivity diagrams

If k < K or p < P

7. If k < K, set k = k+1, go to step 1, or if k = K & p < P, set k = 1 & p = p+1, go to step 1

Else if k = K and p = P

8. Stop and process results

Figure 4. Flow chart showing the connections between the models of the proposed framework. Solid arrows show the eight steps of the operational framework, and dotted lines represent functional linkages among model components.

To Egyptian cities, agriculture and sea

Loss Nasser

Lake Nasser Dongola

Inflow

Khasm el Atbara Girba

Merowe Atbara Model Node

Hassanb

Inflow

Tamaniat

Lake

TK-5 Inflow Inflow Rahad Dinder

Khartoum

Potential Reservoir Existing Reservoir Withdrawals Losses

Melut Loss Sudd

Border

El Deim

Mandaya

Border

Malakal

Sudd exit Inflow

Lake Tana

Karadobi

Lake Albert

Paraa / Kyoga Outlet Lake Albert Inlet

Net inflow Victoria

Bahir Dahr / Tana Outlet Inflow Kessie

Mongala

Pakwatch / Panyango

Net inflow Tana

Kessie

Sobat

Inflow Torrents

Net inflow Albert

Roseires Inflow

Sennar

Gebel Aulia

Tana-Beles Link

Net inflow Kyoga

Kyoga

Kyoga Inlet Owen Falls/ Victoria Outlet

Lake Victoria

Figure 5. Nile model schematic. Inflows for the eleven nodes can be generated using the stochastic simulator, which preserves cross-node correlations.

Without Mendaya

With Mendaya

25000 20000 15000 10000 5000

All Physical

R+CWP

R+CWT

R+NE

R

H

All Physical

R+CWP

R+CWT

R+NE

R

0

H

Hydropower production (GW-hr/yr)

30000

Linkages included Sudanese Dams

High Aswan Dam

Mendaya Dam

Figure 6. The effect of the climate change linkages on hydropower production in Sudan, at the High Aswan Dam in Egypt, and at Mendaya (base case outcomes)

Cumulative Probability

1

0.75

0.5 Historical Conditions Runoff Only Runoff + CWT Effect

0.25

All Physical Linkages

0 $(5,000)

$-

$5,000

$10,000

$15,000

$20,000

$25,000

Net Present Value (millions of US$)

A

Cumulative Probability

1

0.75

0.5 Historical Conditions All Physical Linkages Physical Linkages + VHP

0.25

Physical Linkages + VHP + VW All Linkages 0 $(5,000)

$-

$5,000

$10,000

$15,000

$20,000

$25,000

Net Present Value (millions of US$)

B

Figure 7. The cumulative effect of the A) physical and B) economic climate change linkages on the NPV outcomes of the Blue Nile Dam project (results of 10,000 Monte Carlo trials)

1.00

Cumulative Probability

0.75

0.50

Historical Inflows +6% Inflows -10% Inflows

+3% Inflows

-15% Inflows

+0% Inflows -5% Inflows

0.25

0.00 ($15,000)

($5,000)

$5,000

$15,000

$25,000

$35,000

NPV (millions of US$)

Figure 8. The effect of inflow variation on the NPV outcomes of the Blue Nile Dam project, assuming moderate upstream irrigation development (results of 10,000 Monte Carlo trials)

Present value of net benefits (millions of US$) $0

$5,000

Discount rate (%)

6

Value of energy (US$/kW-hr)

0.04

Lifespan of civil w orks (years)

35

$10,000

$20,000

2 0.09 95 1

Stochastic sequence

$15,000

70

Change in value of hydropow er (%/year)

-0.5

0.5

Real value of w ater (US$/m3)

0.025

0.15

Time delay (yrs)

2

Value of energy from HAD (US$/kW-hr)

0.09

-2 0.04

A Present value of net benefits (millions of US$) $0 Discount rate (%)

$5,000

$10,000

$15,000

6

2

Change in inflow s (%)

-15

6

Value of energy (US$/kW-hr)

0.04

0.09

Lifespan of civil w orks (years)

35

Change in value of hydropow er (%/year) Stochastic sequence Multiplier on cost of deficits

$20,000

95

0.0%

1.5%

32

40

3

1

Real value of w ater (US$/m3)

0.025

Time delay (yrs)

2

0.15 -2

B

Figure 9. The effect of individual parameters on the net present value of the project under A) Historical and B) A2 climate conditions; with median outcome indicated by the solid black line.