Optimizing Dam Removals: A Balancing Act

Imagine a large, vibrant lake ecosystem like Lake Erie. It's teeming with life, but also faces challenges. One major stressor? Dams built on the rivers and streams that flow into the lake. While dams can provide benefits like hydropower or water supply, many older dams no longer serve their original purpose and block fish from reaching crucial historic spawning grounds.

Removing these dams seems like a good idea for restoring fish populations, right? Well, it's complicated. This article explores the core ideas from a research paper by Zheng, Hobbs, and Koonce (2009) which tackles this very problem: How do we decide which dams to remove to get the most ecological benefit, considering the costs and the potential downsides?

A simplified river flowing into a lake. Click the dam to remove it and see potential upstream habitat open up.

The Core Problem: Trade-offs Galore

Dam removal isn't a simple win-win scenario. The paper highlights several key trade-offs:

With thousands of dams in the Lake Erie basin, many being potential candidates for removal, figuring out the *best combination* (or "portfolio") of dams to remove becomes a complex optimization problem.

Measuring the Impact: Habitat and Ecosystem Health

To make informed decisions, we first need to quantify the effects of dam removal.

Habitat Suitability: Is this a good spot for fish?

Scientists use concepts like the Habitat Suitability Index (HSI) to score how good a particular river section is for a specific species and life stage (like spawning walleye or juvenile sea lamprey). HSI typically ranges from 0 (unsuitable) to 1 (perfectly suitable), considering factors like water depth, flow speed, and the type of riverbed material (rocks, sand, mud).

Removing a dam doesn't just open up *any* upstream area; it opens up area with varying degrees of suitability. The paper calculates the *amount* of suitable habitat ($H$) made accessible for both walleye (area, $km^2$) and sea lamprey (length, $km$) if a specific dam $j$ is removed.

Removing the dam opens upstream habitat. Blue represents newly accessible suitable walleye spawning area, red represents newly accessible suitable sea lamprey nursery length. The *amount* of each depends on the specific river characteristics upstream.

From Habitat to Ecosystem: The Big Picture

More habitat is generally good, but how does it translate to the overall health of the Lake Erie ecosystem? The researchers used a complex simulation model called the Lake Erie Ecological Model (LEEM). This model simulates the interactions between different fish species, their food sources, and environmental factors.

By feeding the *change* in walleye spawning potential (derived from the new habitat) into LEEM, they could estimate the impact on various lake-wide indicators ($x_1, ..., x_8$ in the paper), such as:

Combining Goals: The Ecological Health Score

With multiple ecological indicators changing (some improving, some potentially worsening), how do we get a single measure of overall benefit? The paper uses Multi-Criteria Value Analysis (MVA). Think of it like grading a student: you assign weights to homework, tests, and participation based on their importance, and then combine the scores into a final grade.

Here, each ecological indicator $x_i$ gets a weight $W_i$ reflecting its importance to managers and stakeholders. Each indicator's value is also converted to a 0-1 scale using a *value function* $V_i(x_i)$, where 0 is the worst outcome and 1 is the best. The overall ecological health score $z_1$ is then a weighted sum:

$$ z_1 = \sum_{i=1}^{8} W_i V_i(x_i) $$

A higher $z_1$ means a better ecological outcome according to the defined priorities.

Counting the Costs

Restoration isn't free. The model considers two main costs:

  1. Dam Removal Cost ($C_j^{dam}$): The cost to physically remove dam $j$. The paper uses a regression model based on historical data. Larger dams (height, length) cost more. Dams serving important functions (like water supply) cost more to remove (due to replacing the lost service). Earthen dams are generally cheaper to remove than concrete ones.
  2. Sea Lamprey Control Cost ($C_j^{lamprey}$): If removing dam $j$ opens up habitat for sea lamprey, there's an associated cost for applying lampricides to control them in the newly accessible areas. This cost depends on the length of suitable lamprey habitat created ($H_j^{lamprey}$).

The total economic cost ($x_9$) for a portfolio of removed dams is the sum of removal and lamprey control costs for all selected dams.

15 200
Estimated Removal Cost: $
Simplified cost estimation based on the paper's regression model (Eq. 14). Adjust dam characteristics to see how estimated removal cost changes. (Note: This is illustrative; actual costs vary significantly).

The Optimization Model: Making the Choice

Now we put it all together. The goal is to choose a set of dams to remove that maximizes the ecological health score ($z_1$) without exceeding a total budget ($B$). This is formulated as a Mixed Integer Linear Program (MILP).

The key decision variable is $d_j$ for each candidate dam $j$: $d_j = 1$ if dam $j$ is removed, and $d_j = 0$ if it's kept.

Objective: Maximize Ecological Health $$ \text{Maximize } z_1 = \sum_{i=1}^{8} W_i V_i(x_i) $$ Where the ecological indicators $x_i$ depend on the total change in walleye recruitment, which in turn depends on which dams are removed (the sum of benefits from dams with $d_j=1$). The paper uses impulse response functions (IR) to linearly relate the change in walleye YOY (Young-Of-Year) to changes in $x_i$: $x_i = x_i^{base} + IR_{x_i} \times \Delta YOY_{walleye}$. And $\Delta YOY_{walleye}$ is roughly proportional to the sum of habitat opened by removed dams: $\Delta YOY_{walleye} \approx \sum_{j} (\text{constant} \times EP_j \times d_j)$, where $EP_j$ is egg production potential from removing dam $j$. Subject to Constraints:
  1. Budget Constraint: The total cost cannot exceed the budget $B$. $$ x_9 = \sum_{j} (C_j^{dam} + C_j^{lamprey}) d_j \le B $$
  2. Upstream/Downstream Dependency: A dam $n$ located directly upstream of dam $j$ cannot be removed ($d_n=0$) unless dam $j$ is also removed ($d_j=1$). This ensures connectivity. $$ d_n \le d_j \quad \text{for all upstream } n \text{ of } j $$
  3. Binary Decision: Each dam is either removed or not. $$ d_j \in \{0, 1\} \quad \text{for all candidate dams } j $$

Visualizing the Optimization Concept

Imagine a simple river system with a few dams. Each has a removal cost, a potential walleye habitat gain, and a potential lamprey habitat gain (leading to control costs). There's also an upstream dependency. Given a budget, which dams should be removed to maximize walleye gain while respecting the budget and dependency?

5.0
Selected Dams: None | Total Cost: $0M | Total Walleye Gain: 0 units | Upstream Constraint Met: Yes
Simplified optimization example. Dams are squares (cost indicated). Upstream dependencies shown by dashed lines. Set a budget and click 'Run' to see a possible optimal selection (green = removed, grey = kept). This illustrates the MILP's task. (Note: This uses a simplified greedy logic for demonstration, not a full MILP solver).

Results: The Efficiency Frontier

The researchers didn't just find one "best" solution. They ran the MILP model many times, varying the budget constraint ($B$) from zero up to the cost of removing all candidate dams. This traces out an efficiency frontier (also called a Pareto front or trade-off curve).

This curve shows the maximum possible ecological health score ($z_1$) achievable for any given budget ($x_9$). It visualizes the diminishing returns – the first few million dollars spent yield large ecological gains, but achieving the highest possible score requires much larger investments for smaller incremental benefits.

15.0
The Ecosystem Health vs. Cost Trade-off Curve (based on Figure 4a). Move the slider to select a budget. The red point shows the maximum ecological health index achievable for that budget, and the tooltip shows the approximate number of dams removed in that optimal portfolio.

The specific dams included in the optimal portfolio change as the budget changes. Initially, the model tends to pick smaller, cheaper dams that open up significant walleye habitat without too much lamprey habitat. As the budget increases, larger or more expensive dams might be included if they offer substantial ecological benefits.

Sensitivity: What if priorities change?

The model also allows exploring "what if" scenarios. For example, what if stakeholders become much more concerned about preventing sea lamprey spread? The paper explored this by adding a negative weight to lamprey habitat creation in the objective function. Unsurprisingly, this leads to different portfolios being chosen – prioritizing dams that open less lamprey habitat, even if they offer slightly less walleye benefit or cost more (see Figure 5 in the paper).

Conclusion: A Tool for Informed Decisions

The MILP model developed by Zheng, Hobbs, and Koonce provides a powerful framework for navigating the complex trade-offs involved in multiple dam removals. Key takeaways:

Deciding which dams to remove is a complex environmental challenge with significant ecological and economic consequences. Mathematical optimization, coupled with ecological modeling and careful consideration of stakeholder values, offers a valuable way to make these decisions more transparent, efficient, and effective.


Based on: Zheng, P. Q., B. F. Hobbs, and J. F. Koonce (2009), Optimizing multiple dam removals under multiple objectives: Linking tributary habitat and the Lake Erie ecosystem, Water Resour. Res., 45, W12417, doi:10.1029/2008WR007589.