Evaluating Cost Trade-Offs between Hydropower and Fish Passage Mitigation

To promote the sustainable management of hydropower, decision makers require information about cost trade-offs between the restoration of fish passage and hydropower production. We provide a systematic overview of the construction, operational, monitoring, and power loss costs associated with upstream and downstream fish passage measures in the European context. When comparing the total costs of upstream measures across different electricity price scenarios, nature-like solutions (67–88 EUR/kW) tend to cost less than technical solutions (201–287 EUR/kW) on average. Furthermore, nature-like fish passes incur fewer power losses and provide habitat in addition to facilitating fish passage, which presents a strong argument for supporting their development. When evaluating different cost categories of fish passage measures across different electricity price scenarios, construction (45–87%) accounts for the largest share compared to operation (0–1.2%) and power losses (11–54%). However, under a high electricity price scenario, power losses exceed construction costs for technical fish passes. Finally, there tends to be limited information on operational, power loss, and monitoring costs associated with passage measures. Thus, we recommend that policy makers standardize monitoring and reporting of hydraulic, structural, and biological parameters as well as costs in a more detailed manner.


Introduction
While hydropower represents the largest renewable energy source in Europe, it also poses risks to river ecosystems, an array of animal species, and the downstream transport of sediments [1]. To ensure the sustainability of future hydropower development, ecological targets have been established. The primary political instrument is the EU Water Framework Directive (Directive 2000/60/ EC, WFD), which, among other goals, mandates an improvement in the ecological status of waters including the restoration of fish passage and river connectivity of European rivers. While many studies focus on technical advancements and ecological assessments of fish passage technologies [2][3][4][5][6][7], there has been limited research on the costs of these measures in the European context [8]. However, it is important for decision makers to understand the trade-offs between restoring fish passage and hydropower production when they establish and plan cost-effective mitigation measures.
Despite worldwide hydropower operation and mitigation programs, overviews of costs related to fish passage measures mainly come from North America [8]. However, it can be problematic to made in the past decade to improve downstream passage technology, the costs of these measures often fuel debate about their necessity.
Following the categories of fish passage measures described, our study reviews costs of 327 case studies from European hydropower plants. To our knowledge, most studies in the European context have primarily outlined individual case studies. Further, recent literature on the costs of measures to support migratory fish has mainly focused on financial costs associated with passage measures, while potentially significant economic costs were not considered. To shed light on the cost trade-offs associated with sustainable hydropower, we compare the costs of technical and nature-like fish passage measures. Given our limited data about special passes, we cannot draw conclusions about them in this comparison. We investigate the extent to which unit metrics are useful for predicting construction costs and how planning costs differ from actual costs. As they are a common reference point for many operators, we hypothesize that unit metrics can accurately predict costs. Further, we compare how different types of costs (construction, maintenance, power and income losses) contribute to lifetime mitigation costs. To estimate the costs associated with power losses, we compare power losses under low and high electricity price scenarios. We hypothesize that construction and power losses account for the largest shares. Finally, we hypothesize that technical measures for fish passage show lower construction costs than their nature-like counterparts. This is because nature-like measures may require additional land, which can be expensive to acquire [25]. Further, planners can standardize technical measures across sites, thus incurring fewer costs during the planning stages.

Data Acquisition
We compared the costs of building, maintaining, and monitoring fish passage measures using 327 case studies (Germany: 151, Austria: 101, Sweden: 58, France: 16, Switzerland: 1). The data were collected from available reports and through a questionnaire sent to European hydropower operators and from a search of online available data. The German data came from the Thüringen State Office for the Environment, Mining and Nature Conservation, which published reports and data on the planned costs of river connectivity measures for various rivers, including the Ilm, Saale, Unstrut, Werra, Gera, Apfelstädt, and Ohra Rivers [39][40][41][42][43]. (The data can be accessed on Thüringer State Office for the Environment, Mining and Nature Conservation (Landesamt für Umwelt, Bergbau und Naturschutz)'s website at https://tlubn.thueringen.de/wasser/fluesse-baeche/durchgaengigkeit/) The Austrian data came from the Österreichs Energie [44] reports about the National Water Management Plan 2009, which reviewed a total of 133 measures for the implementation of the Water Framework Directive. This analysis did not include all Austrian measures as some of them were related to flow and habitat improvements. The Swedish, French, and Swiss data were collected using an online questionnaire with operators (Vattenfall) and the French small hydropower association (France Hydro Electricité). While the data were not representative for all of Europe or the respective countries, they covered a variety of geographic areas (Alpine region, Scandinavia, France), plant sizes (3.5 kW to 5.88 GW), and technologies (reservoir and run-of-the-river). Table A1 in Appendix A lists the case studies used in our analysis by country, capacity, and measures type. Because the data came from different sources, it is important to note that some observations are missing variables. Thus, the number of observations is noted at each stage of our analysis.

Types of Costs
Financial and economic costs are associated with upstream and downstream passage facilities. Table 1 outlines the specific types of financial and economic costs associated with both. Notably, there were differences for maintenance, legal costs, as well as compensation for land and habitat. Certain downstream measures may not have required any or very little maintenance like passive downstream measures such as guiding walls/dams. On the other hand, downstream measures Sustainability 2020, 12, 8520 5 of 30 like screens/racks may have incurred recurrent maintenance costs associated with removing debris and cleaning the screen/rack, unless they were self-cleaning [30]. Further, downstream measures were unlikely to incur legal costs or require compensation for land and habitat as they normally do not require additional space. Economic Lost power production 2 2 Lost system flexibility 1 0 Note: 2 indicates obligatory associated costs; 1 indicates facultative associated cost, and 0 indicates no associated costs.

Data Analysis
Before analysis, we deflated our cost data (1992-2017) to 2019 Euro values to adjust for changes in values over time. We first deflated the data to the base year (2019) using average yearly inflation rates based on the Consumer Price Index (CPI) in the respective countries in the year of costing [45]. We assumed that the start year of the project was the year of costing as planners tend to make decisions based on cost estimates made in the first year of construction. For the Swedish data, we then converted Swedish krona (2019) to Euros using the average 2019 exchange rate after deflating the data. The Swiss data were already reported in Euros. As we indicate which countries the costs come from, we did not correct for purchasing power parity between countries.

Analysis of Non-Recurring Costs
First, we discussed the capital costs (construction and planning) of different types of mitigation measures. These are non-recurring costs. In the data, capital costs refer to those associated with constructing the measure. As the data came from different sources, the exact itemization was unavailable. However, it was assumed that the capital costs included the costs of planning, engineering, materials, and labor.
During the planning process, operators often estimate the capital (construction) costs. However, planned costs may differ from actual implementation. To investigate this discrepancy, we visualized the unit costs for planned and actual upstream measures for vertical slot passes (technical) and roughened bypass channels (nature-like). We selected these measures as they tended to be the most common installations for technical and nature-like passes, respectively [25], which was also reflected by the greatest number of observations in our data. We only visualized the data from Austria and Germany. As many companies in Austria and Germany plan and build fish passes in both countries, the price difference for planning and construction between Austria and Germany is likely negligent. Furthermore, many regions in Austria (northern Alps) are logical comparison units to the southern part of Germany (Bavarian Alps) from a landscape perspective.
To understand how plant and passage facility characteristics affect the capital costs for upstream passes, we estimated a linear mixed model fit by restricted maximum likelihood (REML). Observations were limited to those with complete information for the variables (n = 127). We estimated three models. The first model controlled for the specific type of measures: log e Costs i = β 0 + β 1 log e Height i + β 2 log e Length i + β 3 log e PlantCapacity i + Sustainability 2020, 12, 8520 6 of 30 log e Costs i is the total capital (construction) cost associated with upstream fish passes. log e Height i and log e Length i are the height of the obstacle to be passed and length of the pass, both in meters. log e PlantCapacity is the capacity of the plant in kW, which controls for the size of the plant. By including the logarithmic transformation of cost, we assumed that capital costs would increase exponentially, which we confirmed through visual inspection of the data. We also included measure type (γ i ) controls as we expected differences across types of measures. Random effects for the country (δ i ) are included and ε i is the error term.
We then estimated a second version of the model, which included a binary variable (Implemented i ) for whether the costs are based on a planned or implemented project. This tested whether there is a difference between planned and implemented costs while controlling for structural characteristics and the type of measure. log e Costs i = β 0 + β 1 log e Height i + β 2 log e Length i + β 3 log e PlantCapacity i +β 4 Implemented i In our third model, we replaced the specific measure variables by grouping them into three categories: technical, nature-like, and combined. The technical design was the baseline and we included interaction terms to investigate how the categories relate to the height of the obstacle and length of the pass. This tested whether generalizations can be made about cost differences between technical, nature-like, and combined fish passes and their structural characteristics. Within the categories, we expected that there would not be significant differences between measure types as technical passes always use concrete and nature-like always use natural materials for construction. Thus, construction costs should be similar but the quantity that is needed might differ. However, we controlled for the quantity by including the height, length, and plant capacity. We therefore estimate the following: log e Costs i =β 0 + β 1 log e Height i + β 2 log e Length i + β 3 log e PlantCapacity i +β 4 NatureLike i + β 5 Combined i + β 6 log e Height i * NatureLike i +β 7 log e Length i * NatureLike i + β 8 log e Height i * Combined i +β 9 log e Length i * Combined i Similarly, to understand how structural characteristics affect the capital costs for downstream measures, we estimated a model using generalized least squares fit by REML. We estimated the costs of downstream mitigation measures (i.e., screen and bypass) as a function of screen/rack area (log e Area i ) and a binary variable for rack configuration (Vertical i ) of vertical (1) or horizontal (0). ε i is the error term. Observations were limited to those with complete information for the variables (n = 17). As all observations came from Germany, we did not include country random effects. Due to missing information, we did not include other factors, such as the angle of the screen/rack.
For the 40 case studies that reported the configuration of the screen/rack (includes both fish protection screens/racks and combined screen/rack bypass systems), we tested whether there are significant differences in costs between horizontal (n = 27) and vertical (n = 13) racks.

Analysis of Recurring Costs
We then presented descriptive statistics for annual maintenance and lost power for both upstream and downstream passage measures and for the monitoring of upstream measures. These are recurring costs. While maintenance and monitoring are expressed in Euros, we presented the lost power production in kWh.
There are a variety of methods for estimating the levelized cost of a technology [46]. In this formula, I represents the investments (construction) expenditures of the mitigation measure, M t represents the maintenance expenditures in the year t, L t represents the power production losses (EUR) in the year t, C represents the plant capacity (kW), r represents the discount rate, and n represents the expected lifetime of the measure. Discounting represents the time value of money by expressing the value of future currency in the present. Thus, the discount rate captures how the market and inflation rates change over time. We used a discount rate of 4% as recommended by the European Commission [47]. We assumed a lifetime of 30 years for the passage measures as most hydropower concessions in Europe cannot be granted beyond this period [48]. Although the lifetime of passage measures may be longer than 30 years, the duration of concession is more important for the calculation of levelized costs. This is because when concessions expire, additional modernization may be required, which incurs further costs.
The electricity prices will vary based on the time, country and market in which they operate. However, given that we lacked information about when the power losses occur, detailed assumptions about prices may overpromise precision. Thus, we compared two prices based on the feed-in tariffs for retrofitted (including ecological measures) German hydropower plants under the German Renewable Energy Act in 2014. The 2014 EEG tariffs are based on the size of the plant. We selected these prices as they represent low (0.055 EUR/kWh) and high (0.125 EUR/kWh) remuneration.
Our estimation is similar to the levelized cost of electricity (LCOE) proposed by the United Kingdom's Department for Business, Innovation and Skills, which expresses the average net present cost of generation over a plant's lifetime. Notably, the LCOE is the ratio of total discounted costs over the plant's lifetime divided by the discounted sum of electricity generated. To levelize costs, we used capacity in place of electrical energy generation as information on the annual generation of each plant was missing. It is difficult to make assumptions about annual power generation based on the plant capacity as hydropower operation hours can vary greatly among plants and by season and location. Some power plants are built and used for peaking (with few operating hours), whereas others are built to run almost constantly.
As many of the case studies included only costs for certain categories, we removed observations that did not have information for all three categories of construction, maintenance, and power losses for the calculation of the LCOM. While the maintenance and power losses can plausibly be zero, we removed them unless this was explicitly stated. We also removed two measures with only one observation (guiding dam/wall and fish lift).

Results
The costs of a large variety of upstream and downstream measures are presented in Table 2. For upstream fish passage, we had most data about technical solutions (n = 160). Geographically, these measures were not evenly distributed. Of the 100 Austrian upstream case studies, the majority (71%) were technical solutions such as vertical slot passes (n = 68) and fish lifts (n = 3). Combined vertical slot and roughened bypass channels accounted for the next largest share with 20 case studies, while there were only nine nature-like measures. Similarly, of the 86 German case studies, the majority (70%) were technical solutions such as vertical slot passes (n = 55) and roughened ramps (n = 6). However, nature-like solutions such as roughened bypass channels (n = 23) accounted for the next largest share, while there were only two combined solutions. In contrast, of the 52 Swedish case studies, there was an almost equal number of technical (n = 24) and nature-like passes (n = 27), while there was only one combined measure. Seven of the eight French case studies were technical measures and the Swiss case study was a combined solution.
For downstream measures, operators most frequently reported using a combined screen/rack and bypass, followed by a fish protection screen/rack. Combined screen/rack bypass systems were reported from Germany (n = 57) and France (n = 3). Fish protection by screens/racks in front of the intake to the power plant included a Coanda effect intake screen (n = 1), horizontal rack (n = 3), horizontal rack with cleaner (n = 2), and trash racks (n = 3) and were reported from a variety of countries including France (n = 3), Germany (n = 3), Sweden (n = 2), and Switzerland (n = 1). All bypasses were reported from Germany and guiding walls/dams were from Sweden (n = 4) and Austria (n = 1).

Non-Recurring Costs of Passage Measures
Capital (construction) costs varied across measures and there were often wide ranges within the same mitigation measure ( Figure 1). Among upstream measures, fish lifts and combined vertical slot and vertical slot passes had the greatest total costs. However, between the remaining technical types (roughened ramps, Denil passes) and nature-like solutions (roughened bypass channels), there were no major differences in median costs. Notably, there were many outlier observations for vertical slot passes and roughened bypass channels.
There was less variability for downstream measures. Guiding walls/dams were the most expensive to construct whereas combined screens/racks and bypasses, fish protection screens/racks, and bypasses were relatively similar in costs.  Between technical (vertical slot) and nature-like (roughened bypass channels) upstream passes, there were also differences in unit costs ( Figure 3), but no clear trends related to increasing length or height of the pass. In terms of fish pass length, the majority of passes were shorter than 250 m. Costs ranged between 194 EUR/m and 61,301 EUR/m for vertical slot passes. Comparatively, roughened  Between technical (vertical slot) and nature-like (roughened bypass channels) upstream passes, there were also differences in unit costs ( Figure 3), but no clear trends related to increasing length or height of the pass. In terms of fish pass length, the majority of passes were shorter than 250 m. Costs ranged between 194 EUR/m and 61,301 EUR/m for vertical slot passes. Comparatively, roughened Between technical (vertical slot) and nature-like (roughened bypass channels) upstream passes, there were also differences in unit costs ( Figure 3), but no clear trends related to increasing length or height of the pass. In terms of fish pass length, the majority of passes were shorter than 250 m. Costs ranged between 194 EUR/m and 61,301 EUR/m for vertical slot passes. Comparatively, roughened bypass channels had a narrower range with costs between 125 EUR/m and 7720 EUR/m. The mean cost per meter length of vertical slot passes (6914 EUR/m) and roughened bypass channels (2233 EUR/m) was statistically different (p < 0.001). In terms of height, both types of measures overcame heights of around 20 m. There was a weak positive relationship between the height of the pass and cost per meter for vertical slot passes, which implied that as the height increases, the unit cost increases. For vertical slot passes, costs ranged between 9949 EUR and 1,592,769 EUR whereas costs ranged between 5866 EUR and 807,920 EUR for roughened bypass channels. The mean unit costs of vertical slot passes (6914 EUR/m) and roughened bypass channels (2233 EUR/m) were statistically different (p < 0.001). The mean cost per meter height of vertical slot passes (177,788 EUR/m) and roughened bypass channels (122,667 EUR/m) was marginally significant (p < 0.10).

Factors Predicting Capital Costs
Across all three regression models for upstream passage measures, the coefficient estimates were robust (Table 3). Models 1 and 2 compared the inclusion of the implemented binary variable. Both models found positive and statistically significant relationships between length of the pass and construction costs. A 1% increase in length was associated with an increase between 0.54% and 0.55% in capital costs and a 1% increase in the height of the obstacle was associated with an increase between 0.53% and 0.54% for Models 1 and 2, respectively. Also, as expected, capital costs increased with plant capacity. For the type of measures in both models, the variables for fish lift and roughened bypass channels with pool structures were statistically significant. The parameter estimate for fish lift was

Factors Predicting Capital Costs
Across all three regression models for upstream passage measures, the coefficient estimates were robust (Table 3). Models 1 and 2 compared the inclusion of the implemented binary variable. Both models found positive and statistically significant relationships between length of the pass and construction costs. A 1% increase in length was associated with an increase between 0.54% and 0.55% in capital costs and a 1% increase in the height of the obstacle was associated with an increase between 0.53% and 0.54% for Models 1 and 2, respectively. Also, as expected, capital costs increased with plant capacity. For the type of measures in both models, the variables for fish lift and roughened bypass channels with pool structures were statistically significant. The parameter estimate for fish lift was positive while that of roughened bypass channels was negative, implying that costs are higher for fish lifts than for combined vertical slot and roughened bypass channels (the baseline) while costs are comparatively lower for roughened bypass channels. The variables for roughened ramps and vertical slot passes were not significant. In Model 2, a variable for whether the costs are based on an implemented measure was included. The parameter estimate was negative and significant. This finding implied that costs are lower for implemented measures than planned ones when other structural characteristics are considered. The parameter estimates for Model 3 were slightly different due to the inclusion of the nature-like and combined binary variables (rather than the specific measure types) as well as the interaction terms. Our decision to differentiate between nature-like, technical, and combined passes aligned with the findings of significant differences between measure types in Models 1 and 2. Height, length, and plant capacity were statistically significant and positive. Both binary variables, nature-like and combined, were significant. The parameter for nature-like was negative while that of combined measures was positive. For the interaction terms for nature-like measures, the length of the pass was significant and negative. For the interaction terms for combined measures, the height of the obstacle was significant and negative. The results implied that nature-like passes are less expensive than technical solutions (the baseline), but that this difference diminishes as the pass increases in length. In contrast, combined solutions tend to be more expensive than technical ones, but this difference diminishes as the height of the obstacle increases. When Models 1 and 2 were estimated with Ordinary Least Squares (OLS), the models explained 77% of the total variance observed (R-squared). Model 3 explained 78% of the total variance observed. Table A2 shows the 95% confidence intervals for the upstream models.
For downstream measures, one model was estimated with 17 observations of combined screen/rack and bypass systems from Germany ( Table 4). The area of the rack was positively and significantly associated with cost. The estimate indicated that, on average, each additional percent of square meter was associated with a 1.18% increase in capital costs. The configuration of the rack (vertical or horizontal) was not significant. When the model was estimated with Ordinary Least Squares, it explained approximately 93% of the variance. When we tested for differences in costs related to the configuration of the rack with a t-test for additional case studies, the difference in mean costs of horizontal (210,671 EUR) and vertical (243,793 EUR) racks was also not statistically significant. Table A3 shows the 95% confidence intervals for the downstream model.

Recurring Costs of Passage Measures
For upstream measures, recurring costs such as annual maintenance and monitoring did not account for a large share of lifetime costs (Table 5). For all measures, the annual maintenance costs ranged from 0 EUR/year to 50,220 EUR/year with an overall average of 13,139 EUR per year. Fish lifts had the highest average maintenance costs compared to other measures, but the case study with the highest maintenance costs was a vertical slot pass. In terms of variation, there were large standard deviations across categories, particularly for the roughened bypass channel. For downstream measures, only two case studies with a guiding dam/wall reported maintenance costs. Between these two case studies, there was large variation.
Our review of monitoring costs found high variability across case studies and notable differences between the monitoring of simple technical, complex technical, nature-like, and combined upstream measures (Table 6). Monitoring for complex technical (i.e., fish lift) and nature-like measures was most expensive as fish lifts had a mean monitoring cost of 160,456 EUR and roughened bypass channels had a mean cost of 443,503 EUR. Comparatively, vertical slot passes had a mean cost of 92,918 EUR and combined solutions cost of 71,262 EUR on average. However, these values were highly variable as indicated by the large standard deviations.
Beyond costs, additional information about the type of monitoring was limited. Only eight measures specified what kind of monitoring was conducted. This included one-time monitoring of water variables (temperature, quality, discharge, etc.) as well continuous monitoring (over a few months) of habitat availability and fish pass functionality. Fish pass functionality was monitored with fish traps, both alone or combined with video, as well as image-based scanner systems. With the exception of fish traps, the specific types of monitoring tended to be more expensive than the average reported overall. For the three cases of fish traps, the mean cost was 14,215 EUR with a standard deviation of 8740 EUR. Comparatively, monitoring with video in addition to fish traps had a mean cost of 157,223 EUR for three case studies with a standard deviation of 104,405 EUR. Finally, image-based scanner systems cost an average of 142,401 EUR for two case studies. Only five measures reported the time span of monitoring, which ranged from one to 24 months. Annual power losses also represented a recurring cost, which differed sizably across measures ( Table 7). For upstream measures, fish lifts and roughened bypass channels incurred the greatest costs on average, although these were highly variable. The majority of observations were vertical slot passes with an average of 149 GWh lost per year and combined vertical slot and bypass channels with an average of 307 GWh lost per year.
For downstream measures, fish protection screens/racks and combined screen and bypass systems reported minimal power losses. For both measures, only three case studies reported power losses, which may be unrepresentative of downstream mitigation measures.

Levelized Costs of Upstream Passage Measures
In total, 55 observations were included from Austria (52) and Germany (3). Our results demonstrated that regardless of the electricity price selected, capital (construction) costs tended to account for the largest share of lifetime costs with the exception of vertical slot passes under the high price scenario (Figure 4). Under the high price scenario for vertical slot passes, capital accounted for 45% while power losses accounted for 54% of lifetime costs. For all other scenarios, capital costs represented between 57-76% of lifetime costs compared to 64-87% under the low price scenario. On average, capital costs were highest for combined measures (415 EUR/kW) compared to technical (130 EUR/kW) or nature-like (50 EUR/kW). However, this was highly variable as combined solutions had a standard deviation of 540 EUR/kW. There was also a high variability of technical (214 EUR/kW) passes, but less variability for nature-like passes (48 EUR/kW).
average, capital costs were highest for combined measures (415 EUR/kW) compared to technical (130 EUR/kW) or nature-like (50 EUR/kW). However, this was highly variable as combined solutions had a standard deviation of 540 EUR/kW. There was also a high variability of technical (214 EUR/kW) passes, but less variability for nature-like passes (48 EUR/kW). Across all measures, maintenance costs accounted for a comparatively small share of levelized lifetime costs. On average, maintenance costs were greatest for combined measures (3 EUR/kW), followed by technical vertical slots (2 EUR/kW). Some missing observations may have represented zero for maintenance costs. Since missing data and zeros cannot be distinguished unless specified, this may have upwardly biased the results. Of the observations included, three reported maintenance costs of zero.
Despite our simplified representation of electricity prices, the comparison of low and high prices revealed that costs associated with power losses can vary greatly. Comparing the low (high) electricity prices, vertical slot passes had the highest mean cost of 68 EUR/kW (154 EUR/kW), compared to 56 EUR/kW (128 EUR/kW) for combined vertical slot/bypass channel and 16 EUR/kW (37 EUR/kW) for roughened bypass channels. The median, however, showed that roughened bypass channels have the greatest costs (16 EUR/kW; 36 EUR/kW) compared to combined (13 EUR/kW; 31 EUR/kW) and vertical slot (8 EUR/kW; 20 EUR/kW). Similar to maintenance costs, power losses could be zero. Since missing data and zeros cannot be distinguished, this may have upwardly biased the average power losses associated with each measure type. Of the observations included, five reported power losses of zero.

Discussion
To understand the cost trade-offs associated with sustainable hydropower, we established several hypotheses about the costs of fish passage mitigation. We hypothesized that planned costs and unit metrics can accurately predict costs. Further, we posited that construction and power losses account for the largest shares of lifetime mitigation costs. Finally, we hypothesized that technical measures incur lower costs than their nature-like counterparts.
However, we found that planning costs differ substantially from actual costs. This may be because planned costs often cannot account for difficulties that arise throughout the construction process (e.g., shortages and changes in the prices of materials) as well as site-specific factors (e.g., target species, difficulty of site accessibility, ground conditions, and local regulations) [25]. As many Across all measures, maintenance costs accounted for a comparatively small share of levelized lifetime costs. On average, maintenance costs were greatest for combined measures (3 EUR/kW), followed by technical vertical slots (2 EUR/kW). Some missing observations may have represented zero for maintenance costs. Since missing data and zeros cannot be distinguished unless specified, this may have upwardly biased the results. Of the observations included, three reported maintenance costs of zero.
Despite our simplified representation of electricity prices, the comparison of low and high prices revealed that costs associated with power losses can vary greatly. Comparing the low (high) electricity prices, vertical slot passes had the highest mean cost of 68 EUR/kW (154 EUR/kW), compared to 56 EUR/kW (128 EUR/kW) for combined vertical slot/bypass channel and 16 EUR/kW (37 EUR/kW) for roughened bypass channels. The median, however, showed that roughened bypass channels have the greatest costs (16 EUR/kW; 36 EUR/kW) compared to combined (13 EUR/kW; 31 EUR/kW) and vertical slot (8 EUR/kW; 20 EUR/kW). Similar to maintenance costs, power losses could be zero. Since missing data and zeros cannot be distinguished, this may have upwardly biased the average power losses associated with each measure type. Of the observations included, five reported power losses of zero.

Discussion
To understand the cost trade-offs associated with sustainable hydropower, we established several hypotheses about the costs of fish passage mitigation. We hypothesized that planned costs and unit metrics can accurately predict costs. Further, we posited that construction and power losses account for the largest shares of lifetime mitigation costs. Finally, we hypothesized that technical measures incur lower costs than their nature-like counterparts.
However, we found that planning costs differ substantially from actual costs. This may be because planned costs often cannot account for difficulties that arise throughout the construction process (e.g., shortages and changes in the prices of materials) as well as site-specific factors (e.g., target species, difficulty of site accessibility, ground conditions, and local regulations) [25]. As many of these complications cannot be foreseen or quantified, we investigated whether other variables can improve predictions about the construction costs of fish passage mitigation.
For upstream passes, we investigated the extent to which the height, length, or type of pass predicts construction costs. Understanding whether height or length is the key driver of upstream passage costs may affect the decision-making process of operators. For example, if an operator is deciding between a fish lift and vertical slot pass, they may find it is cheaper to build the lift if length is the main driver of costs. We found a strong relationship between construction costs and the height of the obstacle and length of the fish pass. A possible exception to the usefulness of height for predicting upstream construction costs are fish lifts, as previous literature suggested that their construction costs are relatively independent of dam height [49].
For downstream passes, we investigated whether the area and configuration of the rack/screen predicts construction costs. Our analysis confirmed that the area can be useful for predicting costs. While a rack/screen is a standard installation for hydropower plants used to repel debris, some can be adapted to function as fish protection and a guiding structure. For this purpose, adaptations in bar spacing (narrowing) or angle as well as inclination are typical but may contribute to higher expenditures (purchase price, reduction in flow during operation, which reduces power production) [29]. Ideally, other variables about the presence of a cleaning device and inclination (angle) of the rack, and the cost of the bypass structure would have been relevant to potentially explain more variance. As our data did not comprehensively provide information on inclination or other specific characteristics of the racks/screens, only a comparison of the orientation was possible. However, we found no significant difference in construction costs for vertical and horizontal racks/screens. The first step of our analysis only considered how these factors affect construction costs, but mitigation can also entail maintenance costs and power losses. Power losses can be difficult to quantify. Thus, upfront construction costs may be assumed to account for the largest share, which may lead decision makers to underestimate the total costs of fish passage mitigation if power losses represent a large share of lifetime costs.
For upstream measures, we hypothesized that capital (construction) costs and power losses account for the largest shares of lifetime costs. Under almost all price scenarios, capital costs accounted for the largest share. However, in the high price scenario for vertical slot passes, power losses accounted for a greater share. This means that when electricity prices are high, power losses may represent a significant concern for operators. However, it is difficult to generalize findings about power losses of measures as there may be significant variation in power loss among sites. Fish passage facilities may be designed to operate only during the migration season [50]. Thus, they do not necessarily operate year-round. Smart management of water release should ideally be adapted to both fish migration and power production needs. This could potentially reduce power and income loss while still maintaining fish passage. Such solutions depend on site-specific conditions, fish species composition, legal requirements, and operational constraints. Our finding is significant because previous cost reports of fish passage measures primarily focused on construction and operation [8], rather than quantifying the role of power losses. As a result of lack of data, a comparison of lifetime costs for downstream passage measures was not possible. Thus, it represents a potential avenue of future research.
Both of these analyses allow us to compare the costs of different types of upstream passage measures. We hypothesized that technical designs are often favored since less space is needed, which incurs fewer costs related to land acquisition. This can be most severe in congested urban areas of Europe. Our analysis of construction costs, however, showed that nature-like fish passes were comparatively cheaper, even when controlling for interactions between nature-like builds and the length or height of the pass. We found that nature-like passes are less expensive than technical solutions, but that this difference diminishes as the pass increases in length. In contrast, combined solutions tend to be more expensive than technical ones, but this difference diminishes as the height of the obstacle increases. This finding was consistent with the analysis of lifetime costs.
When considering lifetime costs, nature-like solutions incurred relatively fewer costs, specifically, fewer power losses compared to combined and technical solutions. When comparing combined and technical solutions, combined solutions were more expensive. This may be because combined solutions are often built when the height is too great for the available space. When compared across sites, natural materials may be cheaper for construction, particularly if autochthonous materials with fewer transportation costs are used [51]. Given that our results showed that nature-like solutions cost less to build and operate, incur fewer power losses, and provide habitat [52] in addition to facilitating fish passage, there is a strong basis for supporting their development in Europe.
Our study also yielded differences in the monitoring costs of upstream passage measures. The lack of data related to the costs and types of monitoring is evident. Approximately one-third of the case studies reported monitoring costs. However, only eight reported the type and five reported the timeline. While it is unclear whether the case studies simply did not report the costs or whether no monitoring took place, recent studies have noted the dearth of effectiveness monitoring related to fish passage measures [53]. In contrast, a study of river restoration measures showed that 80% of the projects are monitored [54]. This may stem from the belief that standardized passage design based on species-specific formulas [55] ensures success. Hence, additional monitoring is often not considered necessary and may result in additional costs if functionality is not proven after all. Recent literature argued that most design criteria are primarily based on salmonids in the Northern Hemisphere, which fails to account for the variation in individuals, populations, and species at individual sites [27].
Our data reveal that when a complex technical solution is built (e.g., fish lift), operators often invest in intensive monitoring. In turn, simpler technical measures (e.g., vertical slot pass) may require less monitoring as proven solutions do not need to be tested. In comparison to vertical slot passes, which can be standardized, nature-like solutions like roughened bypass channels tend to be adapted to local landscape conditions, which may also result in additional monitoring expenses. Nature-like passes may also be more complicated to monitor without standardized and pre-fabricated monitoring gear.
Additionally, there is an array of possible indicators for passage effectiveness. While environmental variables such as temperature, discharge, water depth, and water velocities are regularly used for evaluating passage mechanisms, they are not suitable for cross-site comparisons as they cannot imply a cause and effect [53]. However, detailed reporting of hydraulic and structural parameters may make studies more comparable across sites [53]. Given the high costs of fish passage restoration, assessing the costs and ecological benefits is important to develop evidence-based solutions in the future [55,56]. Thus, we recommend that policy makers standardize monitoring and reporting of the hydraulic, structural, and biological parameters as well as costs in a much more detailed manner.
Future analysis of upstream fish passage costs can be strengthened by clear reporting of structural characteristics. For many variables, data were missing, which reduced the observations in a complete case analysis. Further, it was often unclear how the variables were reported. For example, sometimes the reported length was the shortest distance between the entrance and exit of a fish pass, while other times it was the true length of the fish pass. In the analysis, it was not possible to account for such data inconsistencies as it was not always clear how operators reported length. Additional information about the target fish species may also help to account for the variation observed, as technical and nature-like designs may be built for different types of target species. While we did not have information about the target species for all measures, we found that there were no major differences in the target species for technical or nature-like passes (i.e., technical fish passes were not specifically built for stronger swimmers).
For downstream fish passage measures, our analysis was largely limited by missing data, especially related to maintenance and power losses. These costs can be difficult to identify as some downstream measures may be installed at hydropower plants for reasons unrelated to fish (e.g., trash rack to repel debris). However, if they are designed properly, they can serve the dual purpose of protecting fish. Thus, we recommend that future research focus on the costs of downstream passage facilities. Finally, future research can conduct cost-benefit or cost-effectiveness analyses. While we found that nature-like solutions tend to cost less, it will be important to quantify the possible additional ecological benefits (e.g., providing habitat) to assess their value relative to other types of fish passage measures.
Given the costs, the question arises of how the sustainable management of hydropower and fish passage can be financed and how effective solutions can be incentivized. Present financing instruments include support schemes, feed-in tariffs and green power labels [12]. While support schemes compensate for the modernization of existing plants through direct financing, grants, or loans, feed-in tariffs guarantee a fixed price for renewable energy fed into the public grid and green power labels set minimum standards related to electricity production [12]. Specific to passage restoration, direct financing and loans may cover construction costs, whereas feed-in tariffs may offset the recurring costs associated with power losses.
In some countries, mitigation is financed by state authorities using taxes. In Switzerland, for example, all plants built before 2011 are eligible for financing of upgrades. Ecological improvements are supported by a tax of 0.1 Rappen/kWh on all consumer electricity bills, which amounts to the "Swiss Grid Fund" with a maximum of 50 million CHF/year. To incentivize upgrades, the Swiss Grid Fund is available on a first-come-first-serve basis until the program expires in 2030. After the program's end, operators will be fully responsible for financing, which is in line with the polluter-pays principle [57]. In contrast, Austria supports approximately 50% through national funding such as the Environmental Aid Act. However, this excludes losses stemming from electricity production [57,58]. In Sweden, green power labels are used to incentivize ecological restoration. For operators to qualify for green power labels, they must make annual contributions to environmental projects [12]. The approach of targeted financial contributions may ensure that measures are carried out at sites with the greatest need, which could maximize effectiveness.
In some cases, the amount of funding provided is assessed on a case-by-case basis, which can be time-consuming and costly. On the other hand, arguments can be made for a simpler system based on standard refunds determined by the type, length, and height of the measure or a percentage of costs. While simplification is desirable, the results here do not support a standardization of costs. Within each category, the standard deviation is often greater than the mean, which indicates a wide range of possible costs.
To offset production losses resulting from fish passage measures, feed-in tariffs may be useful. Feed-in tariffs may be conditional on improvements as in Switzerland, France, and Germany [12]. Typically, different rates apply, based on the size of the plant. The larger the plant, the lower the rate. However, given the fluctuations in time of lost power production and market rates, our analysis cannot assess whether feed-in tariff rates are sufficient to offset the losses of power production. When establishing priorities related to financing of fish passage measures for hydropower, decision makers must weigh these factors with societal and political values.

Conclusions
As sustainable management of hydropower must balance mitigation measures for fish with renewable energy production, decision makers require information about cost trade-offs. This systematic overview detailed non-recurring (construction) and recurring (maintenance, monitoring, power losses) costs of passage measures.
Construction tends to account for the largest share of lifetime costs and while unit metrics (height and length) can be used for estimation of upstream fish passage measures, one should be aware that planning costs often differ substantially from actual costs. In particular, factors such as the target species, site conditions, difficulty of site accessibility, and local regulations contribute to the overall costs of mitigation measures. Comparatively, maintenance accounts for a relatively small share of lifetime costs. As there is limited research on power losses related to fish passage measures, our findings provide a basis for including economic aspects in hydropower decision-making. Under the high price scenario for vertical slot passes, power losses exceed construction as a share of lifetime costs. In most markets, power prices vary significantly over the year, week, and day. Thus, balanced solutions that consider both ecological aspects such as key migration seasons of target species as well as power prices are likely to create win-win situations.
When comparing technical and nature-like upstream passage measures, nature-like measures tend to incur lower costs, even when considering power losses. Given that nature-like solutions cost less to build and operate, incur fewer power losses and provide habitat in addition to facilitating fish passage, there is a strong basis for supporting their development in Europe.
In addition to the limited information on costs, there tends to be limited monitoring of mitigation measures and reporting of their costs, which makes it difficult to explore their efficiency after implementation and enable statements about cost-effectiveness. Thus, we recommend that policy makers standardize monitoring and reporting of hydraulic, structural, and biological parameters as well as costs in a much more systematic and detailed manner.      Table A2. Confidence intervals (95%) for the three upstream models presented in Table 3.