Estimating the economic value of hydropeaking externalities in regulated rivers

Hydropower is a flexible form of electricity generation providing both baseload and balancing power to accommodate intermittent renewables in the energy mix. However, hydropower also generates various externalities. This study investigates individuals’ preferences for policies aiming to reduce short-term regulations (i.e., hydropeaking in regulated rivers) while accounting for associated externalities with a discrete choice experiment. This is the first valuation study focusing on hydropeaking that considers both negative and positive externalities. The results imply that most individuals prefer stronger restrictions on short-term regulations to mitigate local environmental impacts. Individuals especially value improvements in recreational use, fish stocks, and the ecological state. On the other hand, potential increases in CO 2 emissions are linked with a clear disutility. The estimated benefits obtained from an improved state of the river environment due to such restrictions probably exceed the disutility caused by increased CO 2 emissions. The results also reveal unobserved preference heterogeneity among individuals, which should be accounted for in the willingness-to-pay (WTP) estimation using a model specification with correlated utility coefficients. Overall, the findings can inform policy-makers and environmental managers on the economic value of hydropeaking externalities and further guide the sustainable management of rivers regulated for hydropower generation.


Introduction
Global climate change mitigation measures and the electrification of economies are changing power markets to include an expanding share of renewable energy sources.The intermittence of these sources increases the need for balancing power in the local, national, and regional production and electricity markets.Hydropower is a significant and highly valuable source of flexible electricity that balances the increasingly fluctuating supply at sub-daily scales to meet electricity demand (Gaudard and Romerio 2014;Vardanyan and Hesamzadeh 2017).Thus, significant pressure exists to utilize regulated rivers more efficiently to balance the electricity market.This development has already altered the hydropower operating regimes by increasing sub-daily flow variation, an effect referred to as hydropeaking (Carolli et al. 2016), in the Nordic countries (Ashraf et al. 2018).Hydropeaking at sub-daily scales is likely to intensify further in the future as the share of intermittent renewable energy sources is expected to rise.
Hydropower production typically generates various externalities (Mattmann et al. 2016).This study focuses on externalities caused by hydropeaking.Hydropeaking produces artificial flow patterns involving high variability in discharge, water level, and flow velocity with substantial environmental effects.For example, hydropeaking results in changes in sedimentation levels and water temperature, negatively affecting invertebrates (Bruno et al. 2010) and fish (Bartoň et al. 2022;Casas-Mulet et al. 2016).It also causes erosion, degrading the river's ecological status (Moreira et al. 2019).Diverse recreational services, such as swimming, boating, kayaking, angling, skiing, or snowmobiling on river ice may also be affected by altered flow patterns as the water level and velocity change quickly.The influence of hydropeaking on recreational opportunities may be negative or positive depending, for instance, on the recreational activity and river section (Hynes and Hanley 2006;Håkansson 2009;Getzner 2015).A positive environmental externality of hydropeaking is lower greenhouse gas emissions compared to most other electricity production forms (Weisser 2007).
Several previous valuation studies have investigated hydropower externalities, as summarized in a meta-analysis by Mattmann et al. (2016).The literature has often examined the environmental impacts of new hydropower projects (Hynes and Hanley 2006;Han et al. 2008;Botelho et al. 2015;Tabi and Wustenhagen 2017), but there are also studies on the impacts of removing existing dams (Loomis 1996;2002;Robbins and Lewis 2009).Closest to our study are the ones on alternative operation regimes of hydropower generation and other remedial measures for improving the ecological status of regulated rivers.For example, Kataria (2009) estimated the value of environmental improvements in a hydropower-regulated river in Sweden, focusing on the river's ecological status using a discrete choice experiment (DCE).The results show that measures that enhance the conditions for environmental attributes have a significant welfare increasing impact.In another study, Jones et al. (2016) examined individual preferences for operational changes in flow regimes to obtain environmental improvements in a regulated Colorado river in the US using a contingent valuation.The findings indicate that preferences are highly sensitive to information about additional value dimensions beyond downstream environmental flow impacts, such as effects on rural communities and air emissions.
In this study, we investigated individual preferences for policies aiming to reduce hydropeaking while accounting for associated externalities with a DCE.To our knowledge, this is the first valuation study focusing on hydropeaking.Our study also differs from previous research in attribute selection because we considered both positive and negative hydropower externalities and their tradeoffs.Our main objective was to determine how people perceive and value the effects of hydropeaking on the river's recreational use, ecological status, fish stocks, and carbon dioxide (CO2) emissions.We examined these issues in a large river system, the Kemijoki River in Finland.The Kemijoki River is a very interesting and relevant river to study the effects of hydropeaking because it is the largest source of hydropower in Finland, directly influencing the balance between negative and positive externalities.
Our specific research questions are as follows: (i) What is the relative importance of the considered attributes in their contribution to willingness-to-pay (WTP) estimates?(ii) How are individualspecific factors associated with individuals' preferences?(iii) What are the welfare effects, in terms of compensating variation, of alternative policy scenarios to restrict hydropeaking?These are very policy-relevant questions because the increasing changes in the sub-daily flow regimes deteriorate river ecosystem and river's recreational services; therefore, environmental managers and policymakers need to consider this new pressure on river environments.Understanding individuals' preferences for hydropeaking regulation and environmental effects is crucial to inform an efficient and sustainable energy transition from fossil fuels to renewable energy sources.As a methodological insight for DCE modeling, we tested the technique proposed by Mariel and Artabe (2020) for identifying behavioral correlations among utility coefficients and to investigate preference heterogeneity apart from scale heterogeneity.
The rest of the paper is organized in the following way.Section 2 presents the case study area, survey design, and empirical model.Section 3 focuses on the results, whereas Section 4 provides a discussion and the concluding remarks.

Case study area
Kemijoki is one of the largest rivers in Northern Europe and Finland, flowing through the cities of Sodankylä, Kemijärvi, Rovaniemi, and Kemi to the Gulf of Bothnia, Baltic Sea.The Kemijoki watershed (50,683 km 2 ) covers a significant portion of northern Finland (Figure 1).There are 16 hydropower plants on the main part of the Kemijoki River (Figure 1).Hydropower production was 4,131 GWh in 2020, accounting for approximately 6% of all electricity production in Finland.The nominal capacity of the Kemijoki River power plants is 1,098 MW, and seven power plants of the main part of the Kemijoki River are among Finland's ten largest hydropower plants.In addition, reservoirs have been built upstream, and Lake Kemijärvi is regulated (Figure 1).
According to the classification of the European Water Framework Directive (2020/60/EC), the main channels of the Kemijoki River and Lake Kemijärvi are both designated as heavily modified.The modified and regulated main river system and tributaries include 1,000 km 2 of lake basins and 685 km of river channels (Räinä et al. 2022).A recent ecological status assessment showed that the environmental goal of good ecological potential was reached only upstream of Lake Kemijärvi, whereas the middle and lower stretches were classified as moderate (Räinä et al. 2022).The main reason for lowering the ecological status is the lack of fish passages in the main channel, which would allow a bypass to non-constructed spawning grounds on the Kemijoki River.Overall, the building of hydrodams and regulations have changed waterbodies and discharge conditions in the watershed (Ashraf et al. 2016).Construction has destroyed the rapid areas almost totally; the only remaining rapids are situated between the Valajaskoski and Vanttauskoski power plants, where the Sierilä power plant is planned and in the licensing phase (see Figure 1).Regulation and water fluctuation also influence the riparian zone, with a reduction in protective littoral vegetation causing recruitment losses among fish species (Räinä et al. 2022).

Survey development and data
The design of the DCE started by reviewing the literature to identify potential attributes (Bergmann et al. 2006;Han et al. 2008;Kataria 2009;Botelho et al. 2015;Klinglmair et al. 2015;Tabi and Wustenhangen 2017).Then, from discussions with researchers and experts, we constructed the current and alternative policy situations and determined the most relevant attributes and their levels.
The development of the DCE lasted approximately one year and included several workshops and rounds of commentary.The survey and the DCE were also presented to the local hydropower operator and environmental regulator.Based on the feedback received, the survey and DCE were developed further.The pilot survey was conducted in April 2021.We collected responses and feedback from 27 individuals with different backgrounds.The pilot round enabled us to pretest the understandability and credibility of the survey questions and the DCE, conduct some preliminary analysis, and refine the experimental design.
The final survey was executed in June 2021.The survey was targeted to (i) 2,500 individuals living in the municipalities in the Kemijoki watershed1 and (ii) 1,500 individuals living in other parts of Finland.The participants were randomly drawn from the civil registry's database, and their ages varied between 18 and 80 years.The participants living in the Kemijoki watershed area were sent a printed questionnaire by ordinary mail that included instructions on how to respond to the online version, whereas the participants living outside the study area were sent an invitation by mail with instructions on how to respond to the online survey.The participants were incentivized to respond by being offered a chance to win a €100 voucher to a grocery store.
We received 396 responses, of which 264 were printed questionnaires and 132 were online responses.
This resulted in a response rate of 9.9%.The response rate was higher among individuals living in the study area than among those living outside the study area.Overall, the response rate was rather low; several factors may explain it.For instance, the difficulty of the subject matter may have affected it.In addition, no reminder rounds were conducted, and individuals living outside the study area were only given with the possibility of answering the online survey due to budgetary constraints.
The respondents were representative of household size (Table 1).Moreover, the division of respondents between rural and urban areas was quite close to the national average.However, the respondents were somewhat older and included more men than the original random sample.The collected sample was also slightly more educated than average for the Finnish population of those over 15 years old.However, this difference is mitigated if we account for the fact that the collected sample only included individuals aged between 18 and 80. Unfortunately, the income distribution is not publicly available with the presented division.We acknowledge that the collected sample is likely to suffer from non-response bias, which warrants to be considered in the DCE analysis.(Official Statistics of Finland, 2020).c: Population refers to the GIS-based urban-rural classification for Finland and the Finnish population (Helminen et al., 2020).NA = Not available

Discrete choice experiment design
The DCE included six hypothetical choice tasks.Before respondents started to answer the choice tasks, the attributes were described to them, and follow-up questions were used to ensure that they had familiarized themselves with the descriptions.Each choice task included three alternatives: one corresponded to the current situation, and two others presented hypothetical policy situations (see an example choice task in Appendix A).Six attributes described the choice alternatives: short-term regulations of hydropower production, recreational use, the ecological state, fish stocks, CO2 emissions, and an increase in the annual electricity bill.Table 2 summarizes these attributes and their levels.The choice and description of the attributes are justified in the next paragraphs.According to the current permit conditions (i.e., the level "Current regulations"), short-term regulations can be utilized efficiently in the hydropower production of the Kemijoki River.This causes strong intra-day and weekly variations in water flow and surface elevation.For example, during low flow, the shoreline is exposed, whereas during high flow, the water level rises, and the water flow rate is high.The intensity of the effects varies in different parts of the river.The effects are typically stronger near hydropower plants.In the future, the Kemijoki River could have stronger restrictions on short-term regulations and reduce harmful local environmental impacts. 2 The level "Somewhat more restrictions" refers to a situation where the short-term hydropower regulation will be moderately limited.Here, fluctuations in water flow and surface height within a day are smaller than under current regulations.The flow variation is reduced by approximately 50% compared to the present circumstances.The level "A lot more restrictions" means that the short-term regulations of hydropower will be severely restricted, clearly reducing hydropeaking use.Here, intra-day variations in flow and water level are hardly observed.On a weekly basis, some variation in water level is noted.
The flow variation is reduced by about 80% compared to today.
Short-term regulations of hydropower may affect recreational use of the Kemijoki River.Flow fluctuations influence, among other things, the quality and usability of the beaches and safety for swimming, paddling, boating and fishing.Due to the strong variation in flow, fish move more, and angling may become more difficult.Fluctuations in the water level can damage fixed structures such as piers.Variation in water height and flow also impacts the durability of the ice cover in winter, impairing safe movement on ice.At present, fishing, swimming, and other water activities can be tricky in certain places (see the level "Current state").The use of fixed beach structures can be difficult, and harm can result.In wintertime, ice cover can be weakened and become unpredictable in certain spots.The level "Improves somewhat" means that fishing becomes slightly easier.In addition, bathing and navigating the waters is easier and safer.The use of fixed structures on beaches is a little easier, and they are less likely to be damaged.The ice cover is a bit more durable in the winter.
Related to the level "Improves a lot," flow changes are more predictable, making swimming and navigating the waters easy and safe.The fixed structures on the beach are not harmed and are more functional.In winter, the ice cover is stronger and safer.
The short-term regulations of hydropower affect the ecological state of the Kemijoki River environment.The ecological state in this study is interpreted as the quality of the river's habitats and the abundance of plants and benthos that live there.Strong short-term regulations increase riverbed 2 The respondents were informed that any new restrictions would not apply to seasonal variation or flood protections.
and riparian wear and weaken natural benthic and plant communities.In the current state, the benthic fauna and flora of the river suffer from short-term regulations, and solids are released into the water.
At the level "Improves somewhat," the living conditions of benthic animals and plants in the river are assumed to improve slightly; communities will be more abundant and fewer solids are released into the water, whereas "Improves a lot" indicates that the river's benthic and plant communities are close to the natural state and there are few solids in the water, in accordance with the Water Framework Directive (2000/60/EC).
Short-term regulations have a detrimental effect on the diversity and abundance of fish species in the Kemijoki River.Rapid and strong flow fluctuations reduce the abundance of almost all naturally reproducing fish species and degrade their habitats.Also, the stronger short-term regulations are, the less successful the stocking of artificially reared fish will be.Currently, common species in the slowflowing river sections and lakes include pike, perch, pikeperch, roach, stocked brown trout, and stocked and introduced non-native rainbow trout.Species dependent on or benefiting from fastflowing river habitat (grayling, whitefish) are scarce, and their natural reproduction is very low.If the state of fish stocks would "Somewhat improve," fish species such as grayling would likely become more common, and natural reproduction would strengthen.Other fish species would also become more abundant, and stocking of artificially reared species is more successful.Notably, this attribute had only a moderate improvement level because expert evaluations revealed that it was not realistic to assume higher improvements in fish stocks.Baltic salmon, which was highly abundant in the river before its damming, was not included in this attribute because the number of salmon in the Kemijoki River cannot be increased by restricting short-term regulations without building well-functioning fish passages next to dams (which are currently lacking).
Restricting the short-term regulations of hydropower production on the Kemijoki River would increase the CO2 emissions of Finnish electricity production.Hydropower generates low-emissions electricity and is especially useful when no intermittent wind power generation is available.Usually, alternative production forms (e.g., combined heat and power) cause more emissions than hydropower.
When estimating potential changes in greenhouse gas emissions and consumers' billing costs, we solved the optimal water reservoir operation of the hydropower generator, scaled to match the aggregated generation capacity of the Kemijoki River.Using optimal policies, we modeled the reservoir allocation under the regulation scenarios "Current regulations," "Somewhat more restrictions," and "A lot more restrictions."It is assumed that the restricted hydropower output is compensated for with other dispatchable electricity generation sources.In other words, we now have three alternative scenarios where generation profiles differ: (i) the original; (ii) generation with somewhat more restrictions on hydropower output; and (iii) generation with many more restrictions on hydropower output.
To investigate the relationship between electricity generation (without hydropower) and CO2 emissions, we estimated a regression model explaining the variation in emissions with linear and quadratic generation variables.The synthetic generation profiles (ii and iii) are used as inputs when creating emissions profiles in scenarios where more restrictions on hydropower generation are imposed.The results indicate that CO2 emissions rise by 2.1 and 2.7% in the considered scenarios compared to the original emissions.Thus, the increase in CO2 emissions was 0%, 2%, or 4% in the DCE.
In addition to having an impact on emissions, significant changes in the outputs of large hydropower plants are likely to increase market prices and consumers' billing costs.For instance, if short-term regulations on hydropower are restricted, additional, more expensive balancing resources, such as gas turbines or electricity storage, are needed.The price impact from restrictions on hydropower generation was separated into two parts.First, we estimated the cost from added fuel usage (peat, natural gas, and coal) and emissions permits (€50 per ton of CO2) with fuel-specific emissions factors (peat: 381.2, hard coal: 340.6, natural gas: 198.1 gCO2/kWh) in the day-ahead market.It was assumed that the added electricity generation would utilize the fuel mix employed in electricity generation in Finland in 2019 (peat: 4%, hard coal: 6%, natural gas: 6%) and that the generation efficiency would be 40%.
Since the restrictions also affect short-term regulations, more balancing power resources are needed.
It was assumed that additional balancing resources would be supplied purely with battery storage (€200 per kWh of storage capacity) because gas turbines were not seen as feasible from environmental or economic perspectives in the long term.The estimated additional costs per household were €5 in the scenario with somewhat more restrictions on hydropower output, and €6.7 when many more restrictions were put into action.To have higher variation across choice tasks in the DCE, the potential increases in the electricity bill varied between €0 and €105.
The DCE design was conducted with a Bayesian D-efficient design consisting of 36 choice tasks that were further divided into six blocks to minimize the respondent's burden.The prior parameter values for the design were obtained from the pilot study.To create feasible choice tasks, we added constraints to the design.Whenever the alternative in the choice task had the short-term regulations of hydropower attributed to the "A lot more restrictions" level and the other alternative to the "Somewhat more restrictions" level, then the attribute levels of the ecological state, recreational use, and emissions were either better or at least equal in the former alternative compared to the latter.
These restrictions were identified as important for the credibility of the choice alternatives during survey testing.

Empirical model
In the frequently used mixed logit (MXL) model (Train 2009), the utility for individual n, related to choice alternative j, is represented as: where   is a vector of attributes including an alternative specific constant (ASC), and   is the corresponding vector of estimated parameters.The idiosyncratic error   is independently and identically distributed and an extreme value one (EV1) type.  also includes random taste parameters that depend on the values of the population mean b and covariance matrix Ω of an underlying distribution φ(|b, Ω).
The utility specification is presented in the preference space in Equation ( 1).However, we reparameterize the utility to the WTP space in which the estimated coefficients can be interpreted as marginal WTP values (Train and Weeks 2005).In the WTP space, the utility for individual n is   = (    +   ′   ) +   =   (  + ′    /  ) +   = (    +  ′    ) +   . (2) In Equation ( 2),   denotes the monetary attribute and   represents the estimated parameter for it,   includes non-monetary attributes and   refers to the corresponding vector of estimated parameters.Above, we also have   =   /  , a vector of marginal WTP for each nonmonetary attribute.The scale parameter  is normalized to 1.The WTP space specification enables convenient distributions for WTP because it avoids the need to consider the distribution of inverse coefficients (Daly et al. 2012).
In this study, non-monetary attributes were treated as random and assigned normal distributions, whereas monetary attributes were treated as random with lognormal distributions.The modeling was executed with 10,000 Sobol draws with random linear scramble and random digital shift.
Analytically derived gradients were used.The estimations were conducted in MATLAB 3 .
In the analysis, we used the MXL model in the preference and WTP space with and without correlation.Overall, the MXL model that allows all parameters to be randomly distributed, and which estimates a full covariance matrix among them, is the most general form possible (Hess and Train 2017;Mariel and Artabe 2020).We allowed for a correlation between random coefficients because it was likely that the unobserved effects between different attributes and attribute levels would be correlated.Such a correlation is taken to reflect that an individual's preferences for one attribute are related to the preferences for another attribute.For instance, individuals who support increases in recreational possibilities can also be supportive of 3 We utilized the DCE package, available here.
improvements in the fish stocks and/or ecological state.Alternatively, individuals who prefer improvements in the ecological state can also dislike increases in CO2 emissions.
To test these behavioral hypotheses and interpret the obtained correlation matrix, we utilized the procedure proposed by Mariel and Artabe (2020).This procedure may help to disentangle behavioral interlinkages from scale heterogeneity.The correlation matrix captures not only the correlation between the random parameters, but also the correlation caused by the scale heterogeneity that cannot be identified separately (Hess and Rose 2012;Hess and Train 2017).
The proposed procedure consists of two simple steps.First, the signs of the attributes corresponding to the utility that have a negative mean coefficient are reversed.Then, only negative correlations are interpreted.This enables to identify correlation resulting from a behavioral phenomenon.
To derive a welfare change estimate for policy scenarios describing the short-term regulations of hydropower production and associated effects, we employed the compensating variation (CV) from Hanemann (1982): where  0 and  1 are the utility expressions for the current and alternative policy scenarios.
, the CV is calculated using the WTP space specification outcomes as follows: 3. Results

General opinions on hydropower
The survey included several hydropower-related claims, which we measured on a 5-point Likert scale (Figure 2).A vast majority (81%) of the respondents either strongly or somewhat agree that hydropower causes significant damage to migratory fish stocks.Hydropower is also clearly more negative than positive for recreational fishing (62% vs. 22%).In addition, 48% of the respondents strongly or somewhat agree that hydropower destroys riparian vegetation, and 53% find hydropower plants harmful from a landscape point of view.h) Hydropower will be an important form of energy production in the future.i) Hydropower causes significant damage to biodiversity.
j) The value of the land decreases due to hydropower.
k) Hydropower enables low-cost electricity.
l) Hydropower improves opportunities for recreational use (e.g., swimming and boating).
Evaluate how the claims about possible effects of hydropower match your opinions.

Strongly disagree Somewhat disagree
Neither agree nor disagree Somewhat agree Strongly agree Do not know NA On the other hand, most of the respondents think that hydropower is a low-emissions form of electricity generation (74% strongly or somewhat agree), that it increases the security of supply in the energy system (66% strongly or somewhat agree), that the municipal economy benefits from it (59% strongly or somewhat agree), and that it is an important energy production form of the future (56% strongly or somewhat agrees).The sample is clearly split with regard to the perceived effects on biodiversity, recreational use, land value, and electricity prices.Furthermore, the share of "do not know" answers was the highest related to impacts on land value, riparian vegetation, and the municipal economy.Overall, the findings indicate that the public is likely to have varying opinions on hydropower and its externalities.This is also reflected in the response distribution when asking whether the respondents disliked hydropower: 55% of those providing an answer to this claim either strongly or somewhat disagreed, 13% were indifferent, and 29% strongly or somewhat agreed with the claim.
The respondents were also presented with a series of questions about the mitigation measures and compensation requirements of the negative effects of hydropower (Figure 3).Over 70% of respondents thought that negative externalities exist, which should be either compensated for or offset, and nearly as many stated that hydropower companies should take more measures in compensation activities.Additionally, 59% are willing to see higher monetary compensation.
However, the respondents were divided based on whether the state should participate in compensating for the disadvantages of hydropower production.Interestingly, public support exists for ecological compensation, as 51% of the respondents either strongly or somewhat agreed that the environmental impacts of large hydropower plants could be compensated for by dismantling small power plants on other rivers 4 .
4 Some small hydropower plants have already been dismantled in Finland to restore river ecosystems.In addition, it has been discussed whether larger hydropower units could compensate for their negative environmental impacts by dismantling small power plants on other rivers.

Valuation of hydropeaking externalities
Table 3 outlines the definitions of the attribute variables used in the discrete choice analysis.Current state levels for recreational use, the ecological state, and fish stocks, as well as no change level for CO2 emissions, serve as reference categories.The MXL model with correlation outperformed the MXL model without correlation based on LL, McFadden pseudo R 2 , and AIC (see Appendix B), thus suggesting that using the MXL model without a correlation could lead to biased estimation of the WTP values.Note that 20 respondents were excluded from the analysis due to missing responses to at least one of the choice tasks.0% 10% 20% 30% 40% 50% 60% 70% 80% 90% 100% a) The disadvantages of hydropower are so small that they do not need to be compensated or offset.
b) The state should compensate the disadvantages of hydropower for the victims.
c) The environmental impact of large built river hydropower plants could be offset by the dismantling of power plant dams on small rivers.d) Hydropower companies should pay higher compensation for the disadvantages of hydropower.e) Hydropower companies should take more measures to compensate for the disadvantages of hydropower.
Evaluate how the claims about mitigation of hydropower's negative effects and compensation match your opinions.The results of the DCE are presented in Table 4. Models 1 and 2 are otherwise similar except for interactions with the ASC.The results of the MXL models (1 and 2) in the WTP space show that the variables have their expected signs.The coefficient of the ASC is negative, implying that respondents preferred the presented alternative policy situations over the status quo.Note, however, that the ASC captures both the utility related to the status quo alternative and the omitted dummy for the short-term regulations attribute, level REGUL_M.The coefficient for REGUL_H is statistically insignificant, meaning that the respondents were, on average, indifferent to the alternative short-term regulations, including more restrictions.Our interpretation is that the intensity of restrictions as such is not as important as the resulting effects are.
The coefficients for improvements in recreational use and the ecological state are all positive and statistically significant.Moreover, the respondents derived higher value for recreational use than the ecological state and witnessed clear improvements in both attributes (RECREA_H and ECOLO_H).
In Model 1, the WTP for RECREA_M is €32  Preference heterogeneity exists among respondents with respect to ASC values.To investigate possible reasons and to control for potential representativeness issues (see Table 1), we introduced interactions between the status quo and respondents' demographic traits.This informed us whether age, higher education, gender, or living place affect the choice between the status quo and alternative river management scenarios.The weakly and statistically significant positive interaction between a respondent's age (AGE) and the ASC denotes that the choice probability of the status quo is higher among older respondents.This implies that a younger age is linked with higher valuation of more restrictive hydropower operations.In addition, the probability of choosing the status quo falls if the respondent is a male.Interestingly, we observed that living in the Kemijoki River watershed area (LOCAL) is not associated with a lower probability of choosing the status quo.This finding indicates that, on average, familiarity with the study area is not linked to a higher valuation of alternative regulation regimes.Furthermore, a high education level (HIGHEDU) does not explain the probability of choosing the status quo alternative.

Policy scenarios
We created four scenarios to illustrate how environmental externalities affect people's economic welfare in terms of CV (Table 5).Policy scenarios 1 to 4 represent policies with stronger short-term regulations of hydropower production compared with the current regulations.Note that the likelihood of reaching the "Improves a lot" level in recreational use and the ecological state is higher in the case of high restrictions on hydropeaking (REGUL_H=1) than in the case of moderate restrictions (REGUL_M=1).All policy scenarios can generate moderate improvement in fish stocks.On the other hand, it is likely that the CO2 emissions will increase more in scenario 3 than in scenario 2 with the current energy production mix.Policy scenarios 3 and 4 illustrate welfare effects associated with a decarbonized power system, where hydropeaking is replaced with low-carbon production, enabling a 0% increase in CO2 emissions.The CV was calculated (see Eq. ( 4)) using the mean WTP values from Model 1 in Table 4.In Table 5, a positive CV indicates a WTP for the presented policy scenario.All alternative policy scenarios induce positive welfare.Policy scenario 1, involving moderate restrictions on hydropeaking, results in greater welfare (€14) than in scenario 2, including high restrictions on hydropeaking (€3).This finding stems particularly from the high compensation requirement related to the 4% increase in CO2 emissions.Scenario 4, with high restrictions on hydropeaking and decarbonized power systems, yields the highest welfare of €62, whereas scenario 3, with moderate restrictions, results in welfare of €40.

Interpreting correlations
Next, we focused on interpreting the estimated correlation matrices of the random coefficients and tested the procedure proposed by Mariel and Artabe (2020).Table 6 presents the correlation matrix without signs reversed non-cost attributes, whereas Table 7 displays the signs in the reversed version.
The models were estimated in the preference space; the full model outcomes are available in Appendix D.
Note that the positive correlations between the coefficients of non-monetary attributes cannot be interpreted according to the proposed rule.Thus, the usage of the proposed procedure does not offer much valuable additional information on preference heterogeneity because a vast majority of the correlations were positive in this study.Based on the procedure, we interpreted only the negative correlations between the coefficients of EBILL and ECOL_M, EBILL and ECOL_H, EBILL and FISH_M, and EBILL and EMIS_4, as well as the coefficients of ASC and EMIS_2, ASC and EMIS_4, and ASC and EBILL.The interpretation of the first three correlations must be made with a reversed sign because the sign of the monetary attribute is reversed.A positive correlation in these cases means that people with a high coefficient for ECOL_M, ECOL_H, and FISH_M have a low monetary coefficient, i.e., a higher valuation of these attributes.This is an expected finding.A negative correlation between the EBILL and EMIS_4 indicates that people with a high negative coefficient for EMIS_4 have a high monetary coefficient, i.e., a lower compensation requirement for this attribute.In addition, a negative correlation between the ASC and emissions attributes (EMIS_2 and EMIS_4) implies that people who dislike the status quo alternative are, on average, more likely to accept increases in CO2 emissions.

Discussion and conclusion
The findings imply that most individuals prefer more rigorous restrictions on short-term hydropower regulation to mitigate local environmental impacts caused by hydropower generation in the Kemijoki River.In the DCE, people choose alternative policy situations more often than the current situation (64% vs. 36%).Moreover, people value fish stocks improvements, recreational use, and ecological conservation in that order.They are willing to pay an additional €29-54 per household per year in increased electricity bills to obtain improvements in these attributes.However, potential increases in CO2 emissions are associated with a clear disutility, demonstrating an obvious trade-off between local and global hydropower externalities.
The WTP value of approximately €32 for the moderate improvement in fish stocks reflects the importance of recreational fishing in the Kemijoki River and the existence of fish stocks and option values.The fish stocks attribute was defined for the respondents in the survey as not containing Baltic salmon, but some respondents may have considered salmon along with non-migrating fish species when assessing the importance of the fish stocks attribute.Thus, some respondents may have overestimated the importance of the fish stocks.Previous studies on regulated rivers with existing fish passages in northern Sweden have found significant values associated with salmon in these rivers (Håkansson 2008;2009).Likewise, recovering the natural life cycle of salmon is likely an important issue for many locals in the Kemijoki River area.Overall, the presence of fish stocks has been among the most valued attributes in previous studies on regulated rivers (Kataria 2009).
People place a relatively high value on improvements in recreational use in the Kemijoki River.
Similarly, Getzner (2015) found that recreational value is higher on free-flowing sections than on dammed stretches of rivers for diverse recreational activities on the Mur River in Austria, and Immerzeel et al. (2021) reported that recreation is among the most valued ecosystem services in Nordic catchments.On the other hand, recreational use might not be that important for all individuals.
Our findings provide some evidence for this, as respondents' preferences for improvements in recreational use were heterogeneous.In contrast to our findings, the value of recreational opportunities was clearly lower than the value of fish protections in regulated rivers in Bavaria in Germany (Venus and Saur 2022).One potential explanation for this difference is that the study by Venus and Saur (2022) includes the construction of fish passage structures, which likely increase the value of fish protections.
The value of the ecological state was significant, although it was the least valued attribute among the negative externalities considered in our study.Numerous prior investigations have also found a significant value for the ecological state of rivers (Hanley et al. 2006;Kataria 2009;Andreopoulos et al. 2015).Our results are also in line with studies that have considered more broadly defined ecological attributes, such as ecological impacts (Jones et al. 2017), fauna and flora (Botelho et al. 2015), and nature and landscapes (Klinglmair et al. 2015).In their meta-analysis on the external effects of hydropower, Mattmann et al. (2016) found significant evidence for public aversion toward deterioration of the landscape, vegetation, and wildlife caused by hydropower projects, but only weak evidence of WTP for mitigating harmful effects.It is possible that people living near hydropowerregulated rivers are accustomed to the river's ecological state and hence do not value its improvement as much.In addition, consistent with risk aversion (Kahneman and Tversky 1979;Tversky and Kahneman 1991), people tend to value deterioration in absolute terms more than improvement in an attribute (Ahtiainen et al. 2015;Juutinen et al. 2021).
For the Kemijoki River, we found that people significantly value the mitigation of greenhouse gas emissions.Previous studies have obtained similar outcomes (Kosenius and Ollikainen 2013;Klinglmair et al. 2015;Jones et al. 2017).Mattmann et al.'s meta-analysis on the valuation of hydropower externalities (2016) revealed that reducing greenhouse gas emissions is valued positively and significantly more in countries with a higher share of hydropower in electricity production, probably because the people in these countries may have a greater level of awareness regarding the positive effect of hydropower on greenhouse gas emissions.In Finland, the share of hydropower is not especially high in the total energy mix, but there has been much public debate on the role of hydropower as a balancing source and in mitigating climate change.It is therefore likely that Finnish people are aware of hydropower's emissions mitigation potential.This is also supported by our findings, as a vast majority of respondents thought that hydropower is a low-emission form of electricity generation (Figure 2).
Our sample was not fully representative of the population.In particular, male respondents were overrepresented.We also found comparatively weak evidence that male respondents were more willing to accept hydropeaking regulation policies than females.Hence, to some extent, we may have overestimated the number of individuals who prefer stricter restrictions on hydropower generation in the Kemijoki River.In contrast, older respondents (overrepresented in our sample) were more willing to accept current policies, thereby tweaking potential bias in the opposite direction.Kataria (2009) also used these two individual-specific factors to explain preference heterogeneity for environmental improvements in hydropower-regulated rivers in Sweden, but their influence was not statistically significant.
To gain further insight into preference heterogeneity, we elaborated on the correlations among the utility coefficients.We expected that the correlations would have an influence on the results of our DCE, as all the attributes (excluding costs) described different environmental factors.For example, respondents valuing improvements in the ecological state would likely value improvements in fish stocks.Therefore, we used a model specification with correlated coefficients.The interpretation of correlations is not straightforward, however, as the correlation matrix of the utility coefficients captures both scale and behavioral heterogeneity (Hess and Train 2017).To identify the influence of the latter, we applied the procedure proposed by Mariel and Artabe (2020).In our data, this method did not help much in interpreting the correlations.Although the signs of the estimated correlations were as expected, the correlations were only negative in a few cases, enabling us to verify that they were due to behavioral phenomena.Notwithstanding, this issue requires more research, as model specifications with correlated parameters are increasingly used in the valuation literature (Marial and Artabe 2020).
Appendix B

Figure 2 .
Figure 2. Likert scale distribution of opinions on possible effects of hydropower (N=396).

Figure 3 .
Figure 3. Likert scale distribution of opinions on compensation and mitigation measures of the negative effects of hydropower (N=396).
The respondents required, on average, a €26 [± €7] compensation for a 2% increase in emissions, and a €60 [±9€] compensation for a 4% increase in emissions.As expected, the coefficient for EBILL is negative and significant.
and for ECOL_H it is €35 [± €8].The respondents also valued moderate improvements in fish stocks because the statistically significant WTP for FISH_M is €32 [± €5].The coefficients for EMIS_2 and EMIS_4 are both negative and significant, which reflects a dislike for increases in CO2 emissions.