A probability‐based model to quantify the impact of hydropeaking on habitat suitability in rivers

A negative effect of hydropower on river environment includes rapid changes in flow and habitat conditions. Any sudden flow change could force fish to move towards a refuge area in a short period of time, causing serious disturbances in the life cycle of the fish. A probability‐based model was developed to quantify the impact of hydropeaking on habitat suitability for two fish species, brown trout (Salamo trutta) and Grayling (Thymallus thymallus). The model used habitat preference curves, river velocity and depth to develop the suitability maps. The suitability maps reveal that habitat suitability deteriorates as flow increases in the studied part of the river. The probability model showed that, on average, suitability indices are higher for adult grayling than juvenile trout in hydropeaking events in this part of the river. The method developed shows the potential to be used in river management and the evaluation of hydropeaking impacts in river systems affected by hydropower.

temporary migration could increase competitive pressure in the now crowded habitats where fish are forced to move (Hansen, Marr, Lipton, Cacela, & Bergman, 2009). Thus, it is important to understand how sudden flow variations alter habitat suitability.
The hydropeaking regime influences fish migration, spawning, and nesting (Jones & Petreman, 2015). The response of fish in terms of habitat selection and movement has been investigated during hydropeaking and, as shown, the season, time of day or night, and ramping rate influence fish movement (Scruton et al., 2003). In addition, the life stages of fish play an important role in their chances of survival and response to sudden flow change Moreira et al., 2019;Schmutz et al., 2014). For example, grayling larvae and juveniles are more susceptible to stranding (Bardonnet & Persat, 1991) due to their size and behavioral response (Zitek, Schmutz, Unfer, & Ploner, 2004). Emerged brown trout are vulnerable to hydropeaking due to their limited swimming capabilities and larval fish may be displaced by sudden high flow (Jensen & Johnsen, 2002). Hydropeaking could also cause egg mortality as a result of redds dewatering (McMichael, Rakowski, James, & Lukas, 2005) and has been linked to juvenile fish mortality (Casas-Mulet, Saltveit, & Alfredsen, 2016;Maddock, Harby, Kemp, & Wood, 2013;. Reduced food supply, stranding, reduced recruitment, physiological stress, alteration of lifestyle and migration are listed as some of the consequences of hydropeaking (Bruno, Siviglia, Carolli, & Maiolini, 2012;Greimel et al., 2018;Harby & Noack, 2013;Nagrodski, Raby, Hasler, Taylor, & Cooke, 2012;Schmutz et al., 2014;Scruton et al., 2008;Young et al., 2011). Although studies show a temporal adaptation behavior in fish, even a single sudden flow change may result in drift or stranding risk. All listed above single effects accumulate in time as results of recurring flow fluctuations and can lead to the profound impact on river fish population. Therefore, numerous researches have been focused on mitigation of hydropeaking effects. Juárez, Adeva-Bustos, Alfredsen, & Dønnum, 2019).
A better understanding of the impact of hydropeaking on habitats and the response of fish to it are important steps in developing mitigation strategies. In this study, we developed a novel probability-based model to quantify the impact of hydropeaking on habitat suitability on a subdaily basis. The approach is designed to search for the habitat where fish could find refuge during sudden flow change. The model specifically measures the impact of sudden flow change on fish over 1 hr, which is considered a short period for fish to respond. The questions addressed in this paper include:

| The study area
The Kuusinkijoki River originates from Kiitämöjärvi lake of Kuusamo area in north-eastern Finland (Figure 1). It flows through Suininkijärvi and is the right tributary of the Oulankajoki River, with a length of 20 km. It has a yearly average flow of 10 m 3 /s, a peak flow of 60 m 3 /s and a catchment area of 1,020 km 2 . The Myllykoski hydropower plant on Kuusinkijoki was built from 1955 to 1956 and commissioned in 1957. The Kuusinkijoki River provides a spawning and growing environment for migrating brown trout (Salamo trutta) where they start their life and live until they migrate to Lake Pääjärvi. Grayling (Thymallus thymallus) are local fish species that live their whole life in the Kuusinkijoki River (Nykänen, 2004). Therefore, we considered these two species as the target species in the study.
We selected a 6 km river stretch downstream of the Myllykoski dam that encompasses various existing habitats and morphologic conditions for a detailed investigation of their characteristics (bathymetry and hydraulics). This investigation reach was chosen based on the importance of the stretch for our target fish species.
The model utilizes previously collected bathymetry and morphology data with a terrain file of 1 m Â 1 m resolution and a computational mesh size of 4 m Â 4 m. The river flow hydraulics were simulated for the dam outflow at 2, 3,4,5,6,8,10,12,14,17,20,25,28, and 35 m 3 /s for the entire reach, giving varying magnitudes of depth and velocity downstream. The results were exported as 1 m Â 1 m raster files for the river section and served as an input for our fish habitat model.

| Habitat probability model
The habitat probability model is a dynamic system that quantifies the impact of hydropeaking on a habitat. The habitat preference curves ( Figure 2a) for velocity and depth are applied to develop suitability maps using physical habitat simulation system (PHABSIM) (Bovee, 1986). The substrate was considered to be large gravel. The suitability maps are used to evaluate how habitat suitability is impacted by hydropeaking in a short time (i.e., 1 hr). In Nordic region, hydropower is traded in a decentralized energy market called "Nordpool" where energy is mostly traded hourly causing hydropeaking at hourly scale. Therefore, we have considered 1 hr in the model as this When the flow suddenly changes, some parts of the river may become unsuitable or less suitable habitats for fish. This forces fish to move towards a refuge area in a short time. The refuge area is defined as an area, which provides a better habitat with a higher suitability index. The suitability index maps SI i and SI f present habitat suitability before and after hydropeaking, respectively. For the model, the river channel and habitat suitability maps are gridded into the cells of 1 m by 1 m. The cells in SI i are the initial location where fish are assumed to be before hydropeaking. For every initial location (cell) in the SI i , the model searches through the SI f to find the cells with higher suitability indices. To limit the computational complexity of the study, the search area is limited to 500 separate areas at increasing distances from the initial location ( Figure 2c). Although the model could be adjusted to find areas with a certain suitability index, we did not set any boundary, as we wanted the model to find all the cells with a higher suitability index. The model determines the distance to the refuge area, swimming direction, and success probability.
The distance to the refuge area is calculated using the shortest path. The distribution of distance to the refuge area shows how the refuge areas are distributed around the initial location. The distribution includes four distance (d) intervals, d = 0 m, 1 m ≤ d < 2 m, 2 m ≤ d < 5 m, and d ≥ 5 m. The swimming direction is estimated by considering fish location before hydropeaking and the refuge area after hydropeaking using vector algebra.
The success probability (SPr) is defined as the probability at which fish could find the refuge area in each neighborhood (here 1-500, 2.4 | Selection of hydropeaking events for the probability model The Kuusinkijoki River flow from January 2018 to September 2020 is shown in Figure 3a. The annual flow regime is characterized by low winter flows, a rapid rise due to snow melt during May and June and low discharge during summer and autumn before the river freezes There are parts in the river which get watered as the flow increases. These parts, mainly close to the riverbanks, may put fish at risk of stranding once the flow decreases because a temporarily wetted habitat is a refuge area where fish is forced to move to during up-ramping (Hunter, 1992).
The suitability index probability distribution for the studied area,

| Distance to the refuge area
The distributions of distances to the refuge area are shown in Table 2 for An analysis on the direction of swimming shows that on average adult grayling need to swim against the river flow 58% of the time which is more than the juvenile trout at 53%. This is an additional information which could be used to characterize riverine habitat quality in hydropeaking events. Estimation of swimming direction can further improve qualitative and quantitative habitat assessment.

| Success probability distribution
The success probability distributions for different flow changes are shown in Figure 7a,b for juvenile trout and adult grayling respectively.
The distribution shows the probabilities at which fish succeed in finding a refuge during sudden flow change. When the flow increases, the distributions show that juvenile trout are successful to find a refuge with a lower probability, less than 0.4. As shown, all distributions are skewed to the right with the Fisher Pearson skewness coefficients 0.6, 1, and 1.2 for 8-14, 14-20, and 28-35 m 3 /s, respectively. When the flow decreases, juvenile trout are successful in finding a refuge area with a probability of more than 0.8. As shown, the success probability distributions are skewed to the left with the Fisher Pearson skewness coefficients À1.2, À1.5, and À1 for 6-2, 14-4, and 28 to 20 m 3 /s, respectively. coefficients À0.7, À0.9, and À 1.7 for 6-2, 14-4, and 28-20 m 3 /s, respectively. When the flow increases from 28 to 35 and 14 to 20 m 3 /s, the distributions are skewed to the right, meaning that their success probability is less than 0.4 more than 50% of the time. However, for the flow change from 8 to 14 m 3 /s, no skewness is observed and the success probability is more than 0.6 at 47% of the time.

| DISCUSSION
The sudden flow alteration as a result of hydropeaking changes the river environment, declining river ecosystems and fish habitat quality (Person, 2013). The probability model developed in this study quantifies the habitat suitability change caused by hydropeaking. The results include the distribution of refuge areas in the studied area and fish success probability of reaching the refuge area.

| Influence of hydropeaking on the habitat suitability index
In the study areas at a 6 km distance below the hydropower plant at the Kuusinkijoki River, the habitat suitability decreased as a result of flow increase (Figures 4 and 5). As shown in the maps, the habitat suitability indices are higher at the riverbanks. This is especially visible at high flows, for example, flows 20, 28, and 35 m 3 /s. The areas near the riverbanks typically provide better habitat during sudden flow increase because of the higher amount of debris and vegetation (Rato et al., 2021) and lower flow velocities and river depths (Greimel et al., 2018).
Hydraulics parameters and conditions such as river flow and depth impact habitat suitability and habitat selection by fish species (Flodmark, Forseth, L'Abée-Lund, & Vøllestad, 2006). Both Figures 4 and 5 suggest that the studied river section has a more suitable habitat for adult grayling at different flows in comparison to the juvenile trout, where the habitat suitability significantly decreases as the flow increases. This difference in the habitat suitability index could be explained by the preference curves. Juvenile fish prefer a shallow habitat with lower flow velocities due to reduced swimming performance than adult fish (Greimel et al., 2018;Heggenes & Traaen, 1988). The habitat modeling compared the habitat suitability at various flows, which could be used to identify and limit the critical ramping rates.
The suitability index maps show which parts of the river are more susceptible to stranding. This is important information if one is to develop F I G U R E 5 Suitability index probability distribution for the studied area at different flows (m 3 /s) for (a) juvenile trout and (b) adult grayling tools or techniques to prevent stranding. The maps and distribution of distance to refuge areas also show suitable habitats after hydropeaking. This identifies where fish may choose to go during the sudden flow change, which could be used to locate popular areas. These novel outcomes and approach could be used to develop flow mitigation and river rehabilitation strategies.
Sudden flow alterations due to hydropeaking cause serious disturbances (Moog, 1993;Schmutz et al., 2014). Our results show that the average suitability indices are higher for the adult grayling before and after hydropeaking in all the selected events ( Figure 6) which may be due to their rapid adjustment to the sudden flow change (Jungwirth & Winkler, 1984).

| Distance to refuge area
Fish respond to changes by moving towards a suitable habitat, which requires them to swim more. If they can detect a more suitable habitat, they may move towards it (Jones & Petreman, 2015). Our results show that the refuge area for juvenile trout was mainly distributed in the immediate refuge area, 1 ≤ d < 2 m, whereas the refuge for adult grayling is much more spread out. This may be due to the habitat suitability preferences associated with species (Greimel et al., 2015) to live in different parts of the river with various flows and depths.
Although our results show that up to 20% of the time (Table 2), the refuge area can be found at a distance of more than 5 m, this does not necessarily mean that juvenile trout can or would reach these areas. Flodmark et al. (2002) observed that fish initially respond to flow fluctuation, but after a period of time they do not respond as they adjust to the conditions. Studies also show small levels of movement during hydropeaking, especially in winter when fish rarely move more than 8 m at once (Puffer et al., 2014). Fish typically move large distances in critical flow conditions, for example, when habitats start to dewater (Scruton et al., 2003). Studies have reported that juvenile trout move immediately towards wetted areas during hydropeaking if they find a suitable habitat . In general, fish movement activity and pattern depend on many factors and environmental F I G U R E 6 Average of suitability index by distance (d) from the initial location before and after hydropeaking [Color figure can be viewed at wileyonlinelibrary.com] F I G U R E 7 Success probability distribution to find refuge area for (a) juvenile trout, (b) adult grayling conditions. Season and time of day or night may have a greater effect on their need to relocate than hydropeaking (Puffer et al., 2014;Robertson, Pennell, Scruton, Robertson, & Brown, 2004;Stickler et al., 2007), but typically during natural habitat change, timings are not sudden and thus do not cause additional stress on fish. Timing and magnitude are the main differences between natural habitat change and forced habitat change due to hydropeaking. Hydropeaking causes unnaturally sudden flow variation in terms of flow depth and velocity and, in the worst cases, may force fish to relocate habitats several times per day.
Our probability-based suitability modeling approach can be applied for regulation management by creating distance estimations of how far fish might need to move during the hydropeaking periods.
This information is valuable for setting environmental constraints on hydropower plants but could also be used as a tool in river channel morphological restoration actions and to further improve river habitat conditions from a habitat "accessibility" point of view.

| Probability of reaching refuge during a hydropeaking event
In our analysis, flow alterations influenced the fish's success probability of finding a refuge area. In the study section, depending on the flow's sudden movement and direction, the success probability differed. Success probability distributions suggest that both species can find refuge areas with a rather high probability of success, that is, more than 0.8 when the flow decreases. A high probability was specifically predicted for the adult grayling at the flow changes 28-20 m 3 /s, where 70% of the time, their success probability is computed more than 0.8.
On the other hand, when the flow increases suddenly, the model predicts that both species have lower probabilities (less than 0.4) to successfully find refuge. These results show that both species are more sensitive to the sudden flow increases in the studied area. This is also in line with Figure 4, where habitat suitability declines as the flow increases. However, studies show that the sudden flow decrease can be considered more critical for fish as it increases the risk of stranding (Bradford, Taylor, Allan, & Higgins, 1995;Hunter (1992), Auer et al., 2017).

| Limitation of this study
The probability model developed in this study is a novel approach to quantifying the impact of hydropeaking on fish habitats. Despite the valuable results, the model has some limitations. The distance to the refuge area is calculated using the shortest path between two locations. In reality, fish do not necessarily take the shortest path due to natural obstacles, and so on, therefore the distance to the refuge area might be longer. This could also impact success probability as some refuge areas might not be accessible. In future studies, natural obstacles, such as high-velocity areas, and the most suitable flow velocities for movement could also be considered in the analysis. We considered six hydropeaking events. For more reliable results, the model should be run for more scenarios considering seasonal impact or ecologically sensitive periods such as spawning timing.
The current model creates a bridge between river hydraulics and physical habitat conditions. However, the fish's biological capabilities have not been taken into account. In future work, we will need to consider the fish's biological abilities, such as swimming speed. In addition, the required velocity to reach a refuge area in a short period should be modeled which can then be used to quantify the levels of stress imposed on fish during hydropeaking.

| CONCLUSION
We developed a new probability model to quantify the impact of hydropeaking on fish habitat suitability. The model is based on hydraulic characteristic of habitat and considering fish preference curves. The model was tested for two species, that is, juvenile trout and adult grayling. The results show how changes in flow due to hydropeaking affect habitat condition and possibility of fish to move to more suitable habitat within a certain time period (1 hr in this study).The habitat suitability maps show that the habitat deteriorates as the flow increases in the studied section. The model is capable to account for the special habitat requirements of different fish species and their life stages. For example, it predicts that juvenile trout reacts more sensitive than adult grayling to hydropeaking. Suggested model allows to quantify different impacts of hydropeaking caused by the energy market demand using a link between the fish habitat conditions and changing with time hydrodynamics. One of the next steps in model development will be incorporation of fish swimming capacity as a factor affecting the success probability of fish finding the refuge.

ACKNOWLEDGMENTS
The author would like to thank Olvi-Foundation (Grant No. 20210718).
The study was supported by Academy of Finland funded EcoRiver (number: 323810) and Hydro-RDI (number: 337523) projects.

DATA AVAILABILITY STATEMENT
The data that support the findings of this study are available from the corresponding author upon reasonable request.