6. Attraction Flow and Migration Habitat Assessment Using an Agent-Based Model

One important key to re-establish sustainable fish populations in rivers is the fish habitat connectivity. In most rivers, the connectivity is disturbed through multiple obstructions such as hydropower plants, weirs and sills. Even if those are supplied with well-functioning fish ladders, the detectability of these facilities for the migratory fish usually plays a critical role in the overall passability of a river barrage. Attraction flow is one of the major factors defining fishway performance (Bunt et al. 2011; Cooke and Hinch 2013; Silva et al. 2018). In many cases investigations focus on the attraction flow rate, i.e., the proportion of the flow from the fishpass to the flow from the adjacent turbines or a weir (Larinier et al. 2008). The flow velocity magnitude in the vicinity of the fishpass outlet and its rate in comparison to the surrounding river flow velocity is another parameter often considered (Williams et al. 2012). Various studies indicate that other factors can influence fish movement when approaching a fishpass entrance. Turbulence (Liao 2007; Kirk et al. 2017), location of the attraction flow outlet (Burnett et al. 2016), and other physical and chemical parameters such as for example temperature (Capra et al. 2017; Caudill et al. (2013) Indirect effects of impoundment on migrating fish: temperature gradients in fish ladders slow dam passage by adult Chinook salmon and steelhead. PloS One 8(12), e85586) or light and noise (Popper and Carlson 1998) have been studied and could potentially affect migration. Other factors like water depth (Scheibe and Richmond 2002, Goodwin et al. 2006), river morphology and obstacles on the river bottom may influence fish movements as well (Piper et al. 2012 and 2015). Within the FIThydro project the habitat model CASiMiR (Noack et al. 2013) has been extended by an agent-based module. The model aims on the assessment of probability for fish being routed into the outlet of a fishway.

The model approach has been developed and evaluated using fish track data collected at the test case site Altusried hydropower plant (HPP) with an acoustic telemetry system. The HPP Altusried is one of 5 HPPs in the Upper Iller River, a tributary of River Danube in South-West Germany (Fig. 6.1, left). It is in operation since 1961, has a hydraulic head of 3 m, an installed capacity of 1,6 MW, the maximum turbine flow is 80 m³/s and it is equipped with 2 Kaplan turbines. They are located at the left side of the weir. All 5 HPPs in the Upper Iller River are operated by the Bayerische Elektrizitätswerke (BEW) and have lately been equipped with facilities for upstream migration. The fishpass outlet in Altusried is located about 260 m below the weir (Fig. 6.1, right).

The telemetry network consisted of 16 receivers 180 kHz HR2 (High Residency) with built-in synchronization tags and temperature, noise and tilt sensors, and 6 reference tags. In total, 25 grayling and 22 barbel were caught, tagged and released in the telemetry system array during their respective migration period in spring 2019.

The concept of the migration model combines habitat suitability maps for migrating fish with additional information on the swimming behaviour of the observed fish in the flow field. Hydrodynamic parameters that define migrating corridors are derived upon the statistical analysis of fish tracks of European grayling (Thymallus thymallus) and barbel (Barbus barbus) recorded over the spring period of 2019 downstream of the HPP Altusried. Swimming behaviour is expressed in terms of a histogram of the probability of fish to change the swimming direction compared to the one in the previous movement step.

The present model operates on the results of a 2D hydrodynamic model. Flow velocity vectors, velocity magnitude and flow depth, obtained with the model Hydro_AS-2D (Nujic 2006) on an unstructured mesh, are interpolated to a structured grid of the migration model. Starting at a defined specific location, a virtual fish, the so-called fish-agent, evaluates the current surrounding flow field and selects the most probable next movement direction according to the pre-defined behavioural rules. Step by step, the path of a fish-agent is calculated and visualized, allowing the modeler to get a picture of possible migratory movements in the far vicinity of a fish ladder outlet. The following main elements form the basis of the migration algorithm:

Two fuzzy rule-based model versions were tested for the calculation of the migration suitability: One with two parameters (flow velocity and water depth) and the other with four parameters (flow velocity, water depth and spatial gradients of flow velocity and water depth). The comparison of recorded fish tracks and calculated hydraulic parameters shows that in the final migration phase approaching the fishway, many fish individuals move along the areas where hydraulic parameters change abruptly. Thus, it can be expected that spatial gradients play an important role for fish migration. However, for brevity reasons, only the two-parameter system based on flow velocity and water depth is presented here.

Both fuzzy rule systems are derived through the analysis of the observed fish tracks in the time-period of 30 min prior to the first entry into the fishpass. Figure 6.2 (right) shows the frequency distribution (blue bars) of the four hydraulic parameters for all observed grayling during the above-mentioned time-period. Based on these frequency distributions of spatial use, up to five fuzzy sets (categories that indicate the preference of fish to use certain hydraulic conditions) are derived (see Fig. 6.2). Combining those fuzzy sets with fuzzy rules (Fig. 6.2, left), the migration corridor suitability maps are calculated.

Some simulation results are presented in the following figures. They are overlaid with the ortho images of the HPP Altusried. The comparison of observed tracks with modelled tracks (example for grayling in Fig. 6.3) shows that the concept of flow probability histogram combined with the random method for the final selection of movement direction allows to mimic up to a certain degree the searching behaviour of observed fish (random lateral, zig-zag). However, individual fish paths of all tagged fish vary widely and many movements are most probably not directly related to migration behavior.

The benefit of the simulation model is that it allows to demarcate migration corridors indicating areas most probable to be used for upstream migration and fish migration paths for different flow scenarios. This makes it a suitable tool to develop mitigation options for the detectability of fishway entrances.

Options to mitigate the attraction flow can be for example a seasonal adaptation of the turbine flow from a considered HPP or release of additional water into the river in direct neighbourhood of the fishpass outlet. Simulation scenarios in Fig. 6.4 show that migration corridors for barbel are getting narrower with increasing discharge from the HPP but at the same time are shifting from the middle of the river towards the banks, which could increase the chance for fish to find the fishpass outlet. In contrast, increasing the discharge too much, up to 80 m³/s and higher, leads to an interruption of the migration corridor on the left river side, which could impair the attraction to the fishpass outlet (Fig. 6.4, bottom right).

First results of an agent-based model are promising. The current model mimics the movement of individual fish by a combination of migration habitat suitability maps with behavioural rules of fish derived from observations of fish movements in the flow field. The base parameters for the definition of migration corridors are flow velocity, water depth and hydraulic gradients. Predicted swimming paths of grayling and barbel show high similarities with observed tracks of individual fish.

The evaluation of the model runs for multiple fish-agents for the test site Altusried confirms a basically appropriate position of the fishpass entrance. HPP flow rates in the range from 40 to 50 m³/s for barbel and from 40 to 80 m³/s for grayling seem to be favourable in the migration season. Fish moving upstream along the left riverbank find the entrance with a much higher probability than those moving close to the right bank.

Further developments will concentrate on more detailed processing of fish tracks aiming to distinguish between different behaviour types (feeding, resting, searching). Additional investigations are planned for the final identification of key hydrodynamic and environmental parameters for the migration model.