2.3. River Geometry
The geometry input for the 2D hydraulic model is based on airborne light detection and ranging (LiDAR) data, a remote sensing technique for mapping relatively shallow water bodies, which is increasingly used for topo bathymetric surveys providing a high-quality digital elevation model [
21]. In Storåne, LiDAR bathymetry reached an 8 m depth [
28]. It is a fast method for collecting high density data, covering rivers of 15–20 km in a few hours. The high level of detail results in a large amount of data, which requires extensive processing before it can be used. The data was delivered as LAZ files, a special type of compression just for LiDAR data. I was deliver by Airborn Hydro Mapping (AHM), Austria, who carried out the flights and data processing. The coordinate system is European Terrestrial Reference System 1989 (ETRS 1989) with ellipsoidal height which, in Storåne, is 44.4 m over the NN2000 reference system normally used in Norway. Data were imported into ArcGIS 10.5, (Environmental Systems Research Institute, California, USA) to create the digital elevation model (DEM) that was used in the hydraulic model.
The entire length of the study site from the outlet to Hovsfjord is 2 km and the total mapped length was 2.6 km including areas upstream and downstream of the study site not used in the model. The number of points recorded were more than 82 million with an accuracy of 6 cm in the XY plane and mean error of 3 cm in elevation [
28]. The point cloud has different density areas. The density affected mainly the water depth with fewer points in the deep areas of the river. Based on this, a DEM with a resolution of 0.5 × 0.5 m was created to ensure that in every square of the DEM there is a point. The DEM could be made finer in areas where point density is higher but we considered our DEM appropriate for the aim of the study.
2.4. Scenarios
A series of different turbine shut down scenarios have been designed to simulate and quantify the stranding risk areas downstream Hol 1 power plant. The station has four turbines, each one with a capacity of 15 m3/s. The residual flow in the river was set at 6 m3/s representing the 5-percentile flow.
Currently the typical dewatering scenario is a discharge decrease of 15 m
3/s in 5 minutes for each turbine. This configuration represents Scenarios A in the study. There are three different scenarios A: A10, the power plant decreases the production from one turbine (21 m
3/s) to zero turbines (residual flow of 6 m
3/s); scenario A21, the power plant decreases the production from two turbines (36 m
3/s) to one turbine (21 m
3/s) and scenario A32, the discharge goes from three turbines (51 m
3/s) to two turbines (36 m
3/s) (
Figure 2). The scenario where we go from four turbines to three turbines was considered for the wetted areas only and not for dewatering scenarios since this is over the 95-percentile discharge and therefore not critical for the stranding. The full production scenario of 66 m
3/s, is considered as the base line for the calculation of dry areas.
All the possible scenario combinations can be obtained by overlapping the simplified scenarios that are proposed in the previous paragraph, for instance, going from three to zero turbines in 5 minutes will be obtained by overlapping scenarios A32, A21 and A10.
We designed a different alternative for each of these scenarios, called Scenarios B (
Figure 2). They are designed so that the turbine goes from full production to 40% production in 25 min and then stops in 5 min for a full shut down time of 30 min. The need for the last fast drop is due to the manufacturer‘s restrictions, the turbines cannot operate at values under 40% of max discharge (B. Dønnum, personal comunication B10 is designed to go from one turbine to zero turbine in 30 minutes, in the first 25 minutes it decreases to 40% of turbine capacity (12 m
3/s) and in the last 5 minutes the turbine is completely stopped and we only have residual flow (6 m
3/s). B21 consists of decreasing from two turbines to one turbine in 30 minutes, in the first 25 minutes it decreases to 40% of turbine capacity (27 m
3/s) and in the last 5 minutes the turbine is completely stopped and only one keeps functioning (21 m
3/s). Scenario B32 consists of decreasing from 3 turbines to two turbines in 30 minutes, in the first 25 minutes it decreases to 40% of turbine capacity (42 m
3/s) and in the last 5 minutes the turbine is completely stopped and only two keep functioning (36 m
3/s). Changing the operation of the turbines will reduce the benefit for the HPP operator. The cost of this mitigation measure has also been estimated.
2.5. Hydraulic Model Set-Up and Calibration
The Hydrologic Engineering Center’s River Analysis System (HEC-RAS 5.0.3.) developed by the U.S. Army Corps of Engineers Institute for Water Resources Hydrologic Engineering Center, 609 Davis, California, USA [
29] was used for the hydraulic simulations. The simulations were based on the Diffusion-Wave Form of the Momentum equation [
30]. The Manning n was set to 0.06, a value which corresponds” to rivers whose bed is sand and gravel with some boulders and banks with overhanging bushes and trees [
31]. Other values were tested (0.03, 0.04 and 0.05) but results differed less than 0.01 m and therefore it was decided to use 0.06 for the entire domain. The input geometry was the DEM with a resolution of 0.5 by 0.5 m. Mesh resolution has a noticeable effect on the computational accuracy of the water-surface elevations and velocities prediction [
32]. However, if the grid becomes smaller, the time-step should also be shortened according to the Courant’s condition, with longer computational time as a result. The Courant number or Courant condition is used to ensure stability and accuracy in an unsteady model [
30]. When we define a computation time step for our unsteady model, the Courant number has to be smaller or equal 1. Several iterations were made to achieve an optimized grid and time-step that allowed us to carry out the study in a feasible and efficient time frame. Finally, the computational mesh was created with 1 m resolution in mid-stream and 0.25 m resolution along the river banks. This way we could use a higher resolution in areas of interest (river banks where stranding is an issue) and decrease number of points of the model in areas where stranding effect is no issue. The final time-step used was 0.5 s. Boundary conditions were introduced upstream and downstream of the reach. The upstream boundary condition was a flow hydrograph corresponding to the releases from the power plant, and downstream boundary condition was a normal depth with a slope of 0.01 that was measured in the area of the downstream boundary. Still this boundary was set well downstream of the actual study reach. Other values of the downstream slope were tested but the results of the model changed less than 0.01 m.
Calibration was made by using three main data sources covering a discharge from 2.44 to 44.72 m
3/s. The sources used were the water edge provided by AHM and computed from the LiDAR data at the time of measurement, a set of RTK-GPS (Real-Time Kinematic Global Positioning System) measurements (Leica Viva CS15, developed by Leica Geosystems Ag, CH-9435 Heerbrugg, Switzerland) of the water edge for four different discharges (
Figure 3) and finally aerial pictures from Norge i bilder (
www.norgeibiler.no).
The comparison with the water edge from the Lidar survey was made visually as a first approach to check the model performance. We analyzed this with an unsteady simulation with a continuous discharge of 31.72 m3/s, which was the discharge at the day of the flight.
For the GPS-measured water levels, we did simulations with continuous flow for every calibration discharge, imported the computed water surface elevation (WSE) from HEC-RAS into ArcGIS, and then compared this with the measured GPS elevation.
Finally, to obtain a broader range of discharge for comparison we selected an aerial picture from Norge i bilder (
www.norgeibilder.no) from 2006 with a recorded discharge of 2.44 m
3/s. The picture has a resolution of 0.5 m; therefore, calibration uncertainties are higher, but this resource gave us the opportunity to compare well below any discharge available during the field measurements. The comparison is made visually by contrasting the wetted areas and water edge in the picture with the simulation (see
Figure 3).
The HEC-RAS simulations were compared with a total of 253 GPS points measured along the water edge. These points were compared with a simulated water level at the same discharge as at the time of the measurements. The comparison gave a mean error of 6 cm and a standard deviation of 6 cm (
Figure 3). It is worth noting that the accuracy of the LiDAR data is 3 cm [
28] and the accuracy of the GPS is between 1.5 and 3 cm. Based on this we find the model to be well-calibrated for our purpose.
The comparison with the water edge of the LiDAR survey flight and the aerial picture were very similar to the model simulation for the corresponding discharges (31.72 and 2.44 m
3/s) shown in
Figure 3. The picture from Norge i bilder was taken with 2.44 m
3/s and represents only residual flow coming from the river upstream.
2.5.1. Variation of Wetted Areas
The maximum potential stranding areas are those which dry out when flow is passing from high to low. We simulated all discharges for the different shutdown scenarios and computed the wetted area for each. The dried-out area is then calculated from a reference discharge of 66 m3/s representing full production. These areas are presented as different maps representing dried areas as a function of the number of turbines that are running at the start and at the end of the scenario.
The middle step of scenario B21 with one turbine running at full production and another one at 40% of full capacity (see
Figure 2, Scenario B21) is referred as turbine 1.4. Correspondingly, turbine 0.4 represents the middle step of scenario B10, when there is one turbine running at 40% of full capacity.
We also calculate the marginal dried-out area per unit of flow (m
3/s). It is calculated as the dried-out area increase divided by the corresponding discharge.
2.5.2. Consideration of Damping Effect in Dewatering Ramping Rate along the Stream
The damping effect is a decrease in the amplitude of an oscillation because of energy being transformed to overcome frictional or other resistive forces. In our case when the turbine is shut down, it creates a wave that will be steep and marked right after the outlet but will become smoothed the further downstream it travels. This factor needs to be considered when calculating the dewatering ramping rate.
A simulation of 3 hours of duration was run to enable the evaluation of the damping effect. In the first 5 minutes the discharge changes from 66 m3/s (four turbines plus residual flow) to 6 m3/s (residual flow).
We generated water surface elevation (WSE) maps every 5 minutes, a total of 35 WSE maps over the duration of the shutdown process. The maps were computed in HEC-RAS and imported into ArcMap for further analysis. From each map we extract WSE from 10 representative points along the river to evaluate the damping effect.
2.5.3. Evaluation of Dewatering Ramping Rate
Stranding of juvenile fish has been a documented consequence below hydropeaking power stations [
13]. Ramping rates are important to consider to ensure habitat improvements [
33], and according to Halleraker [
7] ramping rates higher than 10 cm/h will increase fish stranding risk.
In this study we evaluate the dewatering rate according to the Envipeak guidelines, which define four impact levels according to the dewatering speed: very big, big, moderate and small [
5] (see
Table 1). We define the dewatering ramping rate as the critical velocity rate recorded during an episode with a 5-minute resolution.
The dewatering ramping rate maps for each scenario are created by subtracting the WSE at the end from the WSE at the start of the scenario divided by time. Notice that this time is not the same in every point due to the damping effect and therefore we need a raster map in which every point has a different duration depending on the distance from the outlet.