Catchment-scale geomorphological modelling of leaky dams using CAESAR-Lisflood
Josh Wolstenholme, Chris Skinner, David Milan and Dan Parsons
https://meetingorganizer.copernicus.org/EGU22/EGU22-5730.html
Abstract
The introduction of large wood to fluvial systems is becoming increasingly popular as a method of natural flood management commonly referred to as leaky dams. These are often installed as semi-permanent features through live felling and anchoring in-situ. Currently, most natural flood management modelling is hydrological and focuses on flood risk without accounting for geomorphology of these ‘fixed’ features. We argue that the long-term effectiveness of NFM interventions require and understanding of the nested hydrogeomorphological processes at work within river catchments, particularly those related to bed scour, sediment transport and deposition, and the associated feedbacks following implementation of leaky dams. Leaky dams that are designed to attenuate the hydrograph and ‘slow-the-flow’, may cause sediment storage as well as scour, potentially impeding the effectiveness of a leaky dam to reduce flood risk after a single storm event. Using the new ‘Working with Natural Processes’ toolbox developed for CAESAR-Lisflood, the influence of different storm scenarios on a series of leaky dams in a hypothetical catchment based on a site in North Yorkshire is assessed. The effectiveness of the model at representing the influence of the dams on hydrogeomorphology is also assessed.
What are leaky wooden dams?
These are a form of Natural Flood Management (NFM). They aim to reduce flood risk by temporarily storing water in the river system, and encourage overbank flow to reconnect rivers to the floodplain. This water is then slowly released over time, back into the river network. They aim to emulate natural large woody debris, however are semi-engineered structures that are often fixed in place, installed in the upper catchment.
There are many different types of leaky dam, and no one design is used across the UK. Assessing how these structures influence geomorphology and in turn, flood risk, is key to understanding how their efficacy may change over time.
Monitoring
Two small reaches in Dalby Forest, North Yorkshire are monitored to assess the impact on dam installation on bank and channel geomorphology, as well as sediment grain distribution. The sites have been monitored since July 2019, with dams installed in February 2020.
Dalby Forest - Site 1
River stage at this site has only caused the dam to activate once. There have been no large enough events to activate the dam on a regular basis. Flow direction is from the base of the image to the top.
Dalby Forest - Site 2
This dam has been active since approximately April 2020. It actively recruits natural wood and fine grained sediment. The flow direction is from the bottom of the image to the top.
Dams influence channel geomorphology...
Through repeat bathymetric surveys with a Total Station, we can assess geomorphic variance at each of the sites.
Site 1 shows that following installation of the dam there is increased variation away from the median elevation height suggesting increased variation, however there are few individual peaks which could relate to areas of significant geomorphic change.
Site 2 however shows a greater increase in variation, with less concentration around the median, and spikes which indicate potential areas of scour and deposition.
Probability density function of normalised channel bathymetry relative to the elevation median.
...and sediment distribution
Dams are also shown to influence the sediment distribution of the sites. Site 1 (below: 1A, 1B) shows a normally distributed cumulative sediment curve throughout the monitoring period, with small fluctuations. In comparison, site 2 (below: 2A, 2B) has a significant change throughout time. The upstream portion of the river (2A) shows an increased percentage of fine grained sediments, whilst downstream of the dams (2B), there is a decrease in fines and increased coarse grained material.
Modelling leaky dams
To understand the evolution of these structures, and how they will impact flood risk management, it's important to perform numerical modelling and consider the geomorphic processes. As they are in-situ features, we do not use a wood transport model.
Current catchment-scale modelling of leaky dams is few and far between, and these models typically focus on single events and do not consider geomorphological evolution and internal feedbacks.
Leaky dams in CAESAR-Lisflood
CAESAR-Lisflood is a 2D landscape evolution model capable of modelling timesteps from minutes to centuries and more, as well as small reaches to large catchments. To represent leaky dams, we assign a cell on a regular raster grid, and we can define three parameters:
Dam gap
Dam height
Dam roughness (Manning's n)
The dam is represented by an assigned roughness value which attenuates flow after the river level reaches the base of the dam. This allows base flows to be modelled with the dams in-situ without influencing flow dynamics, similar to reality.
Here, results from varying dam roughness and dam gap size are presented. Catchment-based work is ongoing, however the method applied is the same (but more computationally expensive).
The digital elevation model (DEM) used in the model is UAV derived and resampled to a resolution of five-metres.
The maps to the left show the reach that has been modelled, and its relative location within the Dalby Forest/Bridestones sub-catchment. Blue circles represent real dam locations for the reach (12 total), and automatically generated dams for the catchment (not discussed here).
We expose the reach to six storms with annual exceedance probabilities of 100%, 10% and 1%, and durations of six and 72 hours.
Results
Water Storage
By taking the cumulative difference between simulations without dams and simulations within dams, we calculate the total volume of water stored within the catchment over the duration of each storm.
Six-hour storm
100% AEP
For this storm scenario, the only dams that are activated are those with a 0 m and 0.1 m gap size. The simulations show that a maximum of 6.5 m³ of water are stored in the reach (when modelling flow only), and two m³ when modelling with sediment transport processes.
Increasing the gap from 0 m to 0.1 m increases the total volume of water stored by up to 2.25 m³ across all scenarios. This different is less pronounced in the sediment transport simulations, and water stored throughout the duration of the model run leaves the system much quicker.
10% AEP
The 10% AEP storm scenario shows that all dam gap heights are activated with the exception of 0.4 m. Smaller gap sizes (excluding 0 m) store the most water (up to 10 m³ for flow only, six m³ when modelling sediment transport), with flow only modelling retaining the water in the reach for longer then sediment modelling.
The influence on changing dam roughness is less pronounced, with greater roughness values storing slightly more water than the previous (<0.25 m³ increase) across all parameter combinations.
1% AEP
For the largest six-hour storm, all dam gaps become activated, with 0.1 m having the greatest amount of reach storage (11.75 m³ for flow, 8.25 m³ for sediment models). The largest gap size is active for the least amount of time in the flow only models, yet in the sediment models is active for a similar duration to other gaps tested.
72-hour storm
100% AEP
Only the dams with a gap size of 0.2 m and less are activated during this scenario. The largest volume of water stored is 6.25 m³ in the reach when Manning's n is 1.2 and the gap size is 0.1 for the flow only model. For the sediment transport models, the maximum storage is 3 m³.
Increasing Manning's n appears to decrease the effectiveness of the dams for the flow only models, yet increases it for the sediment transport models.
Finally, although less water is stored in the system when modelling sediment transport, it is stored throughout the entire duration of the storm for all gaps <0.2 m.
10% AEP
Similarly to the six-hour storm scenarios, all of the different dam parameters are activated except for where the gap is 0.4 m for this storm. The smaller gaps hold more water in the reach (max 7.5 m³ and 4 m³ for flow only and sediment transport models respectively), however for this scenario, a gap of 0.2 m retains the largest volume of water.
Increasing Manning's n has a more pronounced effect on water storage than previous storm scenarios, increasing water storage for the 0.2 m gap by up to 1.5 m³.
1% AEP
All dam gap sizes are activated for the 1% AEP storm, with a maximum of 10.25 m³ and 4.25 m³ of water stored in the reach for flow only and sediment models. Similarly to the 72-hour 10% AEP storm, a dam gap of 0.2 m retains the largest volume of water, whilst 0.4 m retains the least (three m³) and is activated for the shortest period of time.
Increasing Manning's n for the dams increases the total volume stored by up to 1.25 m³.
Similarly to previous results, when modelling sediment transport processes, the dams retain water for longer, albeit a lesser volume to the flow only simulations.
Modelling with vs without sediment
From the water storage results above, there is a clear difference between the model outputs when modelling with an without sediment. To summarise the relationship, below is an interactive scatter plot (with histograms) showing the total amount of water stored in the reach for the same parameter set when modelling sediment or flow only. A static version of this from the presentation can be found below.
Click here to see a static scatter histogram
Summary
Modelling sediment transport reduces apparent effectiveness of leaky dam structures
Geomorphic change must be accounted for
CAESAR-Lisflood could be used to optimize dam parameters and positioning
Potential for scaling the method to larger catchments