Duncan Philpott describes his MSc research project which he aligned to WTT habitat work carried out by Research and Conservation Officer, Professor Jon Grey. Duncan is now living in Sweden and working on research connected with trout health.
In the Yorkshire dales, under the watchful eye of Wild Trout Trust’s Jonny Grey, a small tributary of the River Wharfe called White Beck was undergoing restoration. The goal of the restoration was to improve vital spawning habitat for smaller to medium sized trout. The site provided a case study for the subject of my MSc Geographic Information Systems thesis, assessing the suitability and testing the limits of the spatial resolution of drone surveying combined with structure from motion methods.
By flying a drone along a specifically designed flight path capturing images at regular intervals and multiple angles, supplemented by reference control points obtained by survey equipment it follows that high-resolution images and a 3 dimensional reconstruction of the scene can be obtained through the use of structure from motion software. The resulting images and models would be true to space in the real world and permit the measurement of features or other commonly used remote sensing analysis methods such as land cover classifications. The drone used for this research was a readily available consumer product that provides access to datasets that previously would have required the costly hire of aerial image acquisition and laser point surveying.
Three flights (Fig 1 above) were made at the site, the flights took place over half a year and were made before and after the restoration work with the exception of the installation of riparian fencing which occurred shortly before the first flight.
White Beck’s initial visits from Jonny detailed a limestone stream with great clarity, abundant macrophytes and a healthy macroinvertebrate population yet it was a stream that lacked the required habitat to provide its full potential for spawning trout, travelling upstream from the River Wharfe, and the subsequent survival of newly hatched fish. The observed algal growth on the substrate hinted at excessive nutrients and the substrate size lacked sorting for suitable spawning use due to the historical straightening of the river channel.
The approach for restoration was two-fold (Figure 2 below), firstly, riparian fencing and replanting to exclude livestock and secondly, the improvement of instream habitat with the installation of multiple large woody debris (LWD) structures.
The installation of LWD is often used to counter the negative effects of channel straightening by increasing bed heterogeneity through morphological impact consisting of scouring and deposition. The morphological impact can vary depending on the location and orientation of the LWD relative to the river’s flow. Research shows that LWD installations may have the greatest impact in rivers that lack riparian shading such as these upland streams in contrast to rivers with extensive shading (Donadi, Sandin, Tamario, & Degerman, 2019). LWD improves spawning opportunities since the scours that result from their installation clean and sort substrate and increase the amount of suitable spawning area. Furthermore, LWD can serve as a location for brashy debris to tether and provide juvenile habitat with protection from predators.
Riparian fencing controls the access of livestock to sensitive river banks. Livestock movement is a source of erosion which exposes soil and leads to sedimentation which can have adverse effects on invertebrate abundance and clog spawning gravels. Extreme livestock poaching may also result in over-wide and shallow channels, reducing the temperature tolerance of the river and reducing the area with suitable holding habitat for fish. With relief from grazing, shrubby vegetation will have the opportunity to grow and can provide shading and refuge along the river’s banks. Shrubby vegetation can also provide a valuable buffer to reduce the quantity of fine sediments and excess nutrients that are contained in surface runoff from the surrounding landscape.
Since the surveying did not span a full year and take place in comparable seasons, it is difficult to comment on the absolute increase in vegetation, however comparisons between fenced and unfenced regions are possible.
From the high resolution orthoimages we can see that there is a distinct difference between the two regions (Figure 4 and 5). We can observe increased variation in vegetation species and an increase in vegetation height in contrast to the unfenced regions. During winter time the exposed soil from the movement of livestock is obvious, especially near the bridges, while vegetation inside the fencing remains dense and bushy, hopefully mitigating the transport of some of the exposed soil into the river during heavy rainfall events.
The digital elevation models (DEMs) are another product of the drone survey that shows the height of every pixel in the scene. The DEMs explain vegetation growth by the comparison of the difference of height values between multiple surveys (commonly called DEMs of Difference (DoD)). To complement the interpretation of elevation differences in an image format, cross sectional areas can also be computed from the DEM (figure 6).
In Figure 6 (above), the DoD is displaying the elevation difference between the winter and summer surveys, green indicates a higher point in winter survey whereas red indicates a point that was lower in the winter survey
From the cross section we can compare all survey dates simultaneously, there is very little difference observed in the grassy regions outside of the fencing however the vegetation inside the fencing fluctuates greatly through the seasons. There are differences at the edge of the river channel (13m) where vegetation has begun to overhang the stream.
The use of the drone allows the monitoring of all LWD installations simultaneously, provided they are within the designed flight path
Figure 8, above. LWD installation as observed by the drone a) pre-installation, b) post-installation in August c) in winter. The right hand image shows the perspective of the feature as observed from the river bank.
A birds-eye view permits an undistorted perspective of any scours and change in substrate size at each installation. Since the images produced from the structure from motion method are “georeferenced”, lengths and areas within the image, can be measured to centimetre level accuracy with the use of a geographical information system. For example the size of scour or size of substrate may be measured for each installation and easily totalled across the work completed.
The LWD installations created scours after the first high flow event. Generally, the quantity of sorted and cleaned gravels (Figure 8b), did not increase or decrease in area/size after further winter high flows. Indicating that substrate cleaning and sorting benefits occurred quickly for this size of LWD and river.
When comparing every LWD installation, the performance of the LWD varied. Overall, the V‑shaped baffled produced the best results whereas single pieces of LWD placed at an angle to the direction of flow generally scoured and sorted less substrate. The V‑shaped LWD installed in wider and shallower stream sections showed smaller scours in comparison to the V shaped baffles that were installed in the constricted and straightened channel. LWD installed in the narrowest and straightest section of White Beck showed the most consistent cleaning of algal growth on the substrate and the sorting of sediment sizes (Figure 9).
Figure 9 above. Cleaned substrates in the channelized section of White Beck. Note the apparent narrower width of the river given the quantity of overhanging vegetation due to a protected riparian zone.
Unfortunately, the scours visible in the orthoimages were not within the limit of detection of the digital elevation models of the scene. The DEMs proved to be of most use for monitoring the volume of vegetation at a site.
In this study, the subsurface detail was visible due to the great water clarity, general shallow depths and a lack of large trees in the riparian habitat. The overhead perspective of the drone can often provide greater detail of the river bed than if viewed from a lower angle on the river bank other studies have seen success with rivers of depths up to 1.4m. However, densely forested rivers prove to be unsuitable for drone related methods however for upland, urban or alpine streams with limited tree cover
Once a flight plan is designed and control points are acquired for a site, repeat surveys require minimal effort. Revisit surveys over a greater time span could aim to quantify the effect of LWD, relative to river dimensions and flows, while detailed substrate sizes and distribution can be computed. In future this could perhaps allow LWD can be designed to provide spawning habitat that is more specific to the fish that exist in the river. Orthoimages from drones may also be effective for surveying redds and enabling the year to year comparison of the fish spawning in the river.
From a series of overlapping images captured on an automated and pre-planned flight path, that observes a feature from multiple angles, structure from motion (SfM) methods may be applied to generate a 3‑dimensional data point cloud of the scene and subsequently, a digital elevation model.
Points within the captured photos are matched using a software algorithm. The position of these points relative to the camera position when the image enables trigonometry to determine the position of features in real space allows a geometric reconstruction of the scene (Figure Sweeney)
Above: Multiple camera positions observing feature points for 3d reconstruction, Souce: Sweeney, 2016 (http://www.theia-sfm.org/sfm.html)
