Acoustic measures of biodiversity and human disturbance – a study in the UK Yorkshire Dales National Park Greg Watts, University of Bradford 1 ; Rob Pheasant 2 and David Winterbottom 3

ABSTRACT The Ingleborough Soundscape Project's primary aim was to propose a simple, low-cost acoustical metric; by which non-experts can measure the success of contrasting rewilding interventions. It also sought to quantify how exposure to transportation noise within the Ingleborough National Nature Reserve in the UK impacts human well-being. The two sites selected for the study, Colt Park Wood, and South House Moor, are located at the northern end of the Ribble Valley and are each affected by road, rail, and aircraft noise, to varying degrees. The metric developed for gauging the success of various rewilding strategies was a revised version of Acoustically Enhanced Ecological Richness, as proposed by Agius in 2020. This combines a value of ground-truthed Ecological Richness with a systematically derived value of Acoustic Richness; to give a wildness rating of 0-10+. The impact of transportation noise on humans was measured by predicting noise exposure levels and converting them to tranquillity values using the Tranquillity Rating Prediction Tool. It has been found that these tranquillity ratings relate well to levels of perceived relaxation and anxiety. Tranquillity contours were then plotted across the study site using SoundPLAN software. 1. INTRODUCTION

Informed by Krause et al. (2011), the research aims of this project became an exploration of the ecological and acoustic richness of two contrasting study sites within the Ingleborough National Na- ture Reserve (INNR). These were Colt Park Wood (CPW), a non-managed ancient woodland one kilometre south of the Ribblehead Viaduct, and South House Moor, which is 2 kilometres northeast of Ingleborough. In the latter case, livestock grazing was removed in 1999, and a programme of tree planting has since taken place. Figure 1 delineates the two study areas and Figure 3b shows the cor- responding 1:50,000 Ordnance Survey map. Ecological Richness ( ER ) was simply a count of mam- malian and avian species, measured by visual observations, i.e., direct sightings and camera trap re- cordings. Acoustic Richness ( AR ) also focussed on avian and mammalian species and was contrasted with the ground-truthing visual evidence.

Figure 1: Study areas

1 g.r.watts@bradford.ac.uk , 2 r.j.pheasant@outlook.com , 3 david@drwtech.uk

By sampling the bio-acoustic data over different seasons, it was possible to identify levels of AR that correlated with ER ; and differentiate between the two contrasting study sites. Further issues explored during this project were the impacts of anthropogenic noise on CPW and SHM. Road, rail, and aircraft noise is perceptible throughout the north-western section of the York- shire Dales National Park, and predictions of these noise levels were made using SoundPLAN soft- ware. The effect of noise disturbance on the well-being and relaxation of visitors and residents within the INNR was modelled using a combination of the Tranquillity Rating Prediction Tool (TRAPT) proposed by Pheasant et al. (2010) and SoundPLAN to present a series of tranquillity contours.

2. METHODOLOGY

2.1. Bird Surveys From October 2019 to May 2021, two ornithological specialists carried out systematic fortnightly bird surveys of CPW. At the same time, SM4 Birdsong Meters were placed within CPW and on SHM to record diurnal and nocturnal avian and mammalian activity. It was impossible to walk the whole perimeter of CPW as there was very little access on the eastern side. However, despite the wood being narrow, its entire length can easily be visually accessed; by walking from the start of the wood at its northern tip before picking up its western boundary wall. At the southernmost end, the ornithologists took a reverse route that offset them 200m from the boundary wall, allowing a different perspective to be gained whilst at the same time including the trees in the adjacent westerly meadow. On average, each survey took 3.25 hours to complete. Unfortunately, the COVID19 pandemic interrupted the surveys for eight weeks during the Spring migratory window for birds; so, the decision was taken to compensate for the missing data by con- ducting a series of weekly surveys for the same period in 2021, which took the total to 47 completed surveys. The surveyors carried out no fortnightly bird surveys on SHM during this period. However, bird transect data provided by surveyors from the British Trust for Ornithology (BTO) for 2009, and 2014 to 2018, was used as a proxy when testing out the revised AEER metric. The BTO transect across SHM; passed within 100m of where the SM4 Birdsong Meter was located from 2019 to 2021.

2.2. Audio Recordings During October, November and December 2019, audio recordings were made from both study sites. This initial phase aimed to familiarise the researchers with the operation of the equipment and identify a single point in each study site to record from. Each song meter used a sample rate of 24,000 samples per second, as this covered a frequency range of 10Hz – 12kHz, which incorporates all UK birdsong. It also produced 16-bit WAV files from both microphones. Each device was positioned with the left- hand microphone (Channel 0) pointing in a northerly direction and the right-hand microphone (Chan- nel 1), aimed at a southerly one, providing crude directional information. The SM4 Birdsong Meters were secured to a tree by a bungee strap. These were very effective; despite regularly being nibbled by rodents. The advantage of this equipment is that it provides the capacity to monitor bird sounds around the clock. Much bird activity occurs during the night, particularly concerning movement and migration, and audio recording offers a tool for capturing this critical information. However, it doesn't always provide reliable information about what use a bird is making of the study site. Furthermore, the ana- lyst requires a reasonable understanding of migratory patterns and must be continually aware of mim- icry within such complex acoustic environments.

worm 2022

2.3 Mammal Survey The aim of this part of the project was to simply note the presence of mammals on the two study sites during the day and at night using Trail Cameras with infra-red flash for night-time recording. Date and time when trigger occurred were imprinted on the photos and videos. The use of trail cameras attached to a box for monitoring small mammals was still in its early devel- opment when observations started. The deployment in the limestone pavement of CPW was particu- larly novel. Boxes were placed on the surface, on the floor of, and partway into, the grikes. On SHM the cameras were placed where potential vole runs were noticed, within the clumps of trees and along the walls. Observation was stopped once sufficient data had been collected without disturbing the site. To complete the survey of mammals, a survey of bats was also carried using two Echo Meter Touch 2 bat detectors. These were attached to mobile phones for ease of measurement. A wide range of wildlife was detected including mice, rats, voles, badgers, foxes and deer.

. 2.4 Noise and Tranquillity Predictions The Yorkshire Dales are an important natural resource that needs protecting from invasive anthro- pogenic noise, which is likely to increase over time. Just how much tranquillity is remaining can be gauged by plotting Tranquillity Rating ( TR ) contours. Questionnaire surveys of open green spaces have shown a strong association between rated tranquillity of a place and percentage of visitors feel- ing more relaxed after their visit (Watts et al., 2013). As an indicator of soundscape quality perceived tranquillity has a part to play in eco-tourism. Indeed, perceived tranquillity could then be a manage- able resource, which, is substantially beneficial to human health and wellbeing (Marafa et al, 2018). The questionnaire results have been used to validate the following category limits: <5 unacceptable , 5.0 – 5.9 just acceptable , 6.0 – 6.9 fairly-good , 7.0 – 7.9 good and ≥ 8.0 excellent These categories were then used to generate TR contours to illustrate the level of tranquillity across the study site. Laboratory studies conducted at the University of Bradford have shown that the significant factors affecting the rated tranquillity of a place, are the average level of anthropogenic noise and the per- centage of natural and contextual features in the landscape (Pheasant et al., 2010). The equation TRAPT (Tranquillity Rating Prediction Tool) expresses this relationship as: TR = 9.68 + 0.041 NCF - 0.146 L day + MF (1) Where TR is the tranquillity rating on a scale of 0 to 10. NCF is the percentage of natural and contex- tual features, and L day is the equivalent constant A-weighted level during daytime (e.g. from 7am to 7pm). Contextual features include listed buildings, religious and historic buildings, landmarks, mon- uments, and elements of the landscape, such as traditional farm buildings, that directly contribute to the visual context of the natural environment. MF is a moderating factor that was added to the equa- tion following a study that was designed to take account of the presence of litter and graffiti that would depress the rating, or natural water sounds that would improve it. For the purposes of this project a simplifying assumption was to set NCF =100 as there were few man- made features in the landscape. In addition, MF was set to zero as there was no litter or graffiti and no obvious sounds of running water. In this case it can be shown that equation (1) reduces to: TR = 13.78 - 0.146 L day (2)

worm 2022

Equation (2) was then used to convert the calculated transportation noise levels to TR ratings. For this purpose, SoundPLAN was used to make predictions of road, rail and aircraft noise using available data bases of vehicle movements. This model used the Danish Nord2000 standard for trains, cars, and motorbikes. L Aeq noise contours were plotted for individual noise sources and combined levels. The combined levels were then used to predict the tranquillity rating across the study areas. . 3. ACOUSTIC ANALYSIS Approximately three thousand hours of audio data were recorded from the two study sites and down- loaded to a MacBook Air; before being manually processed using Audacity open-source audio soft- ware to produce spectrograms. Manually identifying individual mammalian, avian, meteorological, and anthropogenic components within the spectrogram; utilized the following six stage process: 1. Play-back of audio to identify obvious anthropogenic noise sources, such as trains and aircraft 2. Using audio to identify biological components familiar to the analyst 3. Zooming in to a 7.5-minute window to pick out individual species 4. Consulting the RSPB Guide to Birdsong, to confirm classification 5. If unsure the bird sound of interest was saved as a separate WAV file and uploaded to Cornell

worm 2022

University’s BirdNET identification tool, https://birdnet.cornell.edu/ 6. If still uncertain consult Xeon-Canton birdsong forum, https://www.xeno-canto.org/

Note: Prior to submitting data to BirdNET in Step 5, the individual WAV file was often opened in Kaleidoscope Pro Analysis software, and frequency bandpass filters applied to separate weaker sig- natures from other dominant sound sources. Figure 2a and 2b show selected portions of the spectrograph used to identified bird species in SHM. The time base on the x axis of both spectrograms is in seconds.

Figure 2: (a) Short-Eared owl (b) Ring Ouzel Objective acoustic data recovered from the SM4 recorders were provided in the form of L Amax (max- imum A-weighted sound level), L Amin (minimum A-weighted sound level), and L Amean (arithmetic mean of the samples being measured), by Kaleidoscope's Sound Pressure Level analysis function. Measured L Amax and L Amean levels were used in checking the SoundPLAN modelling of train, motor- cycle and aircraft noise, statistical analysis, and the Tranquillity Rating Prediction Tool (TRAPT ).

4. RESULTS 4.1. Comparison of acoustically identified bird species When comparing the species lists of birds acoustically identified in CPW (62) and on SHM (40), there is a clear difference between the two study sites. However, counts for February and March are missing from CPW, whilst January, February, March, and May are not in the timeline for SHM. When making comparisons of ecological richness, it is important to bear in mind these differences in sam- pling periods. It is considered that this difference may not be significant since only two months, Jan- uary, and May, are additionally missing from the SHM record, with the implication that the ranking of these sites is unlikely to change if a complete data set was available. However, there are 23 hourly periods; that could be matched for identical times from July through to October. These have been used to statistically establish the probability that any observed difference between CPW and SHM in the mean values of hourly species counts resulted from chance.

The average hourly number of species observed in CPW is over twice that recorded in SHM (8.0 compared with 3.5). Using the Student’s t-test the difference was highly significant ( p <0.001). Clearly the ancient woodland at CPW supports a more diverse bird population than the more recently planted woodland at SHM. The birds identified covered a very wide range including Barn Owl, Buz- zard, Cuckoo, Magpie, Pheasant, Starling and Wren. In addition, the percentage of time in each hourly period that mechanical noise was identified was compared using the same statistics. In this case the differences were more modest; there being on average 23% more time when mechanical noise was recorded at CPW than at SHM ( p<0.05 ). Such results were not unexpected as CPW is within 400m of the nearest road and railway line, whereas SHM is three times that distance.

4.2. Acoustically Enhanced Ecological Richness ( AEER ) As mentioned previously, the length of each sample period will affect the total number of species that is counted. There are some differences in the sampling periods of the two study sites. Given the number of additional species recorded on CPW (22) there is an expectation that these differences are unlikely to corrupt the ranking of ecological, or acoustic richness, when calculating AEER . A further point to consider is that when the hourly avian species counts are paired exactly, there is a large difference of (8.0 compared to 3.5), which is statistically highly significant. AEER model as proposed by Agius (2020) is given by: (ER+AR)-DU

Where all values are based on the totals recorded over the entire study period. ER (Environmental Richness) is the number of mammals and birds identified visually, and AR (Acoustic Richness) is the number of mammals and birds identified acoustically. DU (Duplication) is the removal of dupli- cate species from the sum of ER and AR . For example, if:

ER = 70, AR = 65, DU = 45 then AEER = (70+65)-45 = 90

Within this project the equation has been revised to ( ER+AR-DU )/10, to bring the resultant score (i.e. 9.0) within one of five suggested bands, that cover a scale of 0 to10+. on a scale of very poor to

worm 2022

outstanding . In this example, the revised model reduces the AEER down from 90 to 9.0, which equates to a score of very good .

Table 1: Grading criteria for the revised AEET model

Score Rated Criteria 0.0 – 2.5 Very poor Intensively managed land, low in both ecological, and acoustic richness 2.6 – 5.0 Poor Low-intensity or extensive land management with lower levels of an- thropogenic inputs often associated with “traditional” land management practices (e.g. pastoral systems, haymaking, coppice) or promoted by agri-environment schemes 5.1 – 7.5 Good Nature conservation management through active human intervention in ecosystem dynamics specifically aimed at preserving or restoring spe- cific species, habitats, or ecosystem services. Examples include invasive species management, conservation grazing, and mowing management 7.6 – 10.0

Very good Land removed from management that is undergoing a trajectory of change to become naturally vegetated and populated with incoming spe- cies, so that new self-perpetuating communities are forming as the food web expands. If not isolated, still a need to exclude grazing livestock. 10+ Excellent Unmanaged ancient woodland predominantly of native broadleaf or na- tive conifer composition; inaccessible upland rock ledges with chasmo- phytic vegetation; vegetated sea cliffs and undercliffs; unmanaged sand dunes and bars, estuaries, and salt marshes. If not isolated, still a need to exclude grazing livestock. Based on data gathered by the Ingleborough Soundscape Project, the AEER values are calculated below. Note that some reductions are made for overflying birds.

CPW: Visual birds = 62, visual mammals 16 and hence ER = 78. Acoustic birds AR = 56 and DU = 48. Hence AEER = (78 + 56 – 48)/10 = 8.6 ( very good ) SHM : Visual birds = 15, visual mammals = 6 and hence ER = 21. Acoustic birds AR = 39 and DU = 10. Hence AEER = (21+39-10)/10 = 5.0 ( poor ) The values sit well with the timeline of each site; and the rewilding aspirations set against them.

Note: AEER for SHM has been calculated using bird species data from the 2016, 2017 and 2018 BTO transect walks. Reptiles and amphibians were not included in this example. If they were, AEER for SHM would have been 5.2, which is in the ‘good’ rating bracket.

4.3. Noise and tranquillity mapping Based on computed L Aeq from all sources over daytime period 7am to 7pm and equation (2) above, the Tranquillity Ratings were computed. Figure 3(a) shows the Tranquillity Rating contours by

worm 2022

class for CPW (position A) and SHM (position B). Figure 3(b) is the corresponding 1:50,000 Ord- nance Survey map. Reference points C, D and E, are where the main road crosses the railway line, the hamlet of Selside, and the Ribble Way, respectively.

E

worm 2022

C

A

C

A

B

B

D

D

Figure 3: (a) Tranquillity rating contours (b) OS comparison map

It can be seen from the TR contours that SHM is in a ''good'' tranquillity area, but that CPW is on the boundary of ''fairly good'' and ''just acceptable'' one. The higher TR of SHM is due to its distance from road and railway noise, which is more than treble that of CPW. It can be seen that the tranquil- lity rating within approximately 100m of the road or railway is unacceptable, but that the pas- tureland adjacent to the eastern boundary of CPW is shown as ''just acceptable''. For planning a tranquil walk in this area, it would be helpful to superimpose these figures to quickly identify tran- quil sections of footpaths, bridleways, by-ways, and lanes. As an example, it can be seen in Figure 3b that much of the 'Ribble Way' footpath in the north-eastern quadrant of the map (point E) is close to the boundary of the 'good' and 'fairly good' areas of tranquillity. More generally, such maps may interest those who value tranquil countryside and seek solace in open green spaces. It allows the user to plan walks and rides that pass through the most tranquil areas and reduce time spent in the least tranquil environments. This approach has been used to good effect in urban areas where nu- merous Tranquillity Trails have been identified in towns and cities (Watts, 2018).

5. CONCLUSIONS

As expected, there were clear differences in the biodiversity of CPW and SHM as shown by the significant difference in the number of bird species recorded and overall (including mammals) by the

revised AEER metric. The latter ranged from 8.6 ( very good ) to 5.0 (borderline poor/good ). This project started from the hypothesis that the lush ancient woodland of Colt Park and the overgrazed and tired landscape of SHM were sufficiently different habitats, that there would be minimal overlap in terms of ecological or acoustic richness. However, despite its borderline poor/good rating there are a surprising number of bird and mammal species observed in SHM and in time, as the rewilding develops, it is expected to move well into the category of “good”. The rewilding intervention of SHM, particularly the planting of the stands of alder and willow, it was considered attracted large numbers of woodland birds from CPW in the late summer of 2020. These are the key ingredients of landscape- scale conservation, which given CPW’s capacity to provide seed-stock beyond its boundary, bodes well for the Wild Ingleborough initiative. The use of the revised AEER metric should be expanded in further studies to include a validation of the categories outlined above. Closer matching of sampling periods is also required. The likely influence of the noise generated by the transport corridors and the moderating effect of expansive views of open countryside is graphically detailed in the tranquillity contour map. Such maps could be used in future to guide walkers who seek solace in these beautiful moorland areas. In addition, traffic noise mitigation may be considered for particularly sensitive locations. Despite the fact that the tranquillity category for CPW was borderline fairly good and just acceptable the effect on biodiversity seems limited since, as we have observed, the revised AEER was considered very good . Clearly further work is required to determine the levels of tranquillity that would have a detri- mental effect on wildlife. 6. ACKNOWLEDGEMENTS

We are most grateful for the helpful collaboration of many individuals. In particular we acknowledge the assistance of Mr Colin Newlands (Natural England), Dr Mark Fisher (Leeds University) and sup- port from Defra and Professor Kirill Horoshenkov of Sheffield University and coordinator of UKAN. 9. REFERENCES

1. Agius, L.J. An investigation into whether acoustic data from birds and mammals can indicate a

change in species richness in Colt Park Wood. BSc (Hons) dissertation. University of Leeds, School of Geography (2020) 2. Krause, B., Gage, S. H., & Joo, W. Measuring and interpreting the temporal variability in the

soundscape at four places in Sequoia National Park. Landscape Ecology , 26(9) , 1247-1256 (2001) 3. Marafa, L. M., Tsang, F., Watts, G. & Yuan, X-M. Perceived tranquillity in green urban open

spaces. World Leisure Journal , 60(3) , 221-234 (2018). 4. Pheasant R J, Horoshenkov K V and Watts G R. Tranquillity rating prediction tool (TRAPT)

A coustics Bulletin, 35 (6) , 18-24 (2010) 5. Watts, G. R., Miah, A. & Pheasant, R. J. Tranquillity and soundscapes in urban green spaces -

predicted and actual assessments from a questionnaire survey. Environment and Planning B: Planning and Design , 40(1) , 170–82 (2013)

worm 2022