Sound-mapping a coniferous forest-Perspectives for biodiversity monitoring and noise mitigation

Abstract

Acoustic diversity indices have been proposed as low-cost biodiversity monitoring tools. The acoustic diversity of a soundscape can be indicative of the richness of an acoustic community and the structural/vegetation characteristics of a habitat. There is a need to apply these methods to landscapes that are ecologically and/or economically important. We investigate the relationship between the acoustic properties of a coniferous forest with stand-age and structure. We sampled a 73 point grid in part of the UK's largest man-made lowland coniferous plantation forest, covering a 320ha mosaic of different aged stands. Forest stands ranged from 0-85 years old providing an age-gradient. Short soundscape recordings were collected from each grid point on multiple mornings (between 6am-11am) to capture the dawn chorus. We repeated the study during July/August in 2014 and again in 2015. Five acoustic indices were calculated for a total of 889 two minute samples. Moderate relationships between acoustic diversity with forest stand-age and vegetation characteristics (canopy height; canopy cover) were observed. Ordinations suggest that as structural complexity and forest age increases, the higher frequency bands (4-10KHz) become more represented in the soundscape. A strong linear relationship was observed between distance to the nearest road and the ratio of anthropogenic noise to biological sounds within the soundscape. Similar acoustic patterns were observed in both years, though acoustic diversity was generally lower in 2014, which was likely due to differences in wind conditions between years. Our results suggest that developing these relatively low-cost acoustic monitoring methods to inform adaptive management of production landscapes, may lead to improved biodiversity monitoring. The methods may also prove useful for modelling road noise, landscape planning and noise mitigation.

Fig 1

A) United Kingdom coastline. Thetford Forest (black dot) is situated in East Anglia. B) Map of the main central Thetford Forest block. Thick dark lines indicate busy A-roads. Thinner dark lines indicate minor roads. Black dots represent study grid. C) Study grid. Dots represent sampling points (n = 73), which are spaced 250m apart. Polygons represent different forest stands. The thick dark line on the Western edge of the grid is the A1065 and the thick dark line to the North East of the Grid is the A134 –two busy main roads.

Fig 2. Relationships between acoustic indices with coniferous forest stand age (N = 65) from each year (dots = 2014 data; triangles = 2015 data).

Letters represent mean groupings from the Gabriel post-hoc test (p<0.05). A) Mean ADI (+/-2 SE). 2014, F_7,57 = 6.896, p<0.001; 2015, F_7,57 = 18.772, p<0.001. B) Mean AEI (+/-2 SE) 2014, F_7,57 = 5.417, p<0.001; 2015, F_7,57 = 14.359, p<0.001. C) Mean NDSI (+/-2 SE) 2014, F_7,57 = 5.827, p<0.001 (abc); 2015: F_7,57 = 5.010, p<0.001. D) Mean BAI—there were no significant differences between age-groups but the plotted means indicate that in 2014 the mean BAI values were higher in older stands.

Fig 3. Canonical correspondence analysis exploring the relationship between habitat type habitat features and ten 1KHz frequency bands.

(see Table 1 for key to category labels; R² cut-offs for environmental variables = 0.1; TRSp = no. of tree species; CNHT = Canopy Height; GCHT = ground vegetation height; GCDV = ground vegetation diversity). A) 2014 data. Strong associations between axis 1 with CNHT and TRSp (S1 Table) indicate that as structural complexity increases, the higher frequency bands become more apparent in the soundscape. B) 2015 data. Similar relationships between axis 1 and habitat structural metrics (S1 Table) show fairly similar distribution of sites in relation to frequency bands. See Results section for explanation of key findings.

Fig 4

Interpolation maps of average normalised difference soundscape index (NDSI) scores from each sampling point from 2014 (A) and 2015 (B) (points are 250m apart). Darker shading indicates higher levels of anthropogenic/technophonic sounds (i.e. from machinery) in the soundscape. Lighter shades indicate higher levels of biological/biophonic sounds. The grey and black striped line on the left of each map is the A1065, a busy main road. The lighter grey lines indicate smaller connecting roads. The lighter grey lines indicate smaller connecting roads. C) Strong positive relationship between NDSI and distance to nearest road 2014 data (r2 = 0.373, p<0.001, N = 73) displayed as circles; 2015 data (r2 = 0.397, p<0.001) displayed as black triangles. The lines of best fit for both datasets overlap so are not distinguishable from one another.

Fig 5. Ten second spectrograms illustrating different contributors to the soundscape.

A) An example of a recording with high ADI, BAI and NDSI values. Three separate bird species calls are highlighted by white squares. This kind of frequency partitioning is one of the key concepts in soundscape ecology and ecoacoustics. B) Recording from an older coniferous stand (>45yrs) capturing what appear to be contact calls of Regulus regulus (goldcrest). C) Highlighting the presence of Orthoptera within our recordings. Although we only utilised audio data up to 11KHz in our statistical analyses, this recording shows that the Orthopterans in our recordings are occupying higher frequencies (up to 22KHz). D) Another recording with high values in the 0-10KHz range. The three darker vertical patches highlight the sound of a flying insect (potentially Syrphidae) passing the microphone. This kind of frequency modulated pattern might be a useful indicator of winged-insect (i.e. pollinator) activity. E) Recording displaying very low NDSI values (-0.95 –i.e. high road noise) which is evident from the thick black band filling the 0-2KHz frequency range.