Mike Ko Personal Portfolio

 

Home-School Education
2001-2012
Hong Kong

University of Durham
Bachelor of Science
2014-2017
United Kingdom

University of Sussex

Master of Arts
2017-2018
United Kingdom




Academic Works
                                Year 3
: Fieldcourse

Assessing the Possible Correlation between Estimated Large
Animal Densities and Dung Count Densities along Transects in
South African Bushveld

Abstract

            Accurate measurement of species density is important in ecology for both general characterisation and for tackling applied problems. Different methods can be used to estimate species density, although each bears trade-offs in convenience and quality. Confidence in such estimates can be increased if they are reliably correlated to other proxies. This study thus evaluated the possible correlation between large animal density estimates from distance sampling and dung density. Data collection was done on four separate transects through land patches with different burn years in the Mankwe Wildlife Reserve in South Africa. Distance sampling was done through the line transect method while dung counts were simultaneously made on strip transects. The derived animal density and dung density figures were sorted by burn year and statistically tested for correlation with the Pearson correlation coefficient test. No significant correlation was found between the two proxies of species density. Many factors can cause estimates or proxies to appear non-correlated despite being indices of the same variable. Ultimately, the need for identifying correlation between proxies may depend on the confidence in existing density estimates.


Introduction

           Species density and the ability to measure it are crucial in ecology for many reasons. At a fundamental level, knowledge of species densities can help characterise ecosystems. Information on density is also required to assess population dynamics and the potential influence of density-dependent factors on intra- and interspecies interactions. Such information is crucial for gaining a better understanding of population regulation as well as for implementing effective protection and management of endangered species populations. For example, densities of endangered black rhino populations inhabiting game reserves must be regulated to prevent territorial interactions from lowering fecundity and raising mortality, as caused by relatively high densities (Greaver, Ferreira and Slotow, 2013).

           However, it is mostly impossible to directly measure absolute densities, whether it is due to practical considerations or observer bias. Thus densities are usually estimated through methods such as distance sampling, which uses the measured distances between the observed species and a given point for calculating density (Buckland et al., 1993). There are multiple types of distance sampling, and whether a specific method is appropriate depends on the species being measured. For example, line transects can be suitable for low density, mobile species such as large mammals that inhabit areas with comparatively homogenous habitats. Although not the most accurate, this method allows for more data to be obtained for a reasonable sampling time, while simultaneously accounting for the detectability of the measured species by observers.
Other indirect approaches to estimating species density also exist and can be useful. One example is using dung counts and the derived dung density as a proxy for species density (Plumptre and Harris, 1995). Although dung density can be affected by other confounding factors like defecation rates, it can be useful when the target species are in habitats where direct sightings maybe difficult.

           In general, each approach for estimating species density is not perfect and has its advantages and shortcomings. However, proxies for species density may mutually bolster these estimates’ validity if they are constantly consistent with each other. Accordingly, this study sought to assess the possible correlation between the estimated large animal densities from distance sampling and dung densities measured in a South African bushveld habitat.


Methods

           The study was carried out on 27 and 28 September 2016 in the Mankwe Wildlife Reserve in South Africa (approximately located at 25°14'56.8"S 27°19'24.7"E), between 07:00 to 18:00. Originally a buffer zone for a explosives factory, the 4750 hectare area is a dedicated game reserve as of this writing, with a predominantly bushveld habitat composition. The entire reserve is separated into different patches of land that were subjected to controlled burning in different years, ranging from the year 2010 to 2015. Large mammals present in the reserve include various African antelopes, giraffe, warthog, zebra, hares. The reserve contains multiple artificial water pans to sustain the local wildlife.

           Data collection for both large animal and dung density estimations were done simultaneously on four transects within the reserve, two separate transects for each day of study, as indicated in Figure 1. The four transects were non-randomly selected to account for logistics limitations faced by the group of three students carrying out the observations. Species on which line transect and dung count data were to be collected for were mostly  large mammals including antelopes, baboon, giraffe, warthogs and zebra. Although not a mammal, ostriches were also included to increase the available data.

Large Animal Distance Sampling on Line Transects

           For obtaining estimates of large animal density, the following variables for contacted large animals were recorded throughout the entirety of each line transect: 1) sighted animal’s distance from observer; 2) angle of sighted animal from transect line; 3) number of individuals forming the group of the sighted animal species; and 4) the year in which the land patch the animal was observed in had been burnt (“burn year”). Data for sighted animals’ distance and angle from the transect and observer, respectively, were used to calculate the animals’ perpendicular distance from the transect through the trigonometric relationship x = r sin θ.

           Due to the study’s relatively small scale, there was insufficient data points for individual burn years within each transect. Thus the calculated perpendicular distance figures from the four transects were sorted into four data sets based on their burn years (2011, 2012, 2013 and 2014). This effectively created four new virtual transects, each formed by combining transect segments associated with a single burn year.

Figure 1. Map of Mankwe Wildlife Reserve. Each coloured line marks the rough path of the four transects on which data collection were carried out, either in the morning (AM) or evening (PM).

           Using these four data sets of perpendicular distances, estimated animal densities were obtained using the following equation:

D = ( n1 + n2 / 2rl ) x ln (n1 + n2 / n2 ) x 10000

where:

  • ‘D’ is the estimated animal density.
  • ‘l’ is the combined transect length for a given burn year.
  • ‘r’ is the distance that separates the two recording zones, here defined as the distance that includes at least half of all animal sightings.
  • ‘n1’ and ‘n2’ are the number of animal sightings within and beyond r, respectively.

           For simplicity, it was assumed that large animals’ detectability along the line transect decreased exponentially with distance from the observer, hence only two recording zones (≤ r and r <) were set. The estimated densities obtained from the equation were then multiplied by the average group size of the animal sightings within each burn year data set to obtain the final estimated large animal density figure.

Dung Density

           Recordings for dung density data were carried out simultaneously on the same four transects as before, although as strip transects instead of line transects. Dung counts for large animals were recorded in strip areas of 50 meter by 2 meter (1 meter from both sides of the transect).  In the first transect (done on 27 September), data recording for a strip was done every 100 meters along the transect, while for the remaining three transects this was modified to every 200 meters. Each dung count was defined as a dung pile that had more than seven pellets forming a close group. Again, dung counts from all four transects were sorted into four data sets based on their associated burn years. Dung densities for each burn year were calculated by dividing the associated total dung counts by the combined total strip area for the given burn year.

           Because dung density can be dependent on species density, statistical analysis for the two proxies was done using a Pearson correlation coefficient test.


Results

           Table 1 shows the density estimates from line transect distance sampling and dung density calculated from the obtained data.

Table 1. Estimated Large Animal Densities from Distance Sampling and Dung Densities for Virtual Transects of Different Burn Years in Mankwe Wildlife Reserve

Year

2011

2012

2013

2014

Animal Density (ha-1)

0.85

0.33

0.21

0.65

Dung Density (x106  ha-1)

515

573

439

371

 

           Estimated large animal densities varied with areas of different burn years, although it did not show any linear trends over time. In contrast, dung density was roughly consistent regardless of burn year.

           Statistical analysis of the densities with the Pearson correlation coefficient test did not show a significant correlation between the two estimated densities (r = -0.24, N = 4.00, p = 0.98). The results are also illustrated in Figure 2.

Discussion

           This study sought to evaluate a potential correlation between large animal density estimates obtained by distance sampling on line transects and dung densities, both of which are proxies for actual species density. Calculated density figures suggest that there was no correlation between these two proxies. However, the existence of multiple shortcomings in this limited study means that the results are not definitive evidence against correlation between the two proxies of species density.

           In particular, this study did not account for factors that could affect dung density, such as defecation and dung decay rates. These factors were not considered in this study due to its limited scope. It is possible that had dung density been calculated with approaches that account for such factors, the resulting densities might actually be correlated to the estimated densities from distance sampling. It may also be useful to do a correlation test between distance sampling densities and animal density estimates derived from dung density, instead of using the latter itself directly.

           The short time frame within which the study was done meant that the results could be affected by environmental stochasticity. For instance, the reserve experienced rainfall after an undefined period of relative drought on the evening of 27 September. This event could have potentially influenced animal activity and thus the number of sightings on the next day of study in ways not yet reflected by dung densities. A longer study period could possibly minimise such indirectly caused sampling bias.

           The study results suggest that burn years does not affect distance sampling density estimates nor dung densities. However, the method of analysing the virtual transects as categorised by burn year could mask local trends and influences on density acting on the original transect (e.g. proximity to water pans), should these factors be examined. Ideally, sufficient data for individual burn year should be gathered within each individual transect and analysed separately.

           It should be stressed that the apparent lack of correlation does not mean that methods for estimating species density will be inaccurate. For example, modelling studies found that dung-based indices had a positive correlation with known deer density (Forsyth et al., 2007).

           There are different indications as to whether estimates or proxies for species density should be correlated or not. Past studies suggest that dung density can give similar estimates for species density compared to other estimation methods (Barnes et al., 2001). Thus as long as different methods for estimating species densities are acceptably accurate, it can perhaps be argued that there will be some kind of correlation between such proxies. On the other hand, proxies can be influenced by a variety of factors besides species density, hence correlations may not always be found. This is illustrated by the dependence of dung density on defecation and decay rate, for example. Line transects estimates can also become less accurate with increasing target animal mobility (Glennie, Buckland and Thomas, 2015). Indeed, different factors’ potential influence should be accounted for (perhaps on a environ-by-environ basis) before relationships between the proxies can be established (Rivero, Rumiz, and Taber, 2004).

           Whether it is worthwhile to devote effort towards evaluating relationships between estimates or proxies may ultimately depend on how accurate existing estimation methods are. If there is any reason to question existing estimates of species density, then comparison between two or more proxies can help to either remove or confirm any doubts about the estimate. Identifying correlations between two proxies can help validate the estimate in question, especially when both proxies have figures that are not immediately comparable (e.g. animal density against dung density).


References

 

Back to Top