Add your answers to the Distance sampling Survey design assignment below
Choose whether to answer the section on:
I WOULD LIKE TO MAKE AN ATTEMPT AT DESIGNING A SURVEY. HAVE AN ASSIGNMENT TO CARRY OUT ASSESSMENT OF WILDLIFE AT AN AIRPORT IN OUR AREA WHICH IS A 2KM BY 2KM PREMISE, AND MAKE RECOMMENDATIONS ON HOW THESE CAN BE CONTROLLED FOR THE SAFETY OF FLIGHT MOVEMENTS.
OBJECTIVE
-to identify animal and bird species within Survey Airport
-to make recommendations on control measures for identified species
SURVEY EFFORT
-We conducted a pilot study at the airport premise over a length of 1.5Km, and in that length we observed a total of 10 birds and 8 small mammals. Using the recommended constant of 3 and coefficient of variation of 10%, I calculated total transect length to be:
L=(b/cv2) x (Lo/no)
=300 x 0.0833
=24.9Km
b. Repeat surveys: I will conduct a repeat survey as I am interested in data for both the wet and dry season.
c. Randomisation: I plan to use parallel transects of 2Km each, which will start randomly from the eastern fence in one survey, and western fence in the following survey. The transects will run in a perpendicular direction to the runway.
4.COVARIATES
I will consider recording of covariates of any features that may be possible wildlife attractants, such as vegetation/trees/shrubs and physical developments that may attract wildlife. These are important as they are determinants of wildlife presence in the premises.
Objective: To know the total number of buffalos in the private game reserve to inform translocation.
Survey effort –
Survey effort:14100km.
Used data from a pilot aerial study to calculate using the following equation;
L=(b/cv2)∗(L0/n0)
Where
L= total transect length
b = a constant (see below) 3
cv = desired coefficient of variation 10%
L0 = distance surveyed for pilot study 141km
n0 = number of sightings recorded during pilot study 30 buffalo
Pilot study
Pilot study carried out and used to calculate the survey effort.
Sampling regime
Thinking of stratification based on the area of the blocks.
| Survey block | Area (sqkm) | Total transect length(m) | Desired transect length | Number of transects |
|---|---|---|---|---|
| Survey WHS 5 | 2 | 1175 | 100 | 12 |
| Survey WHS 4 | 4 | 2350 | 100 | 24 |
| Survey WHS 3 | 4 | 2350 | 100 | 24 |
| Survey WHS 1 | 7 | 4112.5 | 100 | 41 |
| Survey WHS 2 | 7 | 4112.5 | 100 | 41 |
Transect placement
Boundaries are well defined but now how do I generate the random lines? See below attempt.
No Covariates
Key question
how can i make them randomly placed, i want to then load these lines into a GPS as tracks.
OBJECTIVES
To estimate population densities of predators (Lion, Leopard , Spotted hyena, Brown hyena, Black backed jackal and caracal) in the Chobe National Park using spoor surveys.
SURVEY EFFORT
A pilot study was carried out using the following equation;
L=(b/cv2 ) ( L0/n0 ) where;
L=Total transect length
b=A constant (3 for this survey)
cv=Desired coefficient of variation (in this survey 10%)
L0 =Distance surveyed for pilot study
n0 =Number of sightings recorded during pilot study
The pilot study was conducted for 30 kilometres in one of the transects and the following spoors were recorded;
• Lion 6
• Spotted hyena 12
• Brown hyena 4
• Leopard 4
• Black backed jackal 8
• Caracal 2
A total of 36 spoors recorded for the 30 kilometre transect. Therefore survey effort at 10 % coefficient of variation was;
3000.833=249.9Km
Sampling Regime; The survey will be divided into 18 transects across the park ,each transect will be 30 Km long.
Sampling to be repeated biennially; during wet and dry seasons of the year, to monitor seasonal trends.
Transect placement; The transects will be placed on already existing roads in the park .The best method that could be used in this survey is stratification because the park has several habitats since it is made up of four different ecosystems ,which will be better off sampled as different strata. Boundaries of the Park are well defined.
No covariates
Well done for being brave enough to go first, @Reneilwe_Thobosi - that’s never easy! 
You’ve presented a comprehensive answer here, and raised some very relevant points
This calculation is correct based on the information from your pilot study, but I don’t think you should combine your bird and small mammal sightings. These species will likely have differently-shaped detection functions because you will be detecting them in different ways, and you will be analysing data from birds and mammals separately. This may mean you need to almost double your total transect length, unless you can adjust your survey methods to increase detectability
Bear in mind also that you will be doing repeat surveys:
Your total survey length can be shared between the seasons, so that you only need to survey 12.5km in each season. Are density or detectability likely to differ between the two seasons, due to changes in vegetation or animal movements/breeding patterns? If so, you may need to include season as a covariate
Indeed - your encounter rates will vary through space and time, and such a short survey doesn’t provide much information on whether that encounter rate will be achieved in the main study. Can you find any information on encounter rates for your species and environment from other surveys your team has done, or in grey literature (internal reports) or scientific papers by researchers elsewhere in sub-Saharan Africa?
Good thinking! Disturbance or vegetation differences caused by the runway will be evened out across each transect
It’s great that you’re considering these possibilities. In a small study are like this, with relatively little space for transects, I recommend that you choose one covariate that is likely to have the strongest influence on density and/or detectability. Alternatively, you can include more covariates by placing a larger number of transects closer together, but don’t survey neighbouring transects immediately after each other to avoid disturbance
This depends on what your study aims are. If your surveys need to be sensitive to small changes or differences in density e.g. 5% decline or increase, you will need to use a small coefficient of variation (5%), so you gather enough sightings to make more precise predictions. Some examples where a small cv might be necessary:
Alternatively, you may be content with detecting only large changes/differences in density, such as a 30% difference between two management areas, or inside/outside the breeding season. In this case, using a larger cv (30%) will require you to survey a much shorter total distance. You will have fewer sightings, but your estimates should still be sufficiently precise to detect a 30% difference in density
I hope that helps. Great work! 
This is a great starting point, @Nicorll, and has some really useful information to clarify your planning! 
I think you have a design challenge on your hands, assuming I have correctly understood all the information you provided
With an encounter rate of c. 20 buffalo per 100km, you’ve got a large survey effort for this project. Aerial surveys, or an environment that is easy to drive over, will make this more feasible, but the logistics of placing your transects need thought, given the size of your study area
Some follow-up questions:
How does your encounter rate of c. 20 per 100km compare with that in similar studies? Is this normal for buffalo in your environment in Zimbabwe/South Africa etc
Where did you do the pilot study? Was it a larger area than the combined survey blocks you are stratifying by?
What is the difference between the blocks? What leads you to choose them for stratification? Do you expect differences in density or detectability, or is it simply a way to organise your transects so that you’re surveying the whole area evenly, or can deploy teams effectively? See below for the challenges presented by your small block areas, combined with your low encounter rate
Is your desired transect length km, or metres? I’m assuming metres, given that multiplying your number of transects by 100m gives your proposed transect lengths in each survey block
Each survey block is only a few km wide. This means that a transect stretching across even the largest blocks (4.1km) will have 0.88 sightings on average. In other words, many of your transects, especially in the smaller survey blocks, will have zero sightings. This can cause problems during analysis
If you need to stratify by your survey blocks, I strongly recommend that you extend your transects so that they stretch all the way across each survey block, rather than being 100m long. You need to minimise the number of transects with zero sightings. Also consider the necessity of the blocks
If you need to survey 14 thousand km, but your study area is only 24 sqkm in total (i.e. equivalent to a 5x5km square), how do you intend to fit in those surveys?
What is the balance between replicates (geographically-separated transect lines) and repeats (return visits to the same transects)?
I assume you will be doing repeat surveys. Otherwise, in a ‘back of the envelope’ calculation, without repeats you’ll end up with parallel transects that are only 2m apart![1]
If you don’t need to include covariates, you can prioritise repeats over replicates, provided that you have the minimum recommended number of 10-20 replicates. That means your transects can be further apart, reducing the overlap of detection zones and inter-transect disturbance
If you’re familiar with GIS software such as QGIS, you can use it to:
Let me know what GIS expertise you have - I’m happy to give more detailed advice!
This can also be done on a paper map, for those who haven’t got GIS expertise or assistance:
Do let me know if I have misunderstood anything. Well done for providing enough detail for us to get our teeth into your survey challenge! 
Based on a square study site: 14,000km = 2,800 x 5km long transects, fitted into a 5km-wide space = c. 1.8m between each transect ↩︎
Well done for submitting your assignment, @Emmanuel_Letebele!
A few questions to clarify your survey design:
It sounds like you’re recording detections through tracks/scat, rather than sightings - is that right? Remember that you’ll need the additional calculations (sign creation & decay rates) to convert sign density to animal density
Are the detection patterns (i.e. shape of detection function) likely to be the same for all your species? Think about their behaviour and ecology, social group size, the timing of your surveys, detection cues for each species etc. Is it valid to combine them all in your analysis, to build a single detection function for all of them? If not, how might you adjust your calculation of survey effort? Once you’ve done that, consider the results - is it feasible to use distance analysis for every species?
Thinking about your species of interest, what effect(s) might using roads have on your data collection and results? What could you do to limit any effects?
Is your transect length and layout based on an existing monitoring system? I’m guessing yes, because your total survey effort each year will be 1,080km[1], which is obviously greater than the c. 250km you calculated above.
If you adjust your survey effort after my comments above, will that lead to any changes in your sampling regime? If you’re restricted to repeating a sampling regime that was designed in the past historically, how might you adapt the regime to maintain data compatibility with your historic record whilst giving you what you need for distance sampling analysis?
Do you expect densities or detection probabilities for any of your species to differ between strata? If so, you might want to include stratum as a covariate in your analysis for a given species
Lots to think about here, and these are not always easy questions to answer! Do reply if you’re not sure how to proceed with anything, as I aim to make the course useful for your actual work, rather than just a study exercise
This is a useful base to build on - good work!
18 transects x 30km x 2 seasons ↩︎
This is a concise summary, Lorato. You probably were working against the clock to submit it, and I’m glad you succeeded as it gives us a springboard to begin from!
Adding a little more detail would help us understand your plans and help you shape your potential project. Your summary sparks a few practical questions, similar to some I asked Emmanuel and Nicorll above
Which rare ungulate species are you surveying? Are they similar enough in detectability (size, behaviour, habitat preferences, social groupings etc) to combine into a single survey effort calculation?
Roughly how long can each transect be? Is the overall size of your survey area likely to limit your transect length, as it does for Nicorll?
Well done for considering this point. When you begin collecting data, keep an eye on what proportion of transects have zero sightings. Sighting numbers will obviously vary through space and time, and you might find that you need to adjust the design or collect related covariates to explain that variability
I can see that you’re trying to avoid bias in your sampling, but by specifically avoiding areas (roads animal paths) that you believe will have high or low density, you are creating bias by not sampling the entire area. I recommend that you include all areas that your species can access in your surveys. By placing your transects at random, there will be no systematic bias caused by transects that follow travel routes etc
It’s great that you’re including distance to water as a covariate, and this means that you can test a hypothesis about the impact of the water sources on density and/or detectability, rather than make an assumption
Overall, you’ve put some good thinking into this - well done!
@Emmanuel_Letebele and @Lorato_Esele , to complete your assignments, please share a remaining uncertainty or question that you have about distance sampling survey design
Then I can award you your badges for this assignment 