May 2006 Staff Seminar

Argus pheasant habitat

The tiger project in Bukit Barisan Selatan NP in Sumatra set up camera traps at 197 randomly-slected locations in the Park. They were mainly interested in photographing tigers, but a range of other animals were captured, including Argus pheasants.

At each camera location, the team measured:
o elevation (altitude) in metres
o the number of saplings <10cm dbh within 10m of the camera
o percentage canopy opening, as measured by a densiometer
o the DBH of the biggest tree within 10m of the camera

We looked at several models to see which of these habitat variables are linked to the presence of Argus pheasants, assuming that camera trapping at each site was sufficiently intensive to be sure of capturing a pheasant if the habitat was in fact being used (ie no ‘false absences’). We used Generalized Linear Modelling, which sounds very grand, but we did it in a simple Excel spreadsheet, so that we could see all the steps and how they relate to maximum likelihood and AIC.

Details of the analysis are given in the first page of the spreadsheet.

Source: WCS stats course, June 2005

Reference : O'Brien, T G; M F Kinnaird; H T Wibisono. 2003. Crouching tigers, hidden prey: Sumatran tiger and prey populations in a tropical forest landscape. Animal Conservation 6:131-139. But see also Keating, K A; S Cherry. 2004. Use and interpretation of logistic regression in habitat-selection studies. J Wildlife Management 68:774-789 for assumptions and limitations of the method.

Download spreadsheet
(.xls 61KB)

Bayes in brief

A short introduction to Bayesian approaches, using the same examples and data sets we had earlier used in the "Basic Concepts" section at the start of the workshop. 

Bayes for belugas

An isolated population of about 350 beluga whales lives in Cook Inlet, Alaska, and they are hunted by native Alaskans. Aerial surveys were carried out from 1994 onwards, and regression analysis was used to try to detect a trend in the population.

Not until 2000 – with 7 years of data – was there significant evidence of a decline according to a hypothesis-testing approach. But by 1998, Bayesian analysis was already showing a 93% probability that the population was declining and a 79% probability that the decline exceeded 5% per year. As a result, a moratorium on hunting was implemented in 1999.

These analyses were done in Excel. We also looked at how loss functions are used together with Bayesian analysis to help in decision-making.

Reference: Wade, Paul R. 2001. The conservation of exploited species in an uncertain world: novel methods and the failure of traditional techniques. Ch 6 in Reynolds, J, G M Mace, K H Redford, and J G Robinson, editors. Conservation of exploited species. Cambridge University Press, Cambridge UK.

"Conservation
of Exploited Species"

Download spreadsheet
(.xls 40KB)

Mark-recapture studies and MARK

This section has been revised and now uses data from a study of tigers in Kanha National Park, Madhya Pradesh, India.

Mark-recapture studies can give excellent information on populations abundance and trends, on survival and recruitment of members of the population, and many other ecological processes. The MARK software package covers a range of possible analyses, but we only looked at population abundance, using  some of Ullas Karanth’s tiger data from camera trapping work in India.

It included –

• some comments on what MARK does
• how to install and start MARK
• setting up a project and an input file for MARK
• running a simple analysis and interpreting the output
• model specification with PIMs
• model selection using AIC, delta-AIC, etc
• comparison with program CAPTURE – goodness of fit tests, model selection, additional models
• getting finished.

The study is described in Karanth, K U; J D Nichols; N S Kumar; W A Link; J E Hines (2004) Tigers and their prey: Predicting carnivore densities from prey abundance. Proceedings of the National Academy of Sciences 101:4854-4858

Data file
(.zip, 1KB)

That's all we had time for during the May 2006 workshop. Below you will find some of the material I had prepared but which we didn't have time to look at, plus a unit we used during the 2005 workshop and which wasn't revisited this year.
  

Logging intensity and woodpeckers

Instead of just comparing logged vs unlogged site, Martjan Lammertink compared woodpecker communities at 8 lowland sites with different intensities of logging in West Kalimantan. One site in Gunung Palung NP was undisturbed, 2 sites had some illegal logging but no roads or skid tracks, and 5 sites had been logged commercially. He also looked at 2 sites in unlogged hill forest.

We will use simple linear regression to see how woodpecker density and biomass relate to basal area cut, extent of disturbed patches, and time since logging. We will also see if we can detect any effects of logging on individual species, though we may not have enough data per species to see significant effects.

This can be done in Excel, but it gets a bit tedious – there are 14 species of woodpecker – and it is easier to do in R.

Reference: Lammertink, M. 2004. A multiple-site comparison of woodpecker communities in Bornean lowland and hill forests. Conservation Biology 18:746-757. See also Steury, T D; A J Wirsing; D L Murray. 2002. Using multiple treatment levels as a means of improving inference in wildlife research. J Wildlife Management 66:292-299.

The power of monitoring programs and MONITOR

• a brief explanation of the concept of power in the context of monitoring biological populations;
• how to install and start MONITOR
• setting up an analysis in MONITOR
• an explanation of what MONITOR does to produce its results
• getting finished.

On the way, we will use MONITOR to design a monitoring programme for bears in Dachigam WS, Kashmir.

Unlike most biological / statistical software, MONITOR is a study design tool: it will not analyze your data!

Lab guide
( .pdf 49KB)

Other labs are in the works:

• Vital rates with MARK
• Habitat suitability using linear regression and ANOVA in R
• Modeling trends in population abundance with GAMs in R

The handouts and "lab guides" on the site are designed to be 'stand alone' documents, useful to those staff who could not attend the seminar - or anyone else who surfs to the site for that matter - as well as participants. If you find the downloads aren't sufficiently self-explanatory, or if you have other suggestions and comments, please let me know: .