Occupancy modelling for longer-term monitoring

questions concerning analysis/theory using program PRESENCE

Occupancy modelling for longer-term monitoring

Postby leroy » Tue Aug 07, 2018 8:39 pm

Hi,

I'm looking to monitor a network of sites over a number of years. The plan at this stage is to monitor around 250 sites in total, with ~83 sites monitored each year (multiple nights in a year). It would be good to analyse data after all sites were sampled in at least one year. I was thinking of producing an occupancy map based on the first 3 years worth of data, with a view to do same with the next 3 years of data to identify areas where occupancy has changed across a broad landscape.

I was wondering whether a multi-season modelling approach could still be used given each site isn't sampled every year. If this is possible, I assume sites not sampled in a given year would be coded as missing data (.) in PRESENCE?

Any advice would be appreciated.

Cheers,

Leroy.
leroy
 
Posts: 10
Joined: Thu Aug 04, 2016 4:04 am

Re: Occupancy modelling for longer-term monitoring

Postby jhines » Tue Aug 07, 2018 10:33 pm

If your interest is only occupancy maps each year, then you could run each years data as separate analyses in PRESENCE, and you can code site/surveys which weren't sampled as missing (-). A better option would be to analyze all years together, with year as a covariate. This would allow you to build models with shared parameters (eg., same detection probabilities across years).

If you're interested in what drives changes in occupancy, then the multi-season model is needed. Missing data are ok for that model as well, but you will need some sites which are monitored across years. The more data you have across years, the better the estimates of colonization and extinction, which are used in the computation of subsequent occupancy estimates.
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA

Re: Occupancy modelling for longer-term monitoring

Postby leroy » Wed Aug 08, 2018 10:10 am

Thanks very much for your response Jim. That helps a lot.

Much appreciated!
leroy
 
Posts: 10
Joined: Thu Aug 04, 2016 4:04 am

Re: Occupancy modelling for longer-term monitoring

Postby leroy » Mon Sep 03, 2018 2:49 am

Hi Jim,

Just wondering if there is any guidance on what proportion of sites (e.g., 0.25 or 0.5) should be resurveyed each year?

Thanks,

Leroy.

jhines wrote:If your interest is only occupancy maps each year, then you could run each years data as separate analyses in PRESENCE, and you can code site/surveys which weren't sampled as missing (-). A better option would be to analyze all years together, with year as a covariate. This would allow you to build models with shared parameters (eg., same detection probabilities across years).

If you're interested in what drives changes in occupancy, then the multi-season model is needed. Missing data are ok for that model as well, but you will need some sites which are monitored across years. The more data you have across years, the better the estimates of colonization and extinction, which are used in the computation of subsequent occupancy estimates.
leroy
 
Posts: 10
Joined: Thu Aug 04, 2016 4:04 am

Re: Occupancy modelling for longer-term monitoring

Postby jhines » Mon Sep 03, 2018 10:36 am

Hi Leroy,

It depends on a number of things. One of them is how precise you want the estimates to be, which will depend on the number of sites, number of surveys, and parameter (psi, gamma, epsilon, p) values.

You can use the tool, GENPRES, to answer this question. Just generate data for a group of sites which will be surveyed on all occasions, and a group of sites which will only be surveyed once (p=0 for all but one survey for this group), You'll need to enter your best guess of initial occupancy (psi), colonization (gamma), extinction (epsilon), and detection (p), then let the program run a standard multi-season model. Look at the resulting estimates and decide if the standard errors are acceptable. If not, repeat with a different sampling schemd (eg., more sites, or more sites surveyed on all surveys, higher detection probability,...)

Cheers,

Jim
jhines
 
Posts: 599
Joined: Fri May 16, 2003 9:24 am
Location: Laurel, MD, USA


Return to analysis help

Who is online

Users browsing this forum: No registered users and 7 guests