Single vs Multi Season

[Emily Collins]

Hello, can anyone advise on whether I should use a single season or multi season occupancy model for 12 consecutive months of camera-trap data? Multi-season seems logical at first, but I've seen authors use a single season model for 9 months of data, and then use 'season' as a covariate for detection probability. It also seems to me like the multi-season model is meant for surveys over multiple years - sampling each season more than once.

Any advice would be greatly appreciated!!

Emily

[Jose Infante]

Hi Emily,

It depends on your research question(s) and the amount of data. A season is not necessarily a year, it can be another time frame. Multi-season (or dynamic) models are commonly used to estimate colonization and/or extinction probabilities, and to test hypotheses on these parameters (e.g. using covariates). This needs good amounts of data. If you apply a single-season model to 12 months data it seems to me that you will violate the closure assumption. If you don't need to estimate dynamic parameters but you'd like to test for temporal variation in occupancy you could apply a single-season model with random effects (being random effects the camera trap or sampling unit), by stacking months (i.e., your sampling units would be camera-months).

That's my thinking, hope it helps,

Best,

José.

  

José Domingo Infante Varela Ing. Agrónomo, MSc en Recursos Naturales, Estudiante de Doctorado en Ecosistemas Forestales y Recursos Naturales UACh   Investigador Asociado | Instituto de Conservación, Biodiversidad y Territorio | Laboratorio Fauna Australis UC    mobile: +56 9 88275395 email: join...@uc.cl

--
*** Three hierarchical modeling email lists ***
(1) unmarked (this list): for questions specific to the R package unmarked
(2) SCR: for design and Bayesian or non-bayesian analysis of spatial capture-recapture
(3) HMecology: for everything else, especially material covered in the books by Royle & Dorazio (2008), Kéry & Schaub (2012), Kéry & Royle (2016, 2021) and Schaub & Kéry (2022)
---
You received this message because you are subscribed to the Google Groups "unmarked" group.
To unsubscribe from this group and stop receiving emails from it, send an email to unmarked+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/fb66a996-308a-46ab-abcf-aef58ecdeed8n%40googlegroups.com.

[aw...@scenichudson.org]

Hi Emily,

I'd second what Jose said.  Season is defined by closure - there should no unaccounted-for changes of occupancy within season; changes of occupancy happen between seasons.  You would need to define closure and season for your study area/species/study design.

-Alex

[Emily Collins]

Hi and thank you both for your suggestions!

My research question is to look at site-specific occupancy for gorillas and chimpanzees within a multi-use landscape (19 camera sites over 12 months, with a detection matrix of 52 sampling occasions of one week, although not all sites have 52 sampling occasions). I am doing a separate occupancy model for each species but I will compare them after.

I know there is monthly variation in site occupancy (i.e both species use different sites during different months, although I don't believe either species leaves or enters the study area entirely during the 12 months). Basically I want to be able to determine which sites specifically are used during which months while accounting for imperfect detection. I am also interested in whether certain sites are used by both species during the same time-frame (month, or season) but for now I want to run two separate occupancy models.

The columns of my dataset (I have 1 dataset for each spp) are currently as follows:

site id [1] n=19

site covs  - habitat type (Forest, Road, Trail) , land type (Park, Community, Forestry), # of sampling occasions -  [,2:4]

chimp(or gorilla) detection (y) - [,5:56]

obs covs  - standardized Julian date [,57:108], ecological season (Rainy, Short dry, Short rainy, Dry) [,109:160], month (June2019-May2020) [ ,161:212]

José, you wrote :   "If you don't need to estimate dynamic parameters but you'd like to test for temporal variation in occupancy you could apply a single-season model with random effects (being random effects the camera trap or sampling unit), by stacking months (i.e., your sampling units would be camera-months)."

I believe this is what I want to do, but I'm not sure how to set up my data/unmarked frame.

Instead of my first column being 19 rows of sites, would it be 19*12 rows (i.e  stacked site.month; site1.june, site1.july...site19.may)? Could I then make the ecological season a site cov instead? I'm also missing some data (not all sites have functional cameras for all 12 months). I could stack the site.season, however the ecological seasons (Rainy, Dry, etc) are unequal in length, so it might be problematic.

This is what I have so far for my chimpanzee model (obviously if I switch to the stacked version this would be different): 

chimpanzee <- read.csv("chimp.detection.4.csv")

DATE <- as.matrix(chimpanzee[,57:108])    ## Julian date of sampling occasion

y.chimp <- as.matrix(chimpanzee[,5:56])  ##detection matrix

## standardize dates

y.chimp[is.na(DATE) != is.na(y.chimp)] <- NA

sd.DATE <- sd(c(DATE), na.rm = TRUE)
mean.DATE <- mean(DATE, na.rm = TRUE)
DATE <- (DATE - mean.DATE) / sd.DATE

# set up site covariates (habitat type, land type, number of sampling occasions)

siteCovs <- chimpanzee[ ,2:4]

# set up obs covariates (season, month)

SEASON <- chimpanzee[,109:160]

MONTH <- chimpanzee[,161:212]

obsCovs <- list(date=DATE, season=SEASON, month= MONTH)

## Now I just need to set up either an unmarkedMultFrame or unmarkedFrameOccu...

Any input would be helpful! I already ran a single-season model which worked fairly well (I think?), but I'm annoyed because I think I must be violating the closure assumption, as I know occupancy varies between the sites each month. 

Thank you!

Emily

[Jose Infante]

Hi Emily,

Yes, 19*12 sites in your case is what I was thinking with the stacking strategy (on the assumption that a month is enough to stabilize the estimates of occupancy and detection while not long enough to violate closure). And yes, variables that change in time could then be used as site covs. 

Best,

José.

José Domingo Infante Varela Ing. Agrónomo, MSc en Recursos Naturales, Estudiante de Doctorado en Ecosistemas Forestales y Recursos Naturales UACh   Investigador Asociado | Instituto de Conservación, Biodiversidad y Territorio | Laboratorio Fauna Australis UC    mobile: +56 9 88275395 email: join...@uc.cl

To view this discussion on the web visit https://groups.google.com/d/msgid/unmarked/cbe55272-36fe-45ba-93a8-5e4cafc74d36n%40googlegroups.com.

[Emily Collins]

Thank you!