Modeli s prostorsko omejitvijo za ocenjevanje gostote superpopulacije

Luštrik, Roman (author), Skrbinšek, Tomaž (mentor)

Eden od načinov ocenjevanja velikosti populacij je z uporabo metod lova-ponovnega ulova. Metoda predpostavlja, da je populacija zaprta (ni rojstev, smrti, priseljevanja in odseljevanja) in da imajo vsi osebki enako verjetnost ulovljivosti. Ker populacij pogosto ne moremo vzorčiti v celoti, zaradi prehoda roba območja vzorčenja prihaja do kršenja teh dveh predpostavk, kar imenujemo učinek roba. Klasični Hugginsov model za zaprte populacije za oceno parametrov sam po sebi ne omogoča uporabe prostorskih statistik, omogoča pa vključevanje individualne spremenljivke. V tem delu s pomočjo simulacij testiramo učinkovitost vključevanja individualne spremenljivke v model z namenom odpravljanja posledic učinka roba. Ugotovili smo, da je model, ki vključuje prostorsko informacijo, boljši od modela, ki te informacije ne nosi. Razlika v ocenjenem parametru verjetnosti ulovljivosti ($\hat{p}$) je s praktičnega vidika zelo majhna. Pristranskost ocene parametra $\hat{p}$ je najmanjša za tiste simulacije, kjer je velikost domačega okoliša znatno manjša od velikosti območja vzorčenja, za ostale pa je močno pristranska. Pristranskost ocene parametra $\hat{p}$ se pozna tudi pri oceni gostote, ki je zelo pristranska za primere, kjer je domač okoliš velik v primerjavi z velikostjo območja vzorčenja. Na podlagi porazdelitve za izračun individualne spremenljivke smo povečali območje vzorčenja in uspeli do neke mere popraviti gostoto, a le ob predpostavki, da imamo na voljo reprezentativno obliko in velikost domačega okoliša.

• Language: Slovenian
• Keywords: populacije, gostota, ocena velikosti populacije, učinek roba, metoda lova-ponovnega ulova, program MARK, program R
• Work type: Doctoral dissertation
• Typology: 2.08 - Doctoral Dissertation
• Organization: BF - Biotechnical Faculty
• Year: 2019
• PID: 20.500.12556/RUL-111678
• COBISS.SI-ID: 934263
• Publication date in RUL: 08.10.2019

LUŠTRIK, Roman, 2019, Modeli s prostorsko omejitvijo za ocenjevanje gostote superpopulacije [online]. Doctoral dissertation. [Accessed 10 June 2025]. Retrieved from: https://repozitorij.uni-lj.si/IzpisGradiva.php?lang=eng&id=111678

Secondary language

• Language: English
• Title: Spatially explicit modeling of superpopulation density

Abstract

Among methods for estimating population sizes, mark-recapture is a popular choice. It assumes population closure (void of deaths, births, immigration and emigration) and equal probability of capture. Since populations often cannot be encompassed entirely, some individuals cross in and out of the sampling area in violation of aforementioned assumptions, which is termed edge effect. The time-tested Huggins model does not in itself use spatial information to estimate parameters; however, it does enable use of an individual covariate. In this thesis, we use simulations to test whether including spatial information through an individual covariate helps alleviate edge effect. Our findings suggest that including spatial information does improve the model. For practical purposes, the difference in estimates of probability of capture ($\hat{p}$) between models is negligible. Bias of $\hat{p}$ is smallest in cases where home range size is small relative to sampling area size and large for cases where home range is comparatively large. This is also evident in density estimates, which are highly biased in cases where home range is relatively large compared to sampling area. We increased the sampling area radius based on distributions used to calculate the individual covariate and managed to somewhat alleviate the bias, provided that the calculated home range shape and size are representative.

• Keywords: populations, density, population size estimate, edge effect, mark-recapture, program MARK, program R