Species interaction models. Goal Determine whether a site is occupied by two different species and if they affect each others' detection and occupancy.

Slides:



Advertisements
Similar presentations
Patch Dynamics AKA: Multi-season Occupancy, Robust Design Occupancy.
Advertisements

EMNLP, June 2001Ted Pedersen - EM Panel1 A Gentle Introduction to the EM Algorithm Ted Pedersen Department of Computer Science University of Minnesota.
Null models in Ecology Diane Srivastava Sept 2010.
Krishna Pacifici Department of Applied Ecology NCSU January 10, 2014.
Patch Occupancy and Patch Dynamics Single species, Single Season Occupancy.
Rachel Fewster Department of Statistics, University of Auckland Variance estimation for systematic designs in spatial surveys.
Metapopulations Objectives –Determine how e and c parameters influence metapopulation dynamics –Determine how the number of patches in a system affects.
Measuring biotic components of a system
Maximum likelihood estimates What are they and why do we care? Relationship to AIC and other model selection criteria.
Modeling Wim Buysse RUFORUM 1 December 2006 Research Methods Group.
Additional Topics in Regression Analysis
Final Review Session.
Log-linear and logistic models
PSY 307 – Statistics for the Behavioral Sciences Chapter 19 – Chi-Square Test for Qualitative Data Chapter 21 – Deciding Which Test to Use.
Linear statistical models 2009 Count data  Contingency tables and log-linear models  Poisson regression.
Conservation Design for Sustainable Avian Populations
Single and Multiple Spell Discrete Time Hazards Models with Parametric and Non-Parametric Corrections for Unobserved Heterogeneity David K. Guilkey.
Modelling non-independent random effects in multilevel models William Browne Harvey Goldstein University of Bristol.
Bioscience – Aarhus University Pin-point plant cover data Christian Damgaard Bioscience Aarhus University.
BIOE 293 Quantitative ecology seminar Marm Kilpatrick Steve Munch Spring Quarter 2015.
Detecting trends in dragonfly data - Difficulties and opportunities - Arco van Strien Statistics Netherlands (CBS) Introduction.
Educational Research: Competencies for Analysis and Application, 9 th edition. Gay, Mills, & Airasian © 2009 Pearson Education, Inc. All rights reserved.
Resource Selection Functions and Patch Occupancy Models: Similarities and Differences Lyman L. McDonald Senior Biometrician Western EcoSystems Technology,
Bioscience – Aarhus University Measuring plant abundance and ecological processes with the pin-point method Christian Damgaard Department of Bioscience.
Probability Definitions Dr. Dan Gilbert Associate Professor Tennessee Wesleyan College.
Danila Filipponi Simonetta Cozzi ISTAT, Italy Outlier Identification Procedures for Contingency Tables in Longitudinal Data Roma,8-11 July 2008.
Blocks and pseudoreplication
Patch Occupancy: The Problem
Bivariate Poisson regression models for automobile insurance pricing Lluís Bermúdez i Morata Universitat de Barcelona IME 2007 Piraeus, July.
Global Analyzing community data with joint species distribution models abundance, traits, phylogeny, co-occurrence and spatio-temporal structures Otso.
Workshop on Applied Hierarchical Modeling in BUGS and unmarked Patuxent Wildlife Research Center November 2015.
Statistics 2: generalized linear models. General linear model: Y ~ a + b 1 * x 1 + … + b n * x n + ε There are many cases when general linear models are.
Introduction to Occupancy Models Key to in-class exercise are in blue
Multiple Season Model Part I. 2 Outline  Data structure  Implicit dynamics  Explicit dynamics  Ecological and conservation applications.
Estimation of Animal Abundance and Density Miscellaneous Observation- Based Estimation Methods 5.2.
Multistate models UF Outline  Description of the model  Data structure and types of analyses  Multistate with 2 and 3 states  Assumptions 
Statistical Methods. 2 Concepts and Notations Sample unit – the basic landscape unit at which we wish to establish the presence/absence of the species.
Capture-recapture Models for Open Populations “Single-age Models” 6.13 UF-2015.
 1 Species Richness 5.19 UF Community-level Studies Many community-level studies collect occupancy-type data (species lists). Imperfect detection.
Identify techniques for estimating various populations (quadrats, transects, mark- recapture) Understand the carrying capacity of ecosystems; factors.
Multiple Detection Methods: Single-season Models.
Spatially Explicit Capture-recapture Models for Density Estimation 5.11 UF-2015.
1 STAT 500 – Statistics for Managers STAT 500 Statistics for Managers.
 Integrated Modelling of Habitat and Species Occurrence Dynamics.
K-Sample Closed Capture-recapture Models UF 2015.
Monitoring and Estimating Species Richness Paul F. Doherty, Jr. Fishery and Wildlife Biology Department Colorado State University Fort Collins, CO.
Single Season Model Part I. 2 Basic Field Situation From a population of S sampling units, s are selected and surveyed for the species. Units are closed.
Single Season Occupancy Modeling 5.13 UF Occupancy Modeling State variable is proportion of patches that is occupied by a species of interest.
1 Occupancy models extension: Species Co-occurrence.
 1 Modelling Occurrence of Multiple Species. 2 Motivation Often there may be a desire to model multiple species simultaneously.  Sparse data.  Compare/contrast.
 Multi-state Occupancy. Multiple Occupancy States Rather than just presence/absence of the species at a sampling unit, ‘occupancy’ could be categorized.
Multiple Season Study Design. 2 Recap All of the issues discussed with respect to single season designs are still pertinent.  why, what and how  how.
Single Season Study Design. 2 Points for consideration Don’t forget; why, what and how. A well designed study will:  highlight gaps in current knowledge.
 Occupancy Model Extensions. Number of Patches or Sample Units Unknown, Single Season So far have assumed the number of sampling units in the population.
Multi-state Occupancy. Multiple Occupancy States Rather than just presence/absence of the species at a sampling unit, ‘occupancy’ could be categorized.
Combining pin-point and Braun-Blanquet plant cover data
Hierarchical models. Hierarchical with respect to Response being modeled – Outliers – Zeros Parameters in the model – Trends (Us) – Interactions (Bs)
Hierarchical Models. Conceptual: What are we talking about? – What makes a statistical model hierarchical? – How does that fit into population analysis?
Occupancy Models when misclassification occurs. Detection Errors and Occupancy Estimation  Occupancy estimation accounts for issues of detection when.
Chloe Boynton & Kristen Walters February 22, 2017
Integrated population models: an introduction to structure and behavior Tessa L. Behnke & Thomas V. Riecke J Recruitment Poisson model Population data.
Quadrat Sampling Chi-squared Test
Discrete Probability Distributions
Patch Occupancy and Patch Dynamics
Extension to the Hybrid Symbolic-Numeric Method for Investigating Identifiability Diana Cole, University of Kent, UK Rémi Choquet, CEFE, CNRS, France.
Multistate models Lecture 10.
Estimating mean abundance from repeated presence-absence surveys
Wildlife Population Analysis
Wildlife Population Analysis
A protocol for data exploration to avoid common statistical problems
Presentation transcript:

Species interaction models

Goal Determine whether a site is occupied by two different species and if they affect each others' detection and occupancy probabilities. Examples Predator-prey interactions Competitive exclusion Compares: Expected rates of occupancy to occupancy when another species is present Expected rates of detection to detection when another species is present

Saturated model Model that perfectly fits the data. Deviance = -2*  ln(x i ) x i - proportion times of each history is observed “standard” upon which all of our co-occurrence occupancy models will be judged

Similarities to single season occupancy Relates encounter histories and detection probabilities to a site. Occupancy is assumed closed during sampling period Site is sampled multiple times Encounter history is obtained for both species Based on repeated sampling Spatial or temporal replication

Parameters of interest – ugh!  A – Probability of occupancy by species A (unconditional)  B – Probability of occupancy by B (unconditional)  AB – Probability of occupancy by A & B (co-occurrence) p A – Probability of detecting species A when only A is present p B – Probability of detecting species B when only B is present r AB – Probability of detecting species A & B when both are present r Ab – Probability of detecting species only A when both present r Ba – Probability of detecting species only B when both present r ab – Probability of detecting NEITHER when both present = 1 – r AB – r Ab – r Ba

Many parameters = much data required!

Occupancy – Venn diagram    AB  A  B +  AB

Occupancy parameters – 4 states  A – Probability of occupancy by species A (unconditional)  B – Probability of occupancy by B (unconditional)  AB – Probability of occupancy by A & B (co-occurrence) Could estimate  AB =  A  B if no interaction Interaction estimated by:  =  AB /(  A  B )  avoidance (less frequent than expected)  convergence (more frequent than expected) 4 th State – absence of both species – 1-  A -  B +  AB

Detection parameters Given both species are present 4 possibilities: Detecting species A only – r bA Detecting species B only r Ba r AB – Probability of detecting species A & B r ab – Probability of detecting NEITHER species 1 - (r Ab - r aB – r AB )

Probability of encounter histories Pr(11 11) =  AB *r AB 1 *r AB 2 Pr(11 00) =  AB *r Ab 1 *r Ab 2 +(  A -  AB )*p A 1 *p A 2 Pr(00 00) =  AB *p ab 1 *r ab 2 +(  A -  AB )*(1-p A 1 )*(1-p A 2 ) +(  B -  AB )*(1-p B 1 )*(1-p B 2 ) +(1-  A -  B +  AB ) Uggh!

Estimation & modeling Estimate parameters (MLEs) via ln( L ) Introduce covariates via link functions All parameters constrained between 0 and 1 Usually use the logit link

Model selection Usually use QAICc Model fit via  2 – not the best but it will do c-hat ≈  2 /df (df = degrees of freedom) biased high Could use parametric bootstrap, but not readily available Sample size – number of sites surveyed

Model parameterizations Phi/delta parameterization PsiA = Pr(occ by A) PsiB = Pr(occ by B) PsiAB = Pr(occ by A and B) phi = PsiAB/(psiA*psiB) to make psiA and psiB independent FIX phi to 1 and delete column from DM

Model parameterizations PsiBa/rBa parameterization PsiA = Pr(occ by A) PsiBA = Pr(occ by B, given occ by A) PsiBa = Pr(occ by B, given NOT occ by A) to make psiA and psiB independent set psiBA equal to psiB in DM

Model parameterizations nu/rho parameterization PsiA = Pr(occ by A) PsiBa = Pr(occ by B, given NOT occ by A) nu = log-odds of how occupancy of B changes with presence of A To make psiA and psiB fix nu = 1 and delete column in DM

Additional Occupancy models (most in Presence)

Single-season mixture models ( Mackenzie et al. Ch 5.1) Use to estimate occupancy and detection rates Same repeated presents/absence survey approach Attempt to estimate unobservable heterogeneity Covariates are observable sources Discrete mixture: Finite (small) number of sites with similar occupancy and/or detection rates Continuous mixture All sites have different occupancy and/or detection but they come from some estimable distribution Very data hungry!

Royle-Nichols abundance induced heterogeneity Royle, J.A. and J.D. Nichols Ecology 84(3): Used to estimate abundance [density] from presence- absence data Main assumptions 1. Distribution of animals follows a prior [Poisson] distribution 2. Detection probability is a function of how many animals are present ( p = 1-(1- r ) N ( i ). No covariates!

Royle-N-Mixture Count (repeated count) Model Royle, J.A Biometrics 60, Estimates density from repeated counts Assumptions Spatial distribution prior distribution [Poisson distribution] Detection n animals at a site represents a binomial trial.

Single-season removal model Similar to single-season occupancy Estimates occupancy and detection Sites are no longer surveyed once species is detected More efficient – allows more sites. Assumptions: Detection constant across surveys (not p(t)) Allows covariates but no site interactions

Single-season multiple method Allows for different survey methods Example large-scale and small-scale sampling Assumption: if an individual is detected by one method, another is immediately available for detection by other method at that site. Similar to robust design approach

Species misidentification Royle, J. A., and W. Link Ecology 87: Extends occupancy analysis to allow for false positives Similar to mixture model Some portion observations are false positives

Species richness occupancy Royle et al Ecology 87: Estimate the number and composition of species. Uses presence-absence data For each species estimates: Probability of occupancy Probability of detection For all species Mean probability of occupancy and detection Expected species richness Number of species ‘missed’ Assumptions Closed to changes in population size Number of species is Poisson process

Multi-state occupancy Occupied sites are classified into multiple states Estimates: Occupancy, detection and probability of state Assumption Some state(s) can be identified with certainty Example: Breeding or non-breeding Occupied-breeding-probable breeding

Multi-season, multi-state occupancy Estimated parameters Estimates occupancy given suitable initially Probability that site is unsuitable in season Detection given occupied Extinction given suitable each season Extinction given change from suitable to unsuitable Colonization given change from unsuitable to suitable Colonization given that suitable each season Change from suitable to unsuitable Change from unsuitable to suitable Derived parameter Remains suitable

Occupancy with spatial correlation Hines et al. (in press) Estimates: Occupancy Detection Spatial autocorrelation biases occupancy estimates