Understanding population dynamics using the predictive ability of models
| dc.contributor.advisor | Houlahan, Jeff | |
| dc.contributor.author | Gebreyohannes, Demissew Tsigemelak | |
| dc.date.accessioned | 2024-07-24T15:00:32Z | |
| dc.date.available | 2024-07-24T15:00:32Z | |
| dc.date.issued | 2024-06 | |
| dc.description.abstract | The investigation into what determines population dynamics persists as a central and unresolved issue within the field of ecology. One of the key areas of discussion is the relative importance of density dependent versus density independent factors. The concept of density dependence implies that current abundance is determined by historical abundance. I developed four models – two density dependent (Gompertz and Logistic) and two density independent ('Mean' and 'Trend’) - to predict population size one year beyond the training set and used predictive performance on more than 16,000 populations from 14 datasets to compare the understanding captured by those models. If density dependent population regulation is common then we expect that the Logistic and Gompertz models will, on average, make better predictions than the ‘Mean’ or ‘Trend’ models. I concluded that there is a weak evidence of density dependent population regulation. I also conducted further assessment of the predictive ability of two density-dependent models with covariates (Gompertz and Logistic, incorporating environmental factors such as temperature, precipitation, salinity, pH, and interspecific competition index) and two density-independent models ('Mean' and 'Trend,' also considering covariates and interspecific competition). I concluded that there was limited evidence that incorporating interspecific competition and/or environmental covariates led to improved model predictive ability, corroborating that there is weak evidence of density dependent population regulation. I further interrogated the predictive abilities of cutting-edge models, ARIMA and EDM (Empirical dynamic modelling), on population data. Within the ARIMA models, the low-dimensional parametric ARIMA model - specifically, one treating the mean of the most recent observations as the prediction - yielded the most accurate predictions across the majority of datasets. In EDM, employing an approach that resembles the mean of the previous 2 – 6 years yielded the best predictive ability. Overall, I presented evidence indicating that widespread density-dependent population regulation is weak, particularly based on the performance of density-dependent models, and I also evaluated the predictive abilities of cutting-edge models, implying weak supporting evidence for density-dependent population regulation. | |
| dc.description.copyright | © Demissew Tsigemelak Gebreyohannes, 2024 | |
| dc.format.extent | xx, 353 | |
| dc.format.medium | electronic | |
| dc.identifier.uri | https://unbscholar.lib.unb.ca/handle/1882/38064 | |
| dc.language.iso | en | |
| dc.publisher | University of New Brunswick | |
| dc.relation | National Science and Engineering Research Council of Canada (NSERC) | |
| dc.rights | http://purl.org/coar/access_right/c_abf2 | |
| dc.subject.discipline | Biology | |
| dc.title | Understanding population dynamics using the predictive ability of models | |
| dc.type | doctoral thesis | |
| oaire.license.condition | other | |
| thesis.degree.discipline | Biology | |
| thesis.degree.grantor | University of New Brunswick | |
| thesis.degree.level | doctorate | |
| thesis.degree.name | Ph.D. |