EcoEnsemble: A general framework for combining ecosystem models in R
Abstract Often there are several complex ecosystem models available to address a specific question. However, structural differences, systematic discrepancies and uncertainties mean that they typically produce different outputs. Rather than selecting a single ‘best’ model, it is desirable to combine...
| Main Authors: | , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Wiley
2023-08-01
|
| Series: | Methods in Ecology and Evolution |
| Subjects: | |
| Online Access: | https://doi.org/10.1111/2041-210X.14148 |
| _version_ | 1797760644063166464 |
|---|---|
| author | Michael A. Spence James A. Martindale Michael J. Thomson |
| author_facet | Michael A. Spence James A. Martindale Michael J. Thomson |
| author_sort | Michael A. Spence |
| collection | DOAJ |
| description | Abstract Often there are several complex ecosystem models available to address a specific question. However, structural differences, systematic discrepancies and uncertainties mean that they typically produce different outputs. Rather than selecting a single ‘best’ model, it is desirable to combine them to give a coherent answer to the question at hand. Many methods of combining ecosystem models assume that one of the models is exactly correct, which is unlikely to be the case. Furthermore, models may not be fitted to the same data, have the same outputs, nor be run for the same time period, making many common methods difficult to implement. In this paper, we use a statistical model to describe the relationship between the ecosystem models, prior beliefs and observations to make coherent predictions of the true state of the ecosystem with robust quantification of uncertainty. We introduce EcoEnsemble, an R package that takes advantage of the statistical model's structure to efficiently fit the ensemble model, either sampling from the posterior distribution or maximising the posterior density. We demonstrate EcoEnsemble by investigating what would happen to four fish species in the North Sea under future management scenarios. Although developed for applications in ecology, EcoEnsemble can be used to combine any group of mechanistic models, for example in climate modelling, epidemiology or biology. |
| first_indexed | 2024-03-12T19:01:32Z |
| format | Article |
| id | doaj.art-a5b4b17532504067a8cda5e7c529c09c |
| institution | Directory Open Access Journal |
| issn | 2041-210X |
| language | English |
| last_indexed | 2024-03-12T19:01:32Z |
| publishDate | 2023-08-01 |
| publisher | Wiley |
| record_format | Article |
| series | Methods in Ecology and Evolution |
| spelling | doaj.art-a5b4b17532504067a8cda5e7c529c09c2023-08-02T06:34:44ZengWileyMethods in Ecology and Evolution2041-210X2023-08-011482011201810.1111/2041-210X.14148EcoEnsemble: A general framework for combining ecosystem models in RMichael A. Spence0James A. Martindale1Michael J. Thomson2Centre for Environment, Fisheries and Aquaculture Science Lowestoft UKCentre for Environment, Fisheries and Aquaculture Science Lowestoft UKCentre for Environment, Fisheries and Aquaculture Science Lowestoft UKAbstract Often there are several complex ecosystem models available to address a specific question. However, structural differences, systematic discrepancies and uncertainties mean that they typically produce different outputs. Rather than selecting a single ‘best’ model, it is desirable to combine them to give a coherent answer to the question at hand. Many methods of combining ecosystem models assume that one of the models is exactly correct, which is unlikely to be the case. Furthermore, models may not be fitted to the same data, have the same outputs, nor be run for the same time period, making many common methods difficult to implement. In this paper, we use a statistical model to describe the relationship between the ecosystem models, prior beliefs and observations to make coherent predictions of the true state of the ecosystem with robust quantification of uncertainty. We introduce EcoEnsemble, an R package that takes advantage of the statistical model's structure to efficiently fit the ensemble model, either sampling from the posterior distribution or maximising the posterior density. We demonstrate EcoEnsemble by investigating what would happen to four fish species in the North Sea under future management scenarios. Although developed for applications in ecology, EcoEnsemble can be used to combine any group of mechanistic models, for example in climate modelling, epidemiology or biology.https://doi.org/10.1111/2041-210X.14148Bayesian statisticsecosystem managementecosystem modelsensemble modellingRuncertainty analysis |
| spellingShingle | Michael A. Spence James A. Martindale Michael J. Thomson EcoEnsemble: A general framework for combining ecosystem models in R Methods in Ecology and Evolution Bayesian statistics ecosystem management ecosystem models ensemble modelling R uncertainty analysis |
| title | EcoEnsemble: A general framework for combining ecosystem models in R |
| title_full | EcoEnsemble: A general framework for combining ecosystem models in R |
| title_fullStr | EcoEnsemble: A general framework for combining ecosystem models in R |
| title_full_unstemmed | EcoEnsemble: A general framework for combining ecosystem models in R |
| title_short | EcoEnsemble: A general framework for combining ecosystem models in R |
| title_sort | ecoensemble a general framework for combining ecosystem models in r |
| topic | Bayesian statistics ecosystem management ecosystem models ensemble modelling R uncertainty analysis |
| url | https://doi.org/10.1111/2041-210X.14148 |
| work_keys_str_mv | AT michaelaspence ecoensembleageneralframeworkforcombiningecosystemmodelsinr AT jamesamartindale ecoensembleageneralframeworkforcombiningecosystemmodelsinr AT michaeljthomson ecoensembleageneralframeworkforcombiningecosystemmodelsinr |