Hierarchizing multi-scale environmental effects on agricultural pest population dynamics: A case study on the annual onset of Bactrocera dorsalis population growth in Senegalese orchards
Implementing integrated pest management programs to limit agricultural pest damage requires an understanding of the interactions between the environmental variability and population demographic processes. However, identifying key environmental drivers of spatio-temporal pest population dynamics remains challenging as numerous candidate factors can operate at a range of scales, from the field (e.g. agricultural practices) to the regional scale (e.g. weather variability). In such a context, data-driven approaches applied to pre-existing data may allow identifying patterns, correlations, and trends that may not be apparent through more restricted hypothesis-driven studies. The resulting insights can lead to the generation of novel hypotheses and inform future experimental work focusing on a limited and relevant set of environmental predictors. In this study, we developed an ecoinformatics approach to unravel the multi-scale environmental conditions that lead to the early re-infestation of mango orchards by a major pest in Senegal, the oriental fruit fly Bactrocera dorsalis (BD). We gathered abundance data from a three-year monitoring conducted in 69 mango orchards as well as environmental data (i.e. orchard management, landscape structure and weather variability) across a range of spatial scales. We then developed a flexible analysis pipeline centred on a recent machine learning algorithm, which allows the combination of gradient boosting and grouped random effects models or Gaussian processes, to hierarchize the effects of multi-scale environmental variables on the onset of annual BD population growth in orchards. We found that physical factors (humidity, temperature), and to some extent landscape variables, were the main drivers of the spatio-temporal variability of the onset of population growth in orchards. These results suggest that favourable microclimate conditions could provide refuges for small BD populations that could survive, with little or no reproduction, during the mango off-season and, then, recolonize neighbouring orchards at the beginning of the next mango season. Confirmation of such a hypothesis could help to prioritize surveillance and preventive control actions in refuge areas.
| Main Authors: | , , , , , , , , , |
|---|---|
| Format: | article biblioteca |
| Language: | eng |
| Published: |
Peer Community In
|
| Subjects: | H10 - Ravageurs des plantes, U10 - Informatique, mathématiques et statistiques, Bactrocera dorsalis, Mangifera indica, dynamique des populations, Bactrocera, Tephritidae, modélisation, gestion intégrée des ravageurs, facteur du milieu, http://aims.fao.org/aos/agrovoc/c_36022, http://aims.fao.org/aos/agrovoc/c_4575, http://aims.fao.org/aos/agrovoc/c_6111, http://aims.fao.org/aos/agrovoc/c_33628, http://aims.fao.org/aos/agrovoc/c_3122, http://aims.fao.org/aos/agrovoc/c_230ab86c, http://aims.fao.org/aos/agrovoc/c_34030, http://aims.fao.org/aos/agrovoc/c_2594, http://aims.fao.org/aos/agrovoc/c_6970, |
| Online Access: | http://agritrop.cirad.fr/610744/ http://agritrop.cirad.fr/610744/1/610744.pdf |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | Implementing integrated pest management programs to limit agricultural pest damage requires an understanding of the interactions between the environmental variability and population demographic processes. However, identifying key environmental drivers of spatio-temporal pest population dynamics remains challenging as numerous candidate factors can operate at a range of scales, from the field (e.g. agricultural practices) to the regional scale (e.g. weather variability). In such a context, data-driven approaches applied to pre-existing data may allow identifying patterns, correlations, and trends that may not be apparent through more restricted hypothesis-driven studies. The resulting insights can lead to the generation of novel hypotheses and inform future experimental work focusing on a limited and relevant set of environmental predictors. In this study, we developed an ecoinformatics approach to unravel the multi-scale environmental conditions that lead to the early re-infestation of mango orchards by a major pest in Senegal, the oriental fruit fly Bactrocera dorsalis (BD). We gathered abundance data from a three-year monitoring conducted in 69 mango orchards as well as environmental data (i.e. orchard management, landscape structure and weather variability) across a range of spatial scales. We then developed a flexible analysis pipeline centred on a recent machine learning algorithm, which allows the combination of gradient boosting and grouped random effects models or Gaussian processes, to hierarchize the effects of multi-scale environmental variables on the onset of annual BD population growth in orchards. We found that physical factors (humidity, temperature), and to some extent landscape variables, were the main drivers of the spatio-temporal variability of the onset of population growth in orchards. These results suggest that favourable microclimate conditions could provide refuges for small BD populations that could survive, with little or no reproduction, during the mango off-season and, then, recolonize neighbouring orchards at the beginning of the next mango season. Confirmation of such a hypothesis could help to prioritize surveillance and preventive control actions in refuge areas. |
|---|