Inferring pest dispersal in agricultural landscapes to improve management strategies – DISLAND

For two consecutive years, we conduct intensive monitoring of B. dorsalis abundance in two mango production basins (> 2000km2 per basin) located ~ 200km apart; in the Niayes, near Dakar, and in Basse-Casamance in the southwestern part of Senegal. This design allows us to consider both short- and long-distance dispersal scales and to study agroecological landscapes with contrasting characteristics. In addition to monitoring abundance, several thousand individual samples of B. dorsalis are collected and typed by high-throughput sequencing using tens of thousands of genetic markers. Our sampling and genotyping strategies should allow us to accurately describe the spatio-temporal variation in genetic diversity and gene flow within and between these two major mango production basins in Senegal. This level of spatio-temporal resolution is currently lacking in the study of insect pests.

We collect environmental data relevant to the ecology and management of pest populations at different spatial scales: 1) the orchard, by determining host plant composition, conducting phenological monitoring of host plants, estimating fruit damage, and interviewing growers about agricultural practices and crop losses; 2) the production basin, by using remote sensing and deep learning approaches to classify landscape features in terms of suitability for the fruit fly, and 3) the region, by monitoring commercial roads used to transport potentially contaminated mangoes from previous production basins, and by using climate models or reconstructing air mass trajectories that can carry flies over long distances.

We develop different methodologies for analyzing our extensive spatio-temporal data, including a method centered on a machine learning algorithm to prioritize the effects of multiple environmental factors on pest demography, as well as two spatialized methods for estimating key demographic parameters for pest management (dispersal and population size), one based on distance isolation models from population genetics, the other on diffusion models from population dynamics and epidemiology. All the knowledge gained on pest population dynamics will be integrated into a spatialized simulation model that also represents the actions of of the actors of the socio-technical system.

In the two monitored basins, three pheromone traps were installed in 28 sentinel orchards in February 2022. Traps have been checked on a weekly basis from March to August, then every three weeks from September to February (3,000 counts have already been recorded). Between February 2021 and June 2023, mangoes were sampled from trucks supplying the two main markets of the Niayes area then incubated to check for the presence of B. dorsalis. At the same time, traps were used to directly collect flies from these markets and from different production basins in Western Africa. Genetic data are currently being generated: 1) ~4000 individuals representative of the 56 orchards over 6 key dates in the first year of monitoring, 2) ~2000 individuals from imported fruits or sampled from the main markets, and 3) 31 populations collected in 10 Western African countries. To do this, we have developed a panel of myBaits® nucleic probes (Daicel Arbor Biosciences) for hybridization capture and sequencing experiments on 20,000 genomic targets.

In addition, using deep learning and numerous in-field control points, we are finalizing the production of a new landscape classification method adapted to B. dorsalis and covering all of Senegal. A phenological calendar for the main host plants is currently being constructed, thanks to the monitoring carried out since early 2022. In each orchard, cohorts of mangoes of different varieties have been inspected on a weekly basis until harvest, in order to assess yield losses due to fly stings. Such monitoring is also carried out on alternative hosts such as cashews and citrus. A spatio-temporal, multi-source relational database has been built under the open management system PostgreSQL/PostGis to manage all the data collected or produced during the project.

Various analysis methods are currently being developed: an automatic processing chain for prioritizing the environmental factors impacting pest spatiotempral dynamics; a first reaction-diffusion model modeling the dynamics of B. dorsalis in interaction with the environmental matrix at the scale of a territory; the implementation of temporal changes in the methodology of Virgoulay et al. (2021) for estimating dispersal parameters from genetic data. In parallel, we have also established a strong collaboration with stakeholders, particularly mango growers in the Niayes and Basse-Casamance. A role-playing game involving a group of growers and buyers was developed using a multi-agent simulation model that quantifies the impact of orchard management on pest dynamics and production over the course of a production season. The role-playing game was used in the Senegalese field to test and evaluate its training and co-design potential.

The application of the different methods under development to the extensive collected data (demographic, genetic and environmental) will provide information on the demographic functioning of oriental fruit fly populations in Senegal and on the environmental effects. This new knowledge will make it possible to consider fly management at a scale consistent with the ecological processes at work in the areas concerned, i.e., no longer based on individual decisions at the scale of the cultivated field, but towards collective organization and coordination of actions. For example, it would be possible to target populations at the relevant scales, whether spatial (e.g. production basin) or temporal (e.g. dry season), relying also on monitoring to support decision-making based on intervention thresholds. Implementing such preventive measures would reduce both damage and pesticide use. In addition, new management techniques could be used, such as crop management (e.g. irrigation) and landscape manipulation (e.g. spatial arrangement of host crops). Even more, the construction of a role-playing game based on an operational multi-agent model opens up the prospect of its dissemination as a training, awareness-raising and development tool for concerted fruit fly management, but also as a tool for co-designing innovative fruit fly management strategies. To this end, some of the ecological knowledge generated will be integrated into the multi-agent model to represent the key interactions between pest population dynamics, landscape structure and the interventions of stakeholders.

Caumette C, Diatta P, Piry S, Faye E, Chapuis MP & Berthier K (2021) Facteurs de ré-infestation des vergers par la mouche des fruits Bactrocera dorsalis dans le bassin de production horticole des Niayes au Sénégal. Poster. 3ème Conférence sur l’intensification durable. Dakar, Sénégal, 23-26 novembre 2021 agritrop.cirad.fr/601317/. Prix du meilleur poster d’une des trois sessions.


N'gom A, Belmin R, Grechi I, Brévault T & Rebaudo F (2023) Course contre la mouche : jeu de rôle pour une gestion concertée contre la mouche orientale des fruits dans la filière mangue au Sénégal. Poster. Printemps de Baillarguet. Le Printemps de Baillarguet. Montpellier, France, 26-27 juin 2023.

Crop pests are a major constraint to ecological intensification of agricultural production systems. Integrated Pest Management (IPM) is an ecosystem approach that combines different strategies and practices to minimize the use of pesticides. Its implementation requires a sound knowledge of pest population dynamics and underlying ecological processes to assist decision-making. In particular, the integration of pest dispersal processes (e.g. active and passive, short and long-distance, human-assisted and windborne) that condition seasonal crop infestation, opens up opportunities for preventive interventions (e.g. targeting residual populations) at relevant spatial and functional scales (e.g. the production basin). Accounting for pest dispersal in relation to their environment and at multiple scales is pivotal to a fundamental shift in the contours of action, from conventional and often individual-based interventions at a cultivated field scale, to collective organisation and management at a territory scale. The design of such an IPM strategy, however, also requires the integration of stakeholders to identify the socio-technical lock-ins to innovations.
Since demographic studies are challenging to assess pest dispersal at spatial scales relevant for pest management, an alternative is to use an "indirect" approach based on neutral genetic data. In this context, DISLAND proposes to develop and extend a landscape genetics approach to address “dispersal” issues as a prerequisite to improve pest management strategies. Such an approach includes the assessment, in space and time, of the effect of landscape structure and agricultural practices on spatiotemporal pest abundance and active short-distance dispersal, as well as of the role of passive long-distance dispersal through wind and commodity trades on population dynamics.
To this aim, we will develop appropriate tools including a cost-effective high-throughput sequencing technique to produce thousands of highly informative genetic markers (SNPs) for thousands of samples, a dedicated database to ensure the storage, traceability and sharing of all data, and a flexible multi-agent simulator to model demo-genetic processes in realistic landscapes. We will implement an intensive demo-genetic monitoring of the pest and acquire a large set of data highly relevant to its ecology and management (landscape structure, habitat quality, relative abundance and agricultural practices) in contrasted production basins (and beyond, at the regional scale). Finally, we will adapt or develop appropriate statistical methods to characterize the relationships between dispersal and the environmental matrix and quantify the spatial and temporal variation of key demographic parameters for pest management (dispersal and population size).
This framework will be designed and applied on the Oriental fruit fly, Bactrocera dorsalis, which is an invasive key pest of mango and other fruits in West Africa. The new knowledge on fruit fly dispersal will then be integrated into a spatially-explicit simulation model representing pest population dynamics and interactions with stakeholders. This model will serve as a tool for participatory evaluation and conception of system-wide fruit fly management strategies. This innovative research will bridge agronomy, landscape ecology, population dynamics and population genetics, and socio-ecology.