Climate-Sensitive Vector Dynamics
Modelling Workshop

An introduction to the Copernicus Data Store for disease modeling applications

This lecture will provide an overview of environmental datasets available on the Copernicus data Store and other tools for disease modeling applications. Practical information will be provided about finding and downloading datasets using the related API; and manipulating these environmental datacubes. Other user friendly climate data tools, such as the WMO-KNMI climate explorer and the IPCC Atlas will also be presented.

arbocartoR package and app: deterministic ODE model for Aedes dynamics and control

This course introduces arbocarto, a mechanistic model of Aedes mosquito population dynamics based on a deterministic system of ordinary differential equations (ODEs). The model is spatialized and life cycle-based. It was developed to support operational objectives in surveillance and vector control. Its outputs include spatial maps of relative mosquito densities and estimates of the basic reproduction number (R_o) for arbovirus transmission. To make the model accessible to both research and operational communities, the arbocartoR R package and its companion Shiny app were developed within the H2020 MOOD project at Cirad. These tools allow users to simulate mosquito dynamics for Aedes aegypti and Aedes albopictus, evaluate the effects of vector control actions, and explore the transmission potential of dengue, Zika, and chikungunya across diverse environments.

The course will first present the ODE-based model structure and its operational applications, then demonstrate the features of the Shiny app, and finally guide participants through simple simulations in R using arbocartoR. Participants will gain hands-on experience running both the app and R scripts, adapting the model, visualizing outputs, and interpreting results for decision support in vector surveillance and control.

Physics-Informed Neural Networks and ODEs in vector population dynamics modelling

Mechanistic dynamic models based on ordinary differential equations (ODEs) depend on accurately parameterised, species-specific development and survival rates to generate reliable predictions. Physic-Informed Neural Networks (PINNs) address this by embedding ODE constraints directly into a neural framework trained on observational data, supporting both forward simulation and inverse parameter recovery. We employ PINNs to refine biological parameterisation in vector population dynamics models and directly compare their performance against traditional ODE-based approaches. Our results demonstrate that PINN-driven inverse modelling can in specific scenarios outperform classical methods in parameter accuracy and predictive skill. To unpack the contributions of network design, we perform controlled ablation studies, varying one architectural element at a time while holding all else constant, and assess each modification's effect on overall performance. Hybrid approaches that fuse mechanistic principles with data-driven methods can harness observational datasets to recover hard-to-measure parameters without sacrificing the physical fidelity of the governing biometeorological processes.

Epidemiological data harmonisation: Global Repository of Epidemiological Parameters

The Global Repository of Epidemiological Parameters (grEPI) is a World Health Organization initiative designed to accelerate and harmonise access to key epidemiological parameters to s upport modelling and analysis. This presentation will provide an overview of grEPI and the planned inclusions for version 1, including current approaches to AI supported parameter extraction. The proposed approach for incorporating vector-based parameters (with a focus on Aedes aegypti literature) will be demonstrated for consideration by participants. Developed through consultation with the global modelling community, grEPI aims to improve contestability of model assumptions, support more timely forecasts, and enhance policy relevance for ministries of health and research institutions. This presentation will highlight grEPI's contribution to harmonising epidemiological parameters for vector-borne disease intelligence to strengthen preparedness and response capacity.

Monitoring the population of adult Aedes aegypti in the city of Larnaca, Cyprus: Preliminary results

Aedes aegypti, a mosquito species originally native to Africa, is the world's most important vector of arboviral disease, being able to transmit the causative agents of dengue, yellow fever, chikungunya, and Zika. The recent re-establishment of Ae. aegypti in Cyprus (reported in 2022), after more than 50 years of absence, poses a major national and regional public health threat, particularly in light of this vector's high competence for arbovirus transmission, its synanthropic biology, and the status of Cyprus as a major destination for global tourism.

To better understand the population dynamics of this species in the particular environmental context of Cyprus, we set up eight trapping stations in the city of Larnaca, covering an area of approximately 17 hectares. At each station, once a week we set up a BG-Sentinel trap baited with BG-Lure and dry ice. Traps were collected after 24 hours, and captured specimens were killed on-site by exposure to acetone fumes. Dead specimens were transported to the Cyprus Institute, where they were sorted by species and sex, counted, and stored for future reference.

Here, we present a preliminary summary of the data collected between February 14 and November 5, 2025. During this period, a total of 3,757 mosquitoes were caught. Of these, 3,622 (96.4%) were Culex pipiens, 102 (2.7%) were Ae. aegypti, and 33 (0.9%) were other species. Our results show that Ae. aegypti is present throughout the entire collection area, with all trapping stations reporting at least one specimen collected during the sampling period. Additionally, our data suggests populations of Ae. aegypti peak during the summer months (July - September), while Cx. pipiens populations peak earlier in the year (April - June).

These results provide some insight into the temporal fluctuations of Ae. aegypti population currently established in Cyprus, thereby providing important information for vector control efforts. Furthermore, the quantitative data generated by our work can be used to develop and/or improve predictive models, enhancing the accuracy of their outputs at the local and regional levels.

Aedes aegypti in Egypt, a newly re-emerged and established arboviral vector: History, current status and plan for risk management and control

The presentation will summarize the history of Aedes aegypti and dengue outbreaks in Egypt, elimination as a by-product of malaria control campaigns, recent re-emergence and dengue outbreaks since 2014, geographic expansion and ecology, and the needs for surveillance and modelling the risk for containment and control.

From climate sensitivity to interventions: Work at Heidelberg Planetary Health Hub on Aedes aegypti and Dengue

The Heidelberg Planetary Health Hub (Hei-Planet) is an interdisciplinary research community addressing the interconnected challenges of climate change, environmental degradation, and biodiversity loss through a health-focused lens. A central research theme of Hei-Planet is the study of climate-sensitive infectious diseases, with a particular focus on mosquito-borne illnesses.

In this talk, I will provide a broad overview of our research on Aedes aegypti and Dengue. Our work spans multiple approaches, including thermal biology trait-based models and the development of statistical frameworks for robust quantification of uncertainty in trait performance curves. Building on these estimates, I will present our latest findings from the Lancet Countdown on climate change and health, tracking changes in global transmission suitability for Dengue.

Furthermore, I will discuss data-driven analyses of the effect of weather and large-scale public health interventions on Aedes mosquitoes and Dengue in Sri Lanka. In addition, I will present work on the use of climate teleconnections for early-warning predictions of Dengue outbreaks and ongoing efforts to advance novel modeling techniques that integrate mechanistic models with Neural Network components.

Spatio-temporal validation of two mechanistic models of Aedes aegypti population dynamics in ten Argentine localities

Mathematical modeling of Aedes aegypti population dynamics provides information for vector surveillance and control. We compared the predictive performance of two mechanistic models — Aguirre et al. (deterministic) and DynamAedes (stochastic) — against ovitrap data from 10 Argentine localities spanning a wide climatic gradient, between 2015 and 2024. Models' weekly predictions were standardized and compared with field observations, evaluating three aspects: (1) spatio-temporal performance, (2) peak detection, and (3) seasonal pattern (onset, end, and duration). DynamAedes resulted in extinction events in the three southernmost localities, while Aguirre et al.'s model showed more stable week-to-week behavior. Prediction fit varied across localities, with the best performance in northern localities like Tartagal (warm and humid) and Añatuya (warm and dry). In terms of peak detection, DynamAedes had higher sensitivity (43.5% vs. 27.2%, p<0.001), although both models showed F1-scores below 0.5 across all localities. Regarding the seasonal pattern, Aguirre et al.'s model reproduced the observed duration, while DynamAedes tended to anticipate the start and prolong the duration of the season. Results show that both models reproduced abundance patterns with varying accuracy, each with strengths and limitations depending on the evaluation criteria. This comprehensive comparison, covering nearly ten years of data across a broad geographic and climatic range, highlights the importance of thorough validations under diverse conditions to ensure more robust applications and a deeper understanding of model performance in vector surveillance and early warning contexts.

A stage-structured DDE model for Aedes aegypti and dengue

In the last 50 years there has been a rapid expansion in the range in Aedes aegypti, increasing the risk of Dengue fever globally. To accurately predict the start date of dengue outbreaks at these range edges, a full life-cycle model of Ae. aegypti that incorporates the environmentally-dependent development times, mortality rates and fecundity rates is needed. Here, we use a fully tractable system of delay differential equations to model Ae. aegypti that allows for developmental plasticity between stages coupled with a compartmental SEIR model for dengue fever. We demonstrate how this approach can accurately predict the seasonality in abundance of each developmental stage of Ae. aegypti across sub-tropical climates in the Americas, without the need for backfitting, and present results for the prediction of dengue fever outbreaks.

PesTwin, the biology-informed Digital-Twin transforming Integrated Pest Management

Reducing the damage caused by invasive insect species is a crucial lever both in agriculture, to improve food safety, and in public health, to combat vector-borne diseases. In support of these challenges, and in line with the principles of precision agriculture and Integrated Pest Management (IPM), we present PesTwin, an innovative simulation framework aiming to become the digital twin of a pest invasion. Through a flexible rule-based approach of the Agent Based Modelling and Simulation (ABMS) paradigm, the framework supports the fine-tuned modelling of the main ecological interactions of the pest with its host and the environment. By integrating GIS data, such as historical temperature time-series, PesTwin allows for forecasting the insects in realistic scenarios, both in its spatial and time dimensions. Further, with a module devoted to genetic inheritance, it enables the simulation of evolutionary dynamics, with an explicit focus on innovative genetic control techniques. PesTwin simulator lays at the intersection of ecology, computer science and applied mathematics, paving the way towards a more targeted and effective pest management, through optimization and innovation of control strategies.