Spatial transferability of an agent-based model to simulate Taenia solium control interventions.

BACKGROUND: Models can be used to study and predict the impact of interventions aimed at controlling the spread of infectious agents, such as Taenia solium, a zoonotic parasite whose larval stage causes epilepsy and economic loss in many rural areas of the developing nations. To enhance the credibility of model estimates, calibration against observed data is necessary. However, this process may lead to a paradoxical dependence of model parameters on location-specific data, thus limiting the model's geographic transferability. METHODS: In this study, we adopted a non-local model calibration approach to assess whether it can improve the spatial transferability of CystiAgent, our agent-based model of local-scale T. solium transmission. The calibration dataset for CystiAgent consisted of cross-sectional data on human taeniasis, pig cysticercosis and pig serology collected in eight villages in Northwest Peru. After calibration, the model was transferred to a second group of 21 destination villages in the same area without recalibrating its parameters. Model outputs were compared to pig serology data collected over a period of 2 years in the destination villages during a trial of T. solium control interventions, based on mass and spatially targeted human and pig treatments. RESULTS: Considering the uncertainties associated with empirical data, the model produced simulated pre-intervention pig seroprevalences that were successfully validated against data collected in 81% of destination villages. Furthermore, the model outputs were able to reproduce validated pig seroincidence values in 76% of destination villages when compared to the data obtained after the interventions. The results demonstrate that the CystiAgent model, when calibrated using a non-local approach, can be successfully transferred without requiring additional calibration. CONCLUSIONS: This feature allows the model to simulate both baseline pre-intervention transmission conditions and the outcomes of control interventions across villages that form geographically homogeneous regions, providing a basis for developing large-scale models representing T. solium transmission at a regional level.

Nowcasting and forecasting COVID-19 waves: the recursive and stochastic nature of transmission.

We propose a parsimonious, yet effective, susceptible-exposed-infected-removed-type model that incorporates the time change in the transmission and death rates. The model is calibrated by Tikhonov-type regularization from official reports from New York City (NYC), Chicago, the State of São Paulo, in Brazil and British Columbia, in Canada. To forecast, we propose different ways to extend the transmission parameter, considering its estimated values. The forecast accuracy is then evaluated using real data from the above referred places. All the techniques accurately provided forecast scenarios for periods 15 days long. One of the models effectively predicted the magnitude of the four waves of infections in NYC, including the one caused by the Omicron variant for periods of 45 days using out-of-sample data.

epidWaves: A code for fitting multi-wave epidemic models.

The COVID-19 pandemic has given rise to a great demand for computational models capable of describing and inferring the evolution of an epidemic outbreak in the short term. In this sense, we introduce epidWaves, a package that provides a framework for fitting multi-wave epidemic models to data from actual outbreaks of COVID-19 and other infectious diseases.

ARBO: Arbovirus modeling and uncertainty quantification toolbox.

The ongoing pandemic of COVID-19 has highlighted the importance of mathematical tools to understand and predict outbreaks of severe infectious diseases, including arboviruses such as Zika. To this end, we introduce ARBO, a package for simulation and analysis of arbovirus nonlinear dynamics. The implementation follows a minimalist style, and is intuitive and extensible to many settings of vector-borne disease outbreaks. This paper outlines the main tools that compose ARBO, discusses how recent research works about the Brazilian Zika outbreak have explored the package's capabilities, and describes its potential impact for future works on mathematical epidemiology.

Simulating the hydrological response of a small tropical forest watershed (Mata Atlantica, Brazil) by the AnnAGNPS model.

Given the intrinsic hydrological cycle made of large input of water vapour and intense precipitation producing large volumes of water and sediment, modelling runoff and water losses in humid tropical watersheds is important for forest and water resources management. For instance, reliable simulations of the water cycle in such environments are a prerequisite for predictions of water quality, soil erosion and the climate change effects on water resources. The distributed parameter, physically based, continuous simulation, daily time step AnnAGNPS model, was implemented in almost completely forested (98% of its area, 0.56 km2) Cunha watershed (Brazil) to assess its capability to simulate hydrological processes under tropical conditions. The simulated surface runoff was compared to 4-year observations with statistical indices on several time scales. The model, running with default CN of forest, showed poor predictions of runoff. After increasing CN from 63 to 72 by calibration, the runoff prediction capability of AnnAGNPS was satisfactory on annual, seasonal and monthly scales, while daily runoff predictions were less accurate. Modelling water losses at event scale showed that the effect of forest vegetation on water retention during a single precipitation was more limited than for longer periods (months, seasons and years), since evapo-transpiration and interception account for small shares (>20%) of total precipitation. This study demonstrated that the AnnAGNPS model has reliable runoff prediction capacity in tropical forest watersheds at the annual and seasonal scales (E > 0.73), whereas daily runoff simulations are less accurate (E = 0.44). The use of this model may prove an important tool for water resource and territory management in tropical rainforests.

Modelling land cover change in the Brazilian Amazon: temporal changes in drivers and calibration issues.

Land cover change (LCC) models are used in many studies of human impacts on the environment, but knowing how well these models predict observed changes in the landscape is a challenge. We used nearly three decades of LCC maps to run several LCC simulations to: (1) determine which parameters associated with drivers of LCC (e.g. roads) get selected for which transition (forest to deforested, regeneration to deforested or deforested to regeneration); (2) investigate how the parameter values vary through time with respect to the different activities (e.g. farming); and (3) quantify the influence of choosing a particular time period for model calibration and validation on the performance of LCC models. We found that deforestation of primary forests tends to occur along roads (included in 95 % of models) and outside protected areas (included in all models), reflecting farming establishment. Regeneration tends to occur far from roads (included in 78 % of the models) and inside protected areas (included in 38 % of the models), reflecting the processes of land abandonment. Our temporal analysis of model parameters revealed a degree of variation through time (e.g. effectiveness of protected areas rose by 73 %, p < 0.001), but for the majority of parameters there was no significant trend. The degree to which model predictions agreed with observed change was heavily dependent on the year used for calibration (p < 0.001). The next generation of LCC models may need to embed trends in parameter values to allow the processes determining LCC to change through time and exert their influence on model predictions.

On the predictive ability of mechanistic models for the Haitian cholera epidemic.

Predictive models of epidemic cholera need to resolve at suitable aggregation levels spatial data pertaining to local communities, epidemiological records, hydrologic drivers, waterways, patterns of human mobility and proxies of exposure rates. We address the above issue in a formal model comparison framework and provide a quantitative assessment of the explanatory and predictive abilities of various model settings with different spatial aggregation levels and coupling mechanisms. Reference is made to records of the recent Haiti cholera epidemics. Our intensive computations and objective model comparisons show that spatially explicit models accounting for spatial connections have better explanatory power than spatially disconnected ones for short-to-intermediate calibration windows, while parsimonious, spatially disconnected models perform better with long training sets. On average, spatially connected models show better predictive ability than disconnected ones. We suggest limits and validity of the various approaches and discuss the pathway towards the development of case-specific predictive tools in the context of emergency management.