Home ITCPhD Defence Irene Garcia-Martí

PhD Defence Irene Garcia-Martí

Mapping tick dynamics and tick bite risk using data-driven approaches and volunteered observations

Irene Garcia-Martí is a PhD student in the department of Geo-Information Processing. Her supervisor is prof.dr. R. Zurita-Milla from the faculty of Geo-Information Science and Earth Observation.

Human activities have induced global changes, which among other impacts are leading to the re-emergence of vector-borne diseases. Climate change, human and demographic developments, socio-economic exchanges, and the increase of human outdoor recreational activities, are some of the relevant factors that are contributing to the occurrence of vectors (e.g. mosquitoes, ticks) outside their endemic enclosements and facilitating the global trans- mission of vector-borne diseases between regions. The World Health Organization (WHO) has identified nine type of vectors that can cause, at least, 16 major vector-borne diseases in humans. Major tick-borne diseases comprise two bacterial infections (i.e. Lyme borreliosis, tick-borne encephalitis) and one viral infection (i.e. Crimea-Congo haemorrhagic fever), although there are several minor tick-borne diseases (e.g. rickettsial diseases, relapsing fever) with local importance.

Ticks are pervasive ectoparasites with a limited motility that require a wide array of biotic (e.g. environment, wildlife) and abiotic (e.g. weather, land- scape structure) conditions ensuring their survival. Scientists have reported a latitudinal and altitudinal expansion of tick range in the last decades, which has been attributed to two related factors: global warming has turned unsuitable habitats into suitable ones and, subsequently, various wildlife species (e.g. rodents, birds, ungulates) have expanded their ranges and introduced ticks in new locations.

The expansion of the range of ticks is not the sole element prompting tick- borne diseases. Our planet is experiencing an increasing urbanization. Urban sprawl has increased the amount of residential areas in the periphery of cities, which are in closer contact with green spaces and nature. As a response, several bird and mammal species have adapted their ethology to live at the interface between forests and urban regions, where the chances of finding more food and less predators are higher. However, the proximity between nature and urban areas also means that pathogens and parasites carried by wildlife are getting closer to citizens. This phenomenon also explains the increasing hazard for tick-borne diseases. In addition, the progressive adoption of healthier lifestyles prompts citizens to carry more outdoor activities leading to a higher exposure to tick-borne diseases.

Lyme borreliosis (LB) is a tick-borne disease that has experienced a substantial spatio-temporal expansion in the Northern hemisphere in the last 20 years. Scientists and clinicians in nine European countries, USA, and Canada, have reported that the incidence of LB has steadily increased. In recent years, however, sub European sentinel networks of general practitioners have identified the first signs of stabilization. Yet, each year, roughly 25,000 Dutch citizens are diagnosed with LB. Albeit most of them respond well to the antibiotic treatment, there is a minority of patients reporting persisting symptoms that might lead to chronic symptoms and disability.

Ticks are the vehicle that pathogens utilize to infect new organisms, therefore, it is utterly important to monitor tick dynamics that enable the identification of hazardous locations for LB infection. The ubiquity of humans in natural spaces and the small size of ticks poses multiple challenges for public health organizations to monitor this disease. This is why two Dutch organizations started citizen science-based initiatives to collect data that can shed light on tick bite risk and tick dynamics.

During the period 2006-2012 the educational phenology platform Natuurkalender, linked to Wageningen University, gathered nearly 10,000 volunteered tick bites. This pioneering project attracted the attention of the Dutch National Institute for Public Health and the Environment (RIVM), and in 2012, the platform Tekenradar was launched by these two organizations. Tekenradar has collected over 50,000 volunteered tick bite reports in the Netherlands. To the best of our knowledge, these initiatives constitute the first citizen science projects that specifically focus on ticks and tick-borne diseases. Also in 2006, a group of scientists from Wageningen University started a country wide project to assess tick dynamics. A group of trained volunteers sampled a transect of forest on a monthly basis using a method called blanket dragging, counting the caught ticks every 25 meters. This project was carried out during the period 2006-2016 and created a unique collection of volunteered data.

These unique collections of volunteered observations on tick bites and tick dynamics enable the study of public health threats, such as LB. Volunteered data has the potential of monitoring complex and elusive environmental phenomena at an unprecedented spatio-temporal resolution. However, volunteered data is not exempt of several issues and challenges that require attention. Data quality and representativeness are some of the challenges that required attention before feeding these observations to scientific work- flows.

In this PhD thesis, we focus on extracting spatial patterns and temporal trends from these volunteered data collections with two objectives in mind: modelling tick dynamics to identify the major drivers of tick populations, and modelling human activities to identify potentially risky factors and regions for LB. This disease is the product of a complex zoonotic cycle in which biotic and abiotic factors are weaved together. In our approach, we use machine learning methods to combine these data-sources about these factors and to account for the non-linearity of the interactions within the zoonotic cycle. In this way, we are able to create data-driven models capable of predicting daily tick activity and identifying risky locations for LB.

This journey towards the modelling of tick dynamics and tick bite risk has been structured in four steps. First, we present an approach to calculate tick dynamics stemming from the volunteered tick activity data. For this, we investigate the impact that long- and short-term variables have on tick activity, and we provide a model capable of predicting daily tick activity in forests at the national scale . Second, we introduce an extensive exploratory analysis to identify the most recurrent human and environmental patterns found in the tick bites dataset. This analysis enables the assessment of whether the tick bites collection is representative of the phenomenon under study, and helps elucidating whether this collection contains a clear tick bite signal. Third, we created a novel map of human exposure to tick bites in forested areas. We demonstrate that the risk of tick bite is strongly influenced by human behaviour, rather than by the particular tick dynamics in a given location. This map can also be used to identify locations where citizens are most exposed to ticks. Fourth, we develop a tick bite risk model that integrates information on tick dynamics (hazard) and human exposure to tick bites. This final analysis relies on the tick dynamic model developed previously, the knowledge acquired after the second and third analytical workflows, and on a series of human exposure indicators based on accessibility and landscape attractiveness. All these analytical workflows contribute at creating actionable geo-information products that can assist decision makers at designing tick bite prevention campaigns.

This PhD thesis has led to several scientific and methodological contributions that constitute a step-forward in the fields of spatio-temporal machine learning and environmental modelling. From a scientific point of view, our data-driven approaches showed that water-related features determine tick activity at multiple time scales, that tick bite risk is strongly influenced by human exposure rather than by tick activity, and that it is possible to localize and map the riskiest locations for tick bites, often located along the forest-urban interface. Regarding methodological contributions, we propose two modifications for a well-known machine learning method (i.e. Random Forest). These modifications extend the canonical functionalities of this method, enabling it to learn from skewed and zero-inflated distributions and timeseries data. To conclude, we believe that our work demonstrates that citizen science projects can produce valuable volunteered information that can feed scientific workflows to study elusive geographic phenomena, and that data-driven machine learning methods are paving the road for novel research lines on ticks and vector-borne diseases.