Tick borne encephalitis (TBE) is a severe neurological disease caused by the TBE virus (TBEv), a flavivirus transmitted in Europe mainly by the ticks Ixodes ricinus and I. persulcatus. In nature, TBEv circulation is maintained in the environment by the co-occurrence of three major components: the virus, the vector and the presence of competent hosts, namely rodents. TBEv is typically distributed spatially and temporally as hotspots (foci of infection) and does not mirror the vector and host distributions. In recent decades, the incidence of TBE in human cases in Europe has been rising both in endemic and new regions, with altitudinal and latitudinal shifts, posing an increasing threat to public health. Therefore, the early localization of new TBE foci represents a priority at sanitary community level.
We systematically reviewed the existing literature including data on the drivers associated with the circulation of TBE in Europe: data from sixty-six full text papers were retrieved and analyzed, considering both biotic and abiotic factors.
Our review underlines the high level of heterogeneity, both in study design and type of variables, and the lack of values or thresholds associated with disease emergence.
We therefore aimed at identifying the most important ecological covariates for TBE hazard and at providing useful insights concerning possible uniform future directions. For instance, even though TBE has a typical hotspot distribution, it is present at continental scale and similarly are the distribution ranges of its main vectors and hosts. In order to investigate this aspect, advanced modelling techniques, such as spatial models based on high resolution satellite-derived data, have lately been implemented.
After assessing the main ecological covariates with higher predictive power for TBE hazard, we defined the main data sources, which include Earth Observation Data such as Land Surface Temperature (LST), Normalized Difference Vegetation Index (NDVI) and Land Cover. We subsequently validated our results on reported human cases of TBE and geo-referenced data of TBEv detections in ticks and hosts extracted from literature published between 2010 and 2021. The statistical analysis was focused on two European countries, namely Italy and Finland, which display several TBE foci and well represent the spatial variability of TBEv circulation at European level.
Culex pipiens mosquitoes are important vectors of emerging and re-emerging diseases in Europe. This mosquito species is able to adapt to a wide variety of environments, which give a complex picture in terms of trophic behaviour and vectorial capacities. A fine understanding of the vectors’ habitat suitability that facilitates survival, reproduction and dispersal becomes of paramount importance for determining the risk of local establishment, persistence and spread.
In the last 20 years, studies on vector distribution models have grown exponentially. Nowadays, the ever-increasing abundance of Remote Sensing and Earth Observation (EO) data, together with new Artificial Intelligence (AI) techniques, offers enormous opportunities for vector-borne disease investigations.
With the objective of explaining the spatial distribution of Culex pipiens in Abruzzo and Molise regions in central Italy, the integration of field entomological data, highly detailed remotely sensed imagery and newly innovative artificial intelligence algorithms, has been tested.
Two season campaigns of field collections have been carried out in 2019 and 2020, and the presence/absence and abundance of Culex pipiens in about 50 sites have been collected on a biweekly basis during the vector season (between May and November). The site locations have been chosen in a variety of climatic and environmental conditions [1].
The presence/absence (labels or annotations) of Culex pipiens in each entomological collection was associated with different EO datasets using the date of collection for the timing. Patches of 224x224 pixels of 20 meters spatial resolution were extracted around each site and for each date (ground truth dataset).
The EO datasets considered were:
- the multi-spectral bands captured through the optical devices onboard the Sentinel-2A and 2B satellites of the Copernicus programme, for each revisit time;
- the Land Surface Temperature Daytime and Nighttime from MODIS mission, every 8 days.
The whole dataset (1384 site observations) was split in training (80%) and test (20%) so that all the images of the same site could only belong to one of the two datasets, thus avoiding misleading results. A stratified k-fold technique was carried out with five-folds preserving the proportion of positive and negative labels between the train and test sets.
A Deep Convolutional Neural Network (DCNN) was applied to classify each site according to the presence/absence of Cx. pipiens mosquitoes (binary classification task). We evaluated three models: the first one (baseline) exploited a single multi-band image for predicting the given binary target; the second one integrated the baseline with the spatial relationships among sites, through a graph features aggregation method; the third model focused on the temporal aspect by using the features coming from a sequence of images of the same site.
Due to the reduced dataset dimension, we firstly performed and tested two different pre-training stages, followed by fine-tuning (knowledge transfer) on our data. In detail, we adopted a pre-training phase targeted on RGB bands (B4, B3, B2), taking advantage of the ImageNet dataset [2], and another pre-training phase which involved a self-supervised procedure called colourization [3]. This latter technique recovers the RGB information by taking as input the other spectral bands.
In the baseline model, we applied the DCNN to extract meaningful features from a single image (the closest before the catch date) per site. Then, the final classification layer delivered the probability of each site being positive or negative.
The second approach integrated the similarity among sites based on multiple factors. As in the baseline model, we first extracted features from each image by employing a DCNN. On these embeddings, we applied a graph aggregation layer, which influences features based on their similarity. We calculated this measure as a multidimensional similarity by considering temperature and haversine distance: sites having similar values of these measures reveal a high features correlation. The hypothesis is that this correlation -in terms of the above similarities- among sites can avoid false-negative or false-positive results.
The baseline model reached an F1 score of 0.80, which was increased up to 0.82 with the integration of the graph aggregation model.
The third approach focused on the temporal component by using sequences of Sentinel 2 images of the same site as input for the deep model: the temporal window was of ten images, taking the first images 15-24 days before the Culex pipiens catch date. We extracted the embedding for each image and applied an attentive module, which computes a weighted average of instances. In this way, the most relevant embeddings were combined to provide the final classification.
The third model reached an F1 score of 0.83.
The performances of the best approach were investigated in specific sites of interest, giving useful information for targeting surveillance activities in the following seasons.
This work describes a successful synergy between entomological field activities and the use of new and advanced technologies, i.e. Sentinel 2 satellite imagery and AI Deep Learning algorithms. The methodology adopted can be extended to the national territory and to other vectors, to support the Ministry of Health in the surveillance and control strategies for the vectors and the diseases they transmit.
References:
[1] Ippoliti C, Candeloro L, Gilbert M, Goffredo M, Mancini G, Curci G, Falasca S, Tora S, Di Lorenzo A, Quaglia M, Conte A. 2019. Defining ecological regions in Italy based on a multivariate clustering approach: A first step towards a targeted vector borne disease surveillance. PLoS ONE 14(7): e0219072. https://doi.org/10.1371/journal.pone.0219072
[2] Deng J, Dong W, Socher R, Li L, Kai Li and Li Fei-Fei, "ImageNet: A large-scale hierarchical image database," 2009 IEEE Conference on Computer Vision and Pattern Recognition, 2009, pp. 248-255, DOI: 10.1109/CVPR.2009.5206848.
[3] Vincenzi S., Porrello A., Buzzega P., Cipriano M., Fronte P., Cuccu R., Ippoliti C., Conte A., and Calderara S. “The color out of space: learning self-supervised representations for Earth Observation imagery,” Proceedings of the 25th International Conference on Pattern Recognition, Milan, Italy, pp. 3034 -3041 , 10-15 January 2021, 2020, DOI: 10.1109/ICPR48806.2021.9413112
Whilst COVID-19 has claimed the front news during the past 2 years, mosquitoes and the diseases they transmit have kept spreading. In this paper we review the multiple ways earth observation (EO) and derived environmental and meteorological co-variates can contribute to assess the risk and potential impact of mosquito-borne diseases in general and dengue fever transmitted by the invasive Asian tiger mosquito in particular.
First, we discuss the relationship of vectors, pathogens and hosts with their environment and how this affects or limits how earth-observation (EO) can contribute to map these vector-borne disease components. Based on this we discuss how EO data can be used as co-variates in a variety of methods for the spatial mapping and modelling of these components at a variety of scales.
We start with the area-wide distribution of invasive Aedes vectors: Where are they now? Where are they likely to spread? Will this be affected by climate change? Then, we discuss how vector seasonality can be modelled: When (i.e., in which week of the year) will the seasonal activity of established populations of Aedes albopictus (the Asian tiger mosquito) start in spring and when will it stop in autumn? These activity patterns significantly vary in Europe from South to North and from West to East. Once we know where they are and when they will be active other modelling approaches can be used to predict how many there are based on meteorological forecasts.
Within the limits discussed in the first part, similar methods can be used to model the distribution of the diseases transmitted by these mosquitoes. Taking dengue as an example we show how EO contributes to the mapping of global disease hotspots, how this should be fine-tuned within countries, and how this can contribute to select active outbreak areas. At a finer scale we show that the importance of man-made drivers becomes more important: local disease spread within areas of vector-presence is mainly driven by socio-economic variables and mobility networks. Nevertheless, mosquitoes being cold-blooded organisms external temperatures remain critical when thresholds for viral development aren’t reached.
In conclusion a summary is given of specific EO needs related to the discussed topics and some priorities are given for future developments.
Invasive Aedes mosquitoes are competent vectors of viruses debilitating for humans. These mosquitoes possess a unique combination of eco-physiological traits which underlay their rapid geographical range expansion and the consequent threat they pose to public health.
A considerable amount of experimental and observational data on the ecology of invasive Aedes is progressively being collected and made available, and the integration of such data into mechanistic modelling frameworks has been shown to effectively reproduce Aedes basic population dynamics. The outputs of such mechanistic models have often been suggested as supporting information for actions to mitigate local Aedes mosquito densities and limit the epidemiological risk due to local transmission of Aedes-borne viruses. Temperature is often used as the main environmental factor informing invasive Aedes mechanistic models as it is of pivotal importance for the life cycle of all poikilotherm arthropods. The availability and reliability of temperature datasets at higher spatial and temporal resolution is an important drawback that hinders detailed simulations of Aedes populations, and thus be inadequate for real world applications. This gap is being progressively filled by the development and increasing availability of two sources of climatic data: 1) climatic reanalysis of Regional/Global Circulation models, geophysical models calibrated and validated with observations from weather stations and 2) land surface physical observations derived from remote sensors onboard orbiting satellites.
We simulated Ae. albopictus population dynamics in France using the dynamAedes model informed with these two temperature data sources. dynamAedes is a recently proposed open-source, multi-scale, stage-based, and time discrete probabilistic modelling framework for Aedes mosquitoes. We compared the results obtained from the two simulations, focusing on the model accuracy in reproducing the observed spatial distribution and population dynamics temporal trend of Ae. Albopictus. The simulations were validated using an observational dataset of more than 3000 French municipalities reporting established populations of the vector, allowing us to highlight advantages and limitations of the two considered temperature data sources.
Aedes albopictus and Aedes aegypti mosquitoes have a worldwide distribution and are well adapted to urban environments. Because they are the main vectors of dengue, chikungunya and Zika viruses, these two species constitute a threat for public health both in tropical and temperate regions. To better target surveillance and control of Aedes-borne diseases, there is a need for tools with the capacity to predict the spatially distributed dynamics of mosquito vectors at a local scale. In addition, to be used by public health authorities and vector control services, such tools need easy-to-use interfaces allowing a customization by the user according to the geographical and entomological contexts.
Various approaches exist to model the dynamics of mosquito populations and to predict their spatial distribution. As mosquitoes depend closely on climatic and environmental conditions, satellite-based information (e.g. vegetation, urbanization type) can prove to be valuable input data for the model. Here, we provide both the scientific community and the operational stakeholders an efficient way for implementing a generic mosquito life cycle-based model, driven by meteorological variables (temperature and rainfall). The implementation of ‘ARBOCARTO’ considers the landscape context described from very high spatial imagery and/or ancillary data provided by the user. We present its application in various geographical contexts (mainland France and its overseas departments) and for two Aedes mosquitoes species population (Aedes albopictus and Ae. aegypti).
In highly diverse environments and latitudes, the comparison between the model outputs and observed entomological data demonstrated the ability of ‘ARBOCARTO’ tool to provide valuable complementary information to existing entomological surveillance programs. The different functionalities allow the user to test different scenarios, such as the impact on mosquito dynamics of prevention measures (e.g., reduction of the number of breeding sites) or control actions (e.g., pulverization of insecticides). Thanks to its user-friendly interface, ‘ARBOCARTO’ could be adopted by a broad community of managers involved in vector control.
Rift Valley fever (RVF) is a vector-borne disease that has severe impacts on livelihoods, national and international markets, and human health particularly to vulnerable pastoral communities with low resilience to economic and environmental challenges. RVF is currently limited to Africa and parts of the Near East with the potential to expand globally. In livestock, the disease is spread primarily by mosquitoes and the movement of animals. The clinical disease, which is zoonotic, has been observed in sheep, goats, cattle, buffaloes, camels and humans. RVF can result in widespread febrile illness in humans, and may be associated with severe and sometimes fatal sequelae in under 1 percent of cases.
Outbreaks of RVF are closely associated with climate anomalies, such as periods of heavy rains and prolonged flooding, which increase habitat suitability for vector populations and thus influence the risk of disease emergence, transmission and spread. Early warning systems represent an essential tool to enable national authorities to implement measures preventing outbreaks, as they provide information on occurring animal health hazards that might evolve into disasters unless an early response is undertaken.
In this context, the Food and Agriculture Organization of the United Nations (FAO) has developed a web-based RVF Early Warning Decision Support Tool (RVF DST), which integrates near real-time RVF risk maps with geospatial data, RVF historical and current disease events from EMPRES Global Animal Disease Information System (EMPRES-i) and expert knowledge on eco-epidemiology.
The dynamic prediction model used by FAO builds upon the work by NASA, that utilized vegetation and rainfall anomalies as a proxy for ecological dynamics to map areas at potential risk of RVF in Eastern Africa. The innovative part of FAO’s work consists in its calibration and implementation in an early warning tool that enables near-real-time monitoring and forecasting of RVF at-risk areas in Africa.
The RVF DST brings together a wealth of relevant data for monitoring climate variability and other RVF risk factors, including: observed and forecasted precipitation and anomalies; Normalized Difference Vegetation Index and anomalies; land surface temperature; El Niño forecasts; past and current RVF occurrences; estimates and geographic distribution of susceptible, at-risk livestock species; human populations; market places; road networks; animal trade routes; water bodies and irrigation areas; land cover; soil.
In addition, the tool offers expert knowledge on RVF eco-epidemiology, risk assessment and categorization, analytical functions and charts of trends in major risk factors, as well as recommended actions to guide appropriate responses to RVF at country level. An automated risk analysis report with charts of major risk factors, estimated animals at risk and risk maps can be downloaded.
The RVF DST, developed as part of the early warning component of EMPRES, is being implemented by FAO’s Emergency Centre for Transboundary Animal Diseases (ECTAD) to strengthen RVF preparedness, response and contingency plans. It was created through the input of a pool of international, regional and national experts, as well as epidemiologists from national veterinary services. Adopting a collaborative approach has helped to ensure the sustainability of the RVF DST by cultivating ownership among the beneficiaries.
Developed in 2019, the RVF DST was integrated in the online FAO Hand‑in‑Hand geospatial data platform in July 2020 and piloted in three RVF endemic countries in East Africa – Kenya, Uganda and the United Republic of Tanzania – to provide decision-makers with near real‑time RVF risk maps and assessments, which are updated on a monthly basis at 250 meters of spatial resolution. Covering the period of January 2003 to the present day, the data is provided in near real time.
The Hand-in-Hand geospatial platform is a web-based dashboard providing a suite of geospatial data from FAO and other agencies for use by all countries and partners, promoting transparency and collaboration. The platform has significantly increased the interoperability of FAO geospatial data and the cost-effective maintenance and sustainability of different FAO geospatial applications, including the RVF DST. Timely and reliable information on disease occurrence and its risk factors enhances early warning and response to transboundary animal diseases (TADs) including RVF, and supports their progressive control and elimination.
The RVF DST has enhanced the Organization’s capacity to identify high-risk areas and issue alerts and early warning messages for prevention and control in countries at risk of RVF occurrence. These alerts and messages are issued well before the reporting of the first signs of RVF infection in the countries with a prediction capacity of at least 1-2 months. The output of the tool is used together with results from other RVF monitoring and surveillance activities (such as sentinel herd monitoring) and expert knowledge for near real-time validation of the RVF potential hotspots to inform decision-makers and support early response.
Overall, the RVF DST has contributed to an improved state of vigilance and preparedness in the Eastern African region and enhanced collaboration between FAO, national veterinary services and strategic regional partners such as the World Health Organization (WHO), the Intergovernmental Authority on Development (IGAD), and the World Organisation for Animal Health (OIE) beneficiaries.
The tool is used to build capacity for early warning and forecasting at country level, and demonstrates how near real-time modelling, risk forecasting and digital innovation can enhance preparedness and anticipatory actions.