ITIoannis Tsilikas

Ioannis Tsilikas

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Ioannis Tsilikas has a total of 16 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Machine Learning Approaches for Ozone Forecasting in Urban and Rural Areas of Greece: A Comparative Study Using an IoT Monitoring Network

    by Yiannis Kiouvrekis, Christos Christakis, Ioannis Tsilikas, Theodor Panagiotakopoulos

    Ground-level ozone (O3) is a secondary photochemical pollutant whose formation under intense Mediterranean solar radiation makes Greece prone to elevated concentrations, with well-documented respiratory and cardiovascular health effects. Forecasting of ozone in Greece has nonetheless remained limited: prior work has been largely statistical, focused on one or two sites in Athens, and restricted to a single horizon, and no study has compared machine learning forecasts across the urban-to-rural gradient of the national monitoring network. This study addresses that gap with, to our knowledge, the first multi-site comparison of machine learning ozone forecasts spanning the full urban-to-rural typology gradient of a Greek monitoring network, introducing a typology × model accuracy cross-matrix that links site character directly to forecasting performance and offers concrete, horizon-specific operational guidance for early warning deployment. We develop a single, reproducible pipeline and compare three model families, Support Vector Regression (SVR), Random Forest (RF) and gradient boosting (GBM), against a naïve persistence baseline, for next-hour and next-day-maximum O3 prediction at eight urban, suburban and rural stations of the National Air Pollution Monitoring Network (2021–2024). All predictors are derived from the ozone series itself (lagged, rolling and cyclical calendar features); models were tuned and evaluated using a strictly chronological 70/15/15 train/validation/test split per station, with hyperparameters selected on the validation partition and the test partition evaluated only once to prevent temporal leakage. At the next-hour horizon, all models outperformed persistence at every site, with the single exception of SVR at the rural THR site; gradient boosting was best or joint-best (test R2 from 0.83 at the heavily titrated central-urban site to 0.94 at a rural site), and a Wilcoxon signed-rank test on paired per-sample errors confirmed that GBM significantly outperformed Random Forest at all seven evaluated stations (p<0.05, and p<0.001 at six of them). The next-day-maximum task was substantially harder (R2 between 0.41 and 0.79), and the ranking reversed, with SVR being the best model almost everywhere. Crucially, forecast accuracy declined systematically from rural and suburban toward central-urban sites: the same nitric oxide titration that suppresses urban ozone is also associated with degraded predictability. These results, summarised in the typology × model accuracy cross-matrix, indicate that an operational early warning system for Greece should adopt horizon-specific models, gradient boosting for next-hour nowcasting and SVR for day-ahead maxima, with explicitly higher uncertainty in the urban core, and can be deployed on the network’s real-time (IoT) data stream using ozone observations alone.


  • Open Access

    Interpretable Machine Learning Models Using SHAP for Hourly and Daily-Maximum Carbon Monoxide Forecasting at Urban Air Quality IoT Monitoring Stations in Greece

    by Yiannis Kiouvrekis, Christos Christakis, Ioannis Tsilikas, Theodor Panagiotakopoulos

    Carbon monoxide (CO) remains an important urban air quality and traffic-exposure tracer despite rarely exceeding regulatory limits in modern European cities. This study presents an interpretable machine learning framework for short-term CO forecasting at two operational horizons—next hour and next-day maximum—applied identically to four automated monitoring stations spanning central-urban, urban, urban-background, and suburban typologies in the Greater Athens Area (2021–2024). Using only univariate CO history and calendar-derived features, four learners (support vector regression, random forest, gradient boosting, and a multilayer perceptron) were benchmarked against a naïve baseline under a chronological train/validation/test split. At the next-hour horizon, all learners outperformed naïve at every station, with gradient boosting achieving the best or joint-best skill (test R2=0.81–0.90); Wilcoxon signed-rank tests confirmed that these small but consistent margins were statistically significant. The next-day-maximum task proved substantially harder (R2=0.44–0.59), with the neural and random forest models overtaking gradient boosting and SVR losing competitiveness. A SHAP (Shapley Additive Explanations) analysis of the next-hour gradient-boosting model showed that the most recent hourly lag dominates the forecast, with an effect nearly an order of magnitude over any other predictor, with hour-of-day encoding and short-lag rolling statistics contributing secondary, sign-consistent effects—providing a transparent, mechanistic account of model behavior rather than a black-box skill score. Unlike ozone, a secondary pollutant whose predictability degrades toward the trafficked urban core, CO concentration and forecastability increase together at the traffic-dominated site, indicating that primary-pollutant forecasts are most reliable precisely where exposure is greatest. These findings support a horizon-specific, interpretable forecasting strategy for operational deployment on real-time monitoring networks.


  • Open Access

    Global Mapping of Population Exposure to Upstream Gas Flaring Using Integrated VIIRS Nightfire and GHSL Data, 2016–2023, with Projections to 2030

    by Sotiris Zikas, Christos Christakis, Loukas-Moysis Misthos, Ioannis Psomadakis, Angeliki I. Katsafadou, Ioannis Tsilikas, George C. Fthenakis, Vasilis Vasiliou, Yiannis Kiouvrekis

    Gas flaring from upstream oil and gas production remains a significant source of air pollution and toxic emissions, with major implications for human health and climate. However, the number of people living near flaring has not been quantified globally. This study presents the first worldwide, settlement-scale assessment of populations living within 1 km and 3 km of active upstream flare sites between 2016 and 2023, with projections to 2030. Using the VIIRS Nightfire satellite product, which provides global detections of high-temperature combustion sources, and the Global Human Settlement Layer (GHSL) population and settlement data, we developed a transparent and reproducible geospatial workflow to compute proximity-based exposure indicators by buffering flare locations and intersecting them with population rasters The analysis provides consistent estimates across five settlement categories: rural, peri-urban/suburban, semi-dense urban, dense urban, and urban centres. The VIIRS-based flaring time series combined with GHSL projections allows us to estimate how many people are likely to live near upstream flares under current flaring patterns by 2030. Results show that exposure is concentrated in a few oil-producing countries. Nigeria remains the most affected, with over 100,000 urban residents exposed in 2023. India and Pakistan dominate peri-urban and semi-urban exposures, while Indonesia and Iraq persist as multi-settlement hotspots. Although moderate declines are observed in China and Iran, little progress is evident in Nigeria, Mexico, and Indonesia. Projections for 2030 suggest exposure will increase substantially, driven by population growth and urban expansion, with about 2.7 million people living within 1 km and 14.8 million within 3 km of flaring sites. The findings establish the first globally consistent baseline for population exposure to gas flaring, supporting the monitoring and mitigation objectives of the Zero Routine Flaring by 2030 initiative.

  • Open Access

    Explainable Machine Learning-Based Electric Field Strength Mapping for Urban Environmental Monitoring: A Case Study in Paris Integrating Geographical Features and Explainable AI

    by Yiannis Kiouvrekis, Ioannis Psomadakis, Kostas Vavouranakis, Sotiris Zikas, Ilias Katis, Ioannis Tsilikas, Theodor Panagiotakopoulos, Ioannis Filippopoulos

    The objective of this study is to determine the optimal machine learning model for constructing electric field strength maps across urban areas, advancing the field of environmental monitoring. These models are unique because they use a detailed dataset that goes beyond electromagnetic readings, incorporating information like population density, urbanization levels, and building characteristics. This novel approach, combined with explainable AI, helps identify the key factors affecting electromagnetic exposure. The models enable the creation of highly detailed and dynamic maps of electromagnetic pollution. These maps are not just static snapshots, they can track changes over time, evaluate the success of mitigation efforts, and provide deeper insights into how electromagnetic fields are distributed in urban areas. To construct a detailed electric field strength map, we conducted an extensive analysis using 410 machine learning models across the urban area of Paris, incorporating three fundamental approaches: k-nearest neighbors, neural networks, and decision trees. This comprehensive exploration allowed us to evaluate and optimize various model configurations, ensuring robust and accurate predictions of electric field strength across diverse urban environments. The kNN model exhibited the most consistent performance, with an RMSE of 1.63 and an SD of 0.20. The analysis indicates that kNN outperforms simple neural networks and decision trees in terms of both RMSE and performance stability. From the SHAP analysis, we conclude that the feature representing the total volume of buildings in the area around each antenna (V) is the most significant in predicting electromagnetic field strength in the kNN regression model, consistently showing a high impact across predictions. The population density feature (POP) also demonstrates considerable influence.


  • Open Access

    A Comparative Analysis of Explainable Artificial Intelligence Models for Electric Field Strength Prediction over Eight European Cities

    by Yiannis Kiouvrekis, Ioannis Givisis, Theodor Panagiotakopoulos, Ioannis Tsilikas, Agapi Ploussi, Ellas Spyratou, Efstathios P. Efstathopoulos

    The widespread propagation of wireless communication devices, from smartphones and tablets to Internet of Things (IoT) systems, has become an integral part of modern life. However, the expansion of wireless technology has also raised public concern about the potential health risks associated with prolonged exposure to electromagnetic fields. Our objective is to determine the optimal machine learning model for constructing electric field strength maps across urban areas, enhancing the field of environmental monitoring with the aid of sensor-based data collection. Our machine learning models consist of a novel and comprehensive dataset collected from a network of strategically placed sensors, capturing not only electromagnetic field readings but also additional urban features, including population density, levels of urbanization, and specific building characteristics. This sensor-driven approach, coupled with explainable AI, enables us to identify key factors influencing electromagnetic exposure more accurately. The integration of IoT sensor data with machine learning opens the potential for creating highly detailed and dynamic electromagnetic pollution maps. These maps are not merely static snapshots; they offer researchers the ability to track trends over time, assess the effectiveness of mitigation efforts, and gain a deeper understanding of electromagnetic field distribution in urban environments. Through the extensive dataset, our models can yield highly accurate and dynamic electric field strength maps. For this study, we performed a comprehensive analysis involving 566 machine learning models across eight French cities: Lyon, Saint-Étienne, Clermont-Ferrand, Dijon, Nantes, Rouen, Lille, and Paris. The analysis incorporated six core approaches: k-Nearest Neighbors, XGBoost, Random Forest, Neural Networks, Decision Trees, and Linear Regression. The findings underscore the superior predictive capabilities of ensemble methods such as Random Forests and XGBoost, which outperform individual models. Simpler approaches like Decision Trees and k-NN offer effective yet slightly less precise alternatives. Neural Networks, despite their complexity, highlight the potential for further refinement in this application. In addition, our results show that the machine learning models significantly outperform the linear regression baseline, demonstrating the added value of more complex techniques in this domain. Our SHAP analysis reveals that the feature importance rankings in tree-based machine learning models differ significantly from those in k-NN, neural network, and linear regression models.