ISI Regional Webinar: Advanced Statistical Modeling for Environmental and Earth Observation Data

Moderator:
●    Renata Rojas Guerra, Federal University of Santa Maria (UFSM), Brazil 
Speakers:
●    Bruna Gregory Palm, Blekinge Institute of Technology (BTH), Sweden 
●    Paola Crippa, University of Notre Dame, United States
 

Abstract of the webinar
From satellite imagery to atmospheric pollution monitoring, advanced statistical frameworks have been proposed to enable more accurate detection, prediction, and risk assessment across multiple spatial and temporal scales. This ISI Regional Webinar explores recent advances in this area, featuring contributions from female statisticians whose work connects methodological development with real-world environmental applications across international research settings. Topics include flexible regression models for invasive plant detection in NDVI imagery and probabilistic airshed mapping approaches for transboundary PM2.5 pollution and health risk attribution. Through these case studies, the webinar illustrates how modern statistical approaches can contribute to data-driven decision-making in environmental and public health contexts. 

The webinar is also organised in connection with the Special Issue celebrating the 140th anniversary of the International Statistical Institute (ISI), promoted by the Revista Colombiana de Estadística in cooperation with the ISI Outreach Committee for Latin America and the Caribbean and the ISI Young Statisticians Committee. The Special Issue highlights contributions from the global statistical community, including both theoretical and applied statistics, and features guest editors representing all seven ISI Associations. More information about the Special Issue is available here.

Short talk title - Bruna Gregory Palm: 
Inflated Modified Kumaraswamy Regression Model for Invasive Plants Detection in NDVI Imagery

Short talk abstract: 
This study proposes the inflated modified Kumaraswamy (iMK) distribution, a flexible probability model defined on the unit interval [0,1]. It captures asymmetric behaviors while accommodating inflation at zero, one, or both boundaries, as commonly observed in normalised difference vegetation index (NDVI) data. Based on the iMK distribution, we develop a new regression model (iMK regression (iMKreg)) suitable for double-bounded responses. From this model, we derive a detection tool for invasive plant species, particularly applicable to NDVI imagery. Model performance was evaluated using synthetic NDVI data, with further assessment of predictive accuracy and detection efficacy conducted on a real-world measured NDVI image. The application to detecting black-grass (Alopecurus myosuroides) in wheat crops in southern Sweden shows that the iMKreg model outperforms both standard Gaussian-based linear regression and existing inflated Kumaraswamy regression models.

Short talk title - Paola Crippa: 
Probabilistic Airshed Mapping for Transboundary PM2.5 and Health Risk Attribution

Short talk abstract: 
Air pollution is the leading environmental cause of premature mortality worldwide, yet its transboundary nature makes attribution and regulation challenging because emissions and health impacts often extend beyond political boundaries. This presentation introduces a reduced-complexity, data-driven airshed framework that applies unsupervised learning and clustering to quantify cross-boundary PM2.5 pollution and associated premature deaths across jurisdictions. By integrating observational and reanalysis datasets, the framework captures meteorology, topography, atmospheric chemistry, and extreme events while tracking exposure and health risk over time. Using probabilistic clustering, the method identifies spatial patterns in concentration fields to delineate dynamic airsheds that evolve with atmospheric and societal conditions. The framework is scalable and efficient, enabling multi-year updates of airshed maps with uncertainty estimates. Applied to the contiguous United States, it reveals persistent inequalities in cross-state pollution impacts, with transboundary transport contributing to ~40% of premature deaths despite substantial declines in overall mortality since 1998. The results show how unsupervised learning can complement chemical transport models to support more effective air quality regulation and cross-jurisdictional environmental governance.