Conference Agenda
Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).
Please note that all times are shown in the time zone of the conference. The current conference time is: 24th Aug 2026, 01:53:38am WEST
External resources will be made available 30 min before a session starts. You may have to reload the page to access the resources.
|
Daily Overview |
| Session | |
|
Climate Damages Impact Location: B128 Session Chair: Edouard Civel, EDHEC Business School and Climate Economics Chair | |
    | |
Bridging Climate Data Gaps in Brazil: An AI-Based Infrastructure for Climate Risk and Economic Analysis Pontifícia Universidade Católica de Goiás, Brazil Inconsistent and incomplete climate records remain a major constraint for environmental and resource economics, particularly in regions with sparse or interrupted meteorological observations. In Brazil, missing data in national weather station networks introduce measurement error, bias climate damage functions, and increase basis risk in applications ranging from agricultural insurance to climate-informed public planning. This paper presents an AI-based climate data infrastructure designed to improve the continuity, auditability, and economic usability of climate exposure data. The proposed framework combines global reanalysis products and local surface observations through a staged pipeline that separates physical validation from operational imputation. A Transformer-based model trained on ERA5 reanalysis data under controlled masking is used to learn physically coherent spatiotemporal structures, while Dense Neural Networks (DNNs) provide operational baselines for imputing missing values in Brazil's INMET station network. While qualitative experiments demonstrate that Transformer-based models preserve physically consistent spatial gradients in reanalysis fields, the application of this architecture to the INMET network is currently in progress. Preliminary results show that the implemented DNN baselines substantially outperform classical methods such as K-Nearest Neighbors and linear regression across key climate variables. Beyond methodological contributions, the study introduces a prototype national-scale climate intelligence platform that delivers harmonized, imputed, and metadata-rich climate series. By reducing information costs and improving the reliability of climate exposure measures, the proposed infrastructure supports more credible economic analysis, climate risk assessment, and evidence-based policy design in Brazil. | |