Towards meaningful wildfire risk assessment spatial models
|
Problem Statement:
Wildfire risk assessment spatial models constitute a critical technological tool, serving as an important defence against increasingly catastrophic fire seasons. By integrating diverse geospatial data layers, these models transform raw information into actionable intelligence. They move beyond simple hazard maps to quantify and visualize risk as a dynamic interplay between the probability of ignition, the potential fire intensity, and the profound consequences to human life, property, and ecological assets. This spatial explicitness is paramount for proactive land management, guiding fuel reduction treatments and forest restoration projects where they are most needed. These risk maps empower stakeholders at all levels to shift from a reactive posture of fire suppression to a proactive strategy of risk mitigation, strategically allocating limited resources to protect what matters most and build resilience in a warming world. Gaea is an online tool which constitutes one of the first environmental digital twins at country-scale around the world. It captures the geographical area of the island of Cyprus. Gaea consists of various geospatial and geo-analytics services which offer rich contextual information related to weather, geophysical aspects, land coverage and use, buildings, and real estate properties, as well as environmental risks and risks related to climate change.
Further information about Gaea is provided below:
https://superworld.cyens.org.cy/product1.html
https://superworld.cyens.org.cy/project15.html
An important service of Gaea is wildfire risk assessment, which calculates the risk of wildfire at every location around the island of Cyprus by means of a machine learning-based model.
Task:
While the wildfire risk assessment service of GAEA has been implemented and has satisfactory results, there is space for improvement, plus there is very little work performed related to transparency and understanding. Which environmental parameters and their values affect some location to have high risk? Which parameters and their values help to have low risks?
The student is expected to study the risk assessment ML model, try to improve it at the same time work on AI transparency, to understand how the model works and how to interpret its predictions in terms of lessons learned for local communities.
Work:
10% Theory, 50% Modelling, Coding and Testing, 20% Evaluation and Validation, 20% Writing
Contact:
Andreas Kamilaris: a.kamilaris@utwente.nl