IST597 Project SP 2018

Goal and objective:

Recent years have observed many wildfire events across Arctic tundra ecosystems. Currently available sensor-derived rich geospatial datasets provide opportunity to explore the nature and degree of relationships between a changing climate and wildfire characteristics (i.e. occurrence and intensity). However, with higher data volume and dimensionality substantial challenges exist in deriving reliable and refined insights on these relationships. In addressing that challenge, this project aims to apply machine learning techniques to explore Arctic climate-wildfire dynamic relationships.

More specifically, this project will try to answer if and how spatially and temporally aggregated climate variables (e.g. moisture, temperature, evaporation, etc.) are associated with wildfires, including relative influence of each climate variable for wildfire occurrences. In addition, a comparison will be made among applied techniques’ performances in addressing this challenge.

Data Source:

  • Wildfire: NASA MODIS dataset (2001-2015)
  • Climate: NASA MERRA-Land dataset (1980-2015)
Climate variables are:
  1. Mean soil surface temperature (K)
  2. Total surface precipitation (kg m-2 s-1)
  3. Mean soil moisture content (kg m-2 s-1)
  4. Mean surface evaporation (kg m-2 s-1)
  5. Average soil surface temperature anomaly
  6. Average surface precipitation anomaly

Machine Learning Techniques:

  1. Logistic Regression
  2. Naive Bayes
  3. Decision Tree Learning
    • Classification Trees (e.g. AdaBoost and Random Forest)
  4. Support Vector Machine

Relevant Articles:

  1. Cortez, P., & Morais, A. D. J. R. (2007). A data mining approach to predict forest fires using meteorological data.
  2. Masrur, A., Petrov, A. N., & DeGroote, J. (2018). Circumpolar spatio-temporal patterns and contributing climatic factors of wildfire activity in the Arctic tundra from 2001–2015. Environmental Research Letters, 13(1), 014019.
  3. Özbayoğlu, A. M., & Bozer, R. (2012). Estimation of the burned area in forest fires using computational intelligence techniques. Procedia Computer Science, 12, 282-287.

 

 

Leave a Reply

Your email address will not be published. Required fields are marked *