Thesis Code: T9

Machine Learning for advancing the detection of drought events

Short description:

One of the major consequences of climate change is the increased frequency and severity of extreme weather events, such as droughts, floods, severe storms, wildfires, on top of gradual changes in temperature and precipitation. Among all atmospheric hazards, droughts are still the least predictable(1). This is due to the complexity of the phenomenon, for which there is no single definition, nor a comprehensive and sound methodology to assess its impacts across all sectors. For this reason, although drought monitoring and management are extensively studied in the literature(2), traditional drought indices often fail at yielding precise information on detecting critical events, capturing only a portion of the entire process.

Machine Learning approaches are a valid alternative to traditional drought indices for the detection of drought events through the analysis of large datasets of diverse hydroclimatic variables. Specifically, given a set of candidate features (e.g., temperature, precipitation, soil moisture, streamflow, lake water levels), the use of feature extraction algorithms allows the identification of the most relevant drivers of a drought event, which can be combined into a model able to detect the presence and/or the intensity of a drought event.

The aim of this thesis is benchmarking state-of-the-art feature extraction algorithms, possibly testing their performance on different case studies. The student is expected to carry out the following activities:

  1. Literature review: reviewing the state of the art of existing feature extraction algorithms, with a specific focus on their use for drought monitoring
  2. Data collection: acquisition of observed/reanalysis data (e.g. ERA5(4)) of relevant hydroclimatic variables, such as precipitation, temperature, streamflow
  3. Computational experiments: 
    • run of alternative feature extraction algorithms;
    • comparative analysis of the results.

References

  1. Mishra, A. K., Singh, V. P., 2010. A review of drought concepts. Journal of Hydrology 391 (1-2), 202–216.
  2. Pedro-Monzonìs et al. (2015), A review of water scarcity and drought indexes in water resources planning and management, Journal of Hydrology
  3. https://edo.jrc.ec.europa.eu/edov2/php/index.php?id=1000
  4. https://climate.copernicus.eu/climate-reanalysis 

Relevant courses and knowledge:

Natural Resources Management

Number of students:

1 or 2

Requisites:

The student should be comfortable with data handling and programming skills (Matlab or Python).