This project has a relevant contribution to the following CERENA strategic areas:
- Machine Learning and Artificial Intelligence;
- Geomodelling, Geophysics and Geochemistry;
- Physical, Chemical and Biological Processes
Supervisor: Amílcar Soares
Co-Supervisor: Ana Meira e Castro
Given her excellent curriculum in climatic extremes, Ana Cristina Costa (U.Nova Lisboa), will collaborate to support the selected student work
Mediterranean basin countries are among the most affected by the droughts due to the climate change effects. The objectives of this proposal is the assessment of risk of droughts for managing its impacts based on statistical homogenization and Spatial Data Sciences methodologies. The methodological framework is applied to the south of Portugal with historical climate data (precipitation) of last 50 years.
Based on a long series of daily precipitation available in a set of monitoring stations in South of Portugal, drought risk will be characterized according the following outline:
Task 1- The first important and fundamental goal of this proposal is to select the periods, from the long historical time series, with homogeneous climate data. The homogenization of climate data is of major importance to make the conclusions of climatic and hydrological studies to be reliable and unbiased. A set of statistical methods [1] will be applied and tested for this purpose.
Task 2- In a second step one intends to achieve the following objectives, based on selected periods of homogeneous data: i) characterization of spatial-temporal pattern of main droughts indicators; ii) Define the most critical areas with significative local impacts of temporal or spatial trends; iii) Characterize a methodology for short term prediction of drought risk in each region of south of Portugal. Previous studies [1], [2] and [3] will be accounted for bench mark of the evolution of the phenomenon.
To achieve those objectives, Spatial Data Sciences methodologies will be applied and integrated: Geostatistics for space-time patterns characterization, and stochastic simulation for uncertainty and risk evaluation; Functional Data Analysis for classification and summarizing the droughts indicators in historical long time series of precipitation; both results of previous points will condition Machine Learning algorithms (Physic guided) for short term prediction of droughts risk.
[1] AC Costa, A Soares, 2009, Trends in extreme precipitation indices derived from a daily rainfall database for the South of Portugal, International Journal of Climatology.
[2] AC Costa, A Soares, 2008, Homogenization of climate data: review and new perspectives using geostatistics, Mathematical geosciences 41 (3), 291-305
[3] RM Durao, MJ Pereira, AC Costa, J Delgado, G Del Barrio, A Soares, 2010, Spatial–temporal dynamics of precipitation extremes in southern Portugal: a geostatistical assessment study. International Journal of Climatology 30 (10), 1526-1537
i) Homogenised time series of climate data, publicly available;
ii) Spatial-temporal patterns of main droughts indicators;
ii) Identified most critical areas with significative local impacts of droughts, conditioned by its land use;
iii) A tested methodological outline for short term prediction of drought risk.Previous studies [1], [2] and [3] will be accounted for bench mark of the evolution of the phenomenon.