Hi, my name is Ivana Aleksovska. I'm a applied mathematician/statistician interested in many different subjects. I'm not new to Wikipedia, but due to the difficulty of accessing my old wiki account, I'm creating this new page to continue contributing to the Wikipedia community.

Education

Undergraduate studies: I did my undergraduate studies in Engineering in Mathematical Programming, Faculty of Natural Science&Maths, Applied Mathematics (stochastic programming), University of Ss. Cyril and Methodius University in Skopje, N.Macedonia.

Master: I did two Master 2 thesis.
- "An implementation of the Informational Approach to Global Optimization (IAGO) using the Small (Matlab/Octave) Toolbox for Kriging", Master program in Modeling, Simulation and Optimization, UVSQ University of Versailles Saint-Quentin-en-Yvelines, France.
- "Multivariate Regression Analysis", Master's degree at Faculty of Electrical Engineering and Information technology, Applied mathematics in the field of electrical engineering and information technology, University of Ss. Cyril and Methodius University in Skopje, N.Macedonia.

PhD: I hold a PhD degree, on the following topic: "Improve short- and medium-term predictions of agronomic models by better taking into account the uncertainty of weather forecasts." I worked on a PhD project in a collaboration with Météo-France and CNRM (National Centre for Meteorological Research) and INRAE (Research for Agriculture, Food and Environment). We illustrated the potential of using ensemble weather forecasts in agronomic models compared to frequency data. We then propose strategies to design seamless ensemble weather forecasts that combine information from different numerical weather prediction systems. Additional sensitivity analysis work was carried out to understand whether the uncertainty in the results on estimated phytosanitary dates stems from the uncertainty in the weather forecast input data, or from the uncertainty in the model parameter data and other data related to agronomic parameters. The Morris screening technique and the Sobol indices were very useful in meeting this challenge. Meteorological data were considered as functional data for this problem, and model parameter data were considered as uncertain parameters with an appropriate distribution provided by experts in the field.

Areas I worked on

I worked for renewable energy sector (RES). RES is highly dependent on weather conditions and the reliability of weather forecasts has a direct impact on energy production. Weather forecasts is a major input in various decision support tools, and help energy suppliers to make optimal decisions and avoid potential losses. In this work, we proposed seamless scenarios for 100m wind speed, combining global and regional ensemble prediction systems (EPS), Arpège-EPS and Arome-EPS respectively, from Météo-France, the French National weather agency. Seamless ensemble takes advantage of the increased performance of high-resolution EPS for short lead times, while ensuring a smooth transition to larger-scale EPS for longer lead times and moreover, it provides enhanced performances with respect to the reference EPS, at least at short ranges. In this work, I have proposed a quasi-operational R software, communicating with a database, run the Hungarian method and provides ensemble wind speed forecasts (+score verification on the fly), ready to be used for renewable energy simulations.

I worked as a Scientist Uncertainty Quantification for Destination Earth project at European Centre for Medium-Range Weather Forecasts. I have proposed an alternative way of quantifying the uncertainty of the DestinE deterministic continuous forecast, using a statistical models, called also a metamodels, with post-processing technique, know as EMOS, that improves the skill and reliability of the forecast, while providing uncertainty quantification at station locations over across the globe. I have proposed a quasi-operational Python software to train EMOS models univariately (for each weather variable, location and lead time) using operational ensembles (ENS) and DestineE km-scale continuous forecasts for temperature at 2m (Gaussian distribution) and wind speed at 10m (left-truncated Gaussian distribution at zero) and generate data-driven forecast that give the best performance scores and provides a substantial improvement over the raw ENS, and DestinE ensembles, according to CRPS and reliability scores.

Scientific Interest

Ensemble forecasting, uncertainty analysis, machine learning, post-processing, EMOS, Gaussian, truncated Gaussian, meteograms, Python, parallel computing, Seamless forecast, Uncertainty analysis, Sensitivity analysis, Hungarian method, Wind power, Renewable energy, R, shell, bash, Linux, VM. Propagation uncertainty, Decision support tools DST, Agronomy, Crop protection, Vine, Wheat, BDD, Grib files.

Other interest

Languages, travel, different cultures, NGO, sport...

Further links

My LinkedIn profile: https://www.linkedin.com/in/ivana-aleksovska-ph-d-19653964/

My GitHub: https://github.com/IvanaAlexML/

Articles I worked on:

https://en.wikipedia.org/wiki/Sensitivity_analysis