{"id":873,"date":"2024-10-14T18:01:47","date_gmt":"2024-10-14T18:01:47","guid":{"rendered":"https:\/\/krzysztofraczynski.net\/?page_id=873"},"modified":"2025-02-24T14:45:36","modified_gmt":"2025-02-24T14:45:36","slug":"research","status":"publish","type":"page","link":"https:\/\/krzysztofraczynski.net\/index.php\/research\/","title":{"rendered":"Research overview"},"content":{"rendered":"\n

Research overview<\/h2>\n\n\n\n

My work embodies a multidisciplinary approach to understanding and mitigating the impacts of hydrologic extremes. Leveraging a diverse skill set, a commitment to innovation, and a passion for collaboration, I strive to advance both scientific understanding and practical solutions for addressing one of the most pressing challenges of our time.<\/p>\n\n\n\n

HOST model of extremes distribution<\/h2>\n\n\n\n

The Harmonic Oscillator Seasonal-Trend (HOST) model represents a significant advancement in time-series analysis for understanding hydrological extremes, particularly droughts. This Python-based tool combines advanced statistical methods and harmonic waveform synthesis to decompose data into short- and long-term patterns, providing insights into trends and seasonal variability. By utilizing a novel framework built on Fast Fourier Transform (FFT) and machine learning, the HOST model identifies temporal patterns in streamflow drought occurrence with impressive accuracy, explaining over 98% of long-term trends and approximately 55% of seasonal variance in tested datasets. Its ability to quantify periodicity, such as annual precipitation cycles and multi-year inertial patterns, makes it a powerful tool for diagnosing the drivers of extreme events and informing sustainable water resource management strategies.<\/p>\n\n\n\n

Tested on daily and monthly streamflow data from the National Water Model (NWM) retrospective dataset, the HOST model demonstrated robust performance, with a Kling-Gupta efficiency often exceeding 0.9 for long-term trends. While its accuracy in seasonal patterns and binary occurrence classification showed limitations in regions with sparse drought events, the model’s weighted occurrence classification and automatic decomposition improvements have reduced false-positive forecasts. This innovative approach provides actionable insights for hydrological planning and emergency response, particularly in regions vulnerable to the increasing frequency of droughts due to climate change. The HOST model’s ability to integrate with physical models further enhances its applicability for short-term event forecasting and long-term water management solutions. Model is published as a Python module (hostmodel<\/a><\/code>).<\/p>\n\n\n

\n\t
<\/div>\n\t
<\/div>\n\t
<\/div>\n\t
<\/div><\/div>