Loading...

Development of Precipitation Simulating Model using Statistical-Probabilistic Models and Deep Learning Algorithms

Kadkhodaei, Kianoosh | 2025

29 Viewed
  1. Type of Document: M.Sc. Thesis
  2. Language: Farsi
  3. Document No: 58672 (09)
  4. University: Sharif University of Technology
  5. Department: Civil Engineering
  6. Advisor(s): Moghim, Sanaz
  7. Abstract:
  8. Throughout history, rainfall has been one of the vital factors influencing human life. Predicting the quantity, intensity, extent, duration, and probability of rainfall holds considerable importance in fields such as agriculture, water resource management, urban planning, and disaster preparedness. Although rainfall is inherently random and influenced by multiple factors, accurate and reliable precipitation forecasts enhance proactive decision-making, facilitate optimal resource allocation, reduce risks, and promote sustainable development. The aim of this study is to develop a rainfall simulation model by integrating probabilistic statistical models, machine learning algorithms, and satellite data. The data used in this research include satellite and atmospheric model outputs, cloud-related features, as well as station-based rainfall observations. The primary objective is to improve the accuracy and reliability of rainfall simulation through a multifaceted approach. The proposed method involves identifying the key variables that influence monthly rainfall. These variables are determined using feature selection algorithms and subsequently incorporated into the model. The main rainfall simulation framework is essentially a combination of a time-series model and a machine learning model. Rainfall data are decomposed into cyclical and trend components using filters: the machine learning model is responsible for simulating the cyclical component of monthly rainfall, while the time-series model is used to simulate the trend component. Analyses revealed that correlations between various variables and rainfall differ across seasons, indicating the presence of distinct dominant rainfall mechanisms throughout the year. In addition, different feature selection algorithms identified various key variables—including climatic factors and cloud-related variables—highlighting the importance of both categories. Overall, the autoregressive model performed relatively better than the other models used for predicting the linear component of rainfall. Ultimately, the best-developed model in this study achieved a rainfall simulation accuracy of 0.71
  9. Keywords:
  10. Rainfall Simulation ; Machine Learning ; Remote Sensing ; Statistical Model ; Water Resources Management ; Probabilistic-Statistical Models

 Digital Object List

 Bookmark

No TOC