Dust Storm Prediction Using Artificial Intelligent Techniques (A Case Study: Zabol City)

Jamalizadeh Tajabadi, Mohammad Reza (2008) Dust Storm Prediction Using Artificial Intelligent Techniques (A Case Study: Zabol City). Masters thesis, University of Zabol.

[img]
Preview
Text
Abstract - cutted.pdf

Download (6kB) | Preview

Abstract

Arid Regions covered about 40% of world and contain many people. Dust storms are most common natural events in these areas that threaten no health and fund in these areas but the entire world. Recently, artificial intelligence techniques become more usable in many fields and their ability in modeling of complicated processes that are not well explored. In present study two artificial intelligence techniques utilized for dust storm prediction in Zabol Area. A 25-year period of climatic records was used. We classified dust events in two classes based on visibility. Doing Gamma Test indicate best arrangement of inputs for model construction. Models have used selected parameters and their efficiencies compared via contingency table. Finally, uncertainty and sensitivity analysis are carried out. Results show that presented classification is stronger in all models and more severe events are more attainable in prediction. Support Vector Machines (SVMs) have best efficiency (CSI=0.52) and Artificial Neural Networks (ANNs), Local Linear Regression (LLR) and Stepwise Regression, respectively. Uncertainty and sensitivity are very high in SVMs, whereas ANNs are less sensitive and uncertain. However more attempts are necessary in dust storm identification.

Item Type: Thesis (Masters)
Uncontrolled Keywords: Dust Storm, Prediction, Artificial Intelligent, Artificial Neural Networks, Support Vector Machines
Subjects: G Geography. Anthropology. Recreation > GA Mathematical geography. Cartography
G Geography. Anthropology. Recreation > GE Environmental Sciences
Depositing User: admin admin1 admin2
Date Deposited: 21 May 2016 07:16
Last Modified: 21 May 2016 07:16
URI: http://eprints.uoz.ac.ir/id/eprint/613

Actions (login required)

View Item View Item