|
| Titre : |
A Deep Model For Forest Fire Risk Estimation In Algeria |
| Type de document : |
document électronique |
| Auteurs : |
Mohamed Islam Boussahel ; Harbouche,Khadidja, Directeur de thèse |
| Editeur : |
Setif:UFA |
| Année de publication : |
2025 |
| Importance : |
1 vol (57 f .) |
| Format : |
29 cm |
| Langues : |
Anglais (eng) |
| Catégories : |
Thèses & Mémoires:Informatique
|
| Mots-clés : |
Algeria
Wildfires
Prediction
MODIS
Long Short Term Memory (LSTM)
Logistic Regression
Support Vector machines (SVM)
Random Forest
XGBoost |
| Index. décimale : |
004 Informatique |
| Résumé : |
Algeria is one of the most affected Mediterranean countries by wildfires, often leading to
devastating consequences. However, unlike its neighbors (Spain, Portugal, Greece, and Italy)
it has not yet developed a sophisticated and reliable wildfire prevention system.
In this study, we are going to use MODIS Collection 6.1 fire data, GLOH2O weather data,
elevation data from DIVA-GIS, Land Cover Map from ESA, and the study area’s shape file
from GADM to estimate wildfire risk. Our approach relies on the application of different
algorithms (Logistic Regression, Support Vector Machines (SVM) with an RBF kernel, long
short term memory networks (LSTMs), Random Forest, and XGBoost) to predict wildfire
occurrences based on the available data.
The study aims to find the optimal combination of predictor model and preprocessing
techniques that gives the most accurate wildfire predictions. |
| Note de contenu : |
Sommaire
General introduction
1 Introduction to Forest Fires in Algeria 1
1.1 Climate in Algeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Northern Algeria: Mediterranean Climate and Lush Vegetation . . . 2
1.1.2 The Highland: Steppe Vegetation in a Semi-Arid Transition . . . . . 2
1.1.3 The Sahara Desert: Rare Vegetation and Extremely Arid Area . . . . 3
1.2 Algerian forests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 History of Algerian wildfires . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Current Techniques for Dealing with Wildfires . . . . . . . . . . . . . . . . . 5
1.4.1 Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4.2 Pre-Suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.3 Suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Advantages of Deep Learning and Machine Learning in the Fire Prevention
Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.1 Enhanced Predictive Accuracy . . . . . . . . . . . . . . . . . . . . . 7
1.5.2 Ability to Handle High-Dimensional and Heterogeneous Data . . . . 8
1.5.3 Adaptability to Local Conditions and Robustness to Uncertainty and
Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.4 Support for Decision Making and Policy Planning . . . . . . . . . . 8
2 Artificial Intelligence (Deep Learning and Machine Learning) 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Exploring different machine learning (ML) Algorithms . . . . . . . . . . . . 12
2.2.1 Logistic Regression: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Support Vector Machines (SVM) . . . . . . . . . . . . . . . . . . . . 13
2.2.3 Random Forest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.4 XGBoost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Exploring Deep Learning (DL) and Recurrent Neural Networks (RNN) . . . 17
2.3.1 Recurrent Neural Networks (RNN) and Long-Short Term Memory time
series (LSTMs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Harnessing Tools and Acquiring the Data 22
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Used Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 QGIS Desktop 3.40.3 Bratislava . . . . . . . . . . . . . . . . . . . . . 23
3.2.2 Python Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 MODIS C6.1 NASA FIRMS Fire Archive . . . . . . . . . . . . . . . . 27
3.3.2 GLOH2O weather data . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.3 ESA WorldCover Land Cover Map . . . . . . . . . . . . . . . . . . . 29
3.3.4 DIVA-GIS Elevation Data . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.5 Data Preprocessing and Integration . . . . . . . . . . . . . . . . . . . 30
4 Implementation of the model and results 35
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.1 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.3 Recall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.4 F1 Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3 Training the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4 Testing the model on unseen data . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Running our model under real conditions . . . . . . . . . . . . . . . . . . . . 55
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
General Conclusion |
| Côte titre : |
MAI/1030 |
A Deep Model For Forest Fire Risk Estimation In Algeria [document électronique] / Mohamed Islam Boussahel ; Harbouche,Khadidja, Directeur de thèse . - [S.l.] : Setif:UFA, 2025 . - 1 vol (57 f .) ; 29 cm. Langues : Anglais ( eng)
| Catégories : |
Thèses & Mémoires:Informatique
|
| Mots-clés : |
Algeria
Wildfires
Prediction
MODIS
Long Short Term Memory (LSTM)
Logistic Regression
Support Vector machines (SVM)
Random Forest
XGBoost |
| Index. décimale : |
004 Informatique |
| Résumé : |
Algeria is one of the most affected Mediterranean countries by wildfires, often leading to
devastating consequences. However, unlike its neighbors (Spain, Portugal, Greece, and Italy)
it has not yet developed a sophisticated and reliable wildfire prevention system.
In this study, we are going to use MODIS Collection 6.1 fire data, GLOH2O weather data,
elevation data from DIVA-GIS, Land Cover Map from ESA, and the study area’s shape file
from GADM to estimate wildfire risk. Our approach relies on the application of different
algorithms (Logistic Regression, Support Vector Machines (SVM) with an RBF kernel, long
short term memory networks (LSTMs), Random Forest, and XGBoost) to predict wildfire
occurrences based on the available data.
The study aims to find the optimal combination of predictor model and preprocessing
techniques that gives the most accurate wildfire predictions. |
| Note de contenu : |
Sommaire
General introduction
1 Introduction to Forest Fires in Algeria 1
1.1 Climate in Algeria . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.1.1 Northern Algeria: Mediterranean Climate and Lush Vegetation . . . 2
1.1.2 The Highland: Steppe Vegetation in a Semi-Arid Transition . . . . . 2
1.1.3 The Sahara Desert: Rare Vegetation and Extremely Arid Area . . . . 3
1.2 Algerian forests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 History of Algerian wildfires . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.4 Current Techniques for Dealing with Wildfires . . . . . . . . . . . . . . . . . 5
1.4.1 Prevention . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4.2 Pre-Suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.4.3 Suppression . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5 Advantages of Deep Learning and Machine Learning in the Fire Prevention
Process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.5.1 Enhanced Predictive Accuracy . . . . . . . . . . . . . . . . . . . . . 7
1.5.2 Ability to Handle High-Dimensional and Heterogeneous Data . . . . 8
1.5.3 Adaptability to Local Conditions and Robustness to Uncertainty and
Noise . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
1.5.4 Support for Decision Making and Policy Planning . . . . . . . . . . 8
2 Artificial Intelligence (Deep Learning and Machine Learning) 10
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2.2 Exploring different machine learning (ML) Algorithms . . . . . . . . . . . . 12
2.2.1 Logistic Regression: . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2.2.2 Support Vector Machines (SVM) . . . . . . . . . . . . . . . . . . . . 13
2.2.3 Random Forest . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 15
2.2.4 XGBoost . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
2.3 Exploring Deep Learning (DL) and Recurrent Neural Networks (RNN) . . . 17
2.3.1 Recurrent Neural Networks (RNN) and Long-Short Term Memory time
series (LSTMs) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
3 Harnessing Tools and Acquiring the Data 22
3.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2 Used Tools . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
3.2.1 QGIS Desktop 3.40.3 Bratislava . . . . . . . . . . . . . . . . . . . . . 23
3.2.2 Python Libraries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25
3.3 Data acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
3.3.1 MODIS C6.1 NASA FIRMS Fire Archive . . . . . . . . . . . . . . . . 27
3.3.2 GLOH2O weather data . . . . . . . . . . . . . . . . . . . . . . . . . . 28
3.3.3 ESA WorldCover Land Cover Map . . . . . . . . . . . . . . . . . . . 29
3.3.4 DIVA-GIS Elevation Data . . . . . . . . . . . . . . . . . . . . . . . . 30
3.3.5 Data Preprocessing and Integration . . . . . . . . . . . . . . . . . . . 30
4 Implementation of the model and results 35
4.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2 Performance Metrics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
4.2.1 Accuracy . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.2 Precision . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.3 Recall . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.2.4 F1 Score . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
4.3 Training the model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37
4.4 Testing the model on unseen data . . . . . . . . . . . . . . . . . . . . . . . . 51
4.5 Running our model under real conditions . . . . . . . . . . . . . . . . . . . . 55
4.6 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 57
General Conclusion |
| Côte titre : |
MAI/1030 |
|
Ajouter le résultat dans votre panier Affiner la recherche
