omarmaalej12/predicting-the-Future-Evolution-of-Vegetation-Cover-in-Tunisia-Using-Advanced-Machine-Learning

predicting the Future Evolution of Vegetation Cover in Tunisia Using Advanced Machine Learning Methods # 🌱 Vegetation Cover Prediction in Tunisia under Climate Change Scenarios This repository contains the code and resources developed during a research internship at the Laboratory of Ecology, Systematics, and Evolution (ESE) – Université Paris-Saclay. ## 🧠 Objective The goal of this project is to predict the **future evolution of vegetation cover in Tunisia** under climate change (SSP245 scenario), using advanced **machine learning techniques** and large-scale **geospatial and bioclimatic data**. --- ## 📦 Key Data Sources and Preprocessing ### 🌍 Vegetation and Climate Data - **Vegetation data** comes from **GLAD 2019 maps** (Global Land Analysis and Discovery Lab, University of Maryland), based on **Landsat imagery** with a **30m spatial resolution**. Each pixel is labeled with one of 20 land cover classes (0 to 19). For this project, only the two tiles covering Tunisia (`40N_010E` and `40N_000E`) were used and processed via QGIS. - **Climate data** (BIO1–BIO19) from WorldClim was used for both the current period (1970–2000) and future projections (2021–2040) under **SSP245** scenario. ### 🗺️ Geographic Processing with QGIS Using **QGIS**, we carried out the following: - Extracted the **geographic boundary of Tunisia**. - Generated a **1km² resolution pixel grid** over the country. - Overlaid and fused the **vegetation maps** using zonal statistics ("majority" rule). - Assigned the **dominant vegetation class** to each 1km² grid cell. - Aligned 19 climatic variables (BIO1–BIO19) with each pixel. - Produced two final datasets (Excel/CSV) with: - Coordinates (lat, lon) - Dominant vegetation class - 19 bioclimatic variables These datasets served as the inputs for all machine learning models. --- ## 📊 Workflow Overview ### 1. **Data Processing** - Removal of missing values. - Dimensionality reduction via **correlation analysis**. - **Standardization** of selected variables using z-score scaling. - Aggregation of 20 original vegetation classes into 3 simplif …