Logo Lanfrica

RGuisgo/African-Climate-Forecasting-R

Domaine:

climate
Créateur:
RGu
Hôte:
Forecasts yearly temperature and rainfall trends in ten African countries using Bayesian Autoregressive and Random Walk models in R. Includes model comparison, visualization, and projections with uncertainty intervals. # African Climate Trend Analysis & Forecasting – Bayesian Models (R) ## 📌 Overview This project examines trends and forecasts of yearly **temperature and rainfall** in ten African countries across West, South, East, and North Africa (Ghana, Nigeria, Senegal, South Africa, Namibia, Rwanda, Ethiopia, Madagascar, Egypt, Tunisia) using **Bayesian Autoregressive (AR(1))** and **Random Walk (RW(1))** models. - **Objective:** Compare model performance and generate projections for temperature and rainfall. - **Models Used:** Bayesian Autoregressive (AR(1)) and Random Walk (RW(1)) - **Evaluation Metrics:** Deviance Information Criterion (DIC) and Watanabe-Akaike Information Criterion (WAIC) --- ## 🔍 Key Findings - **Temperature Data:** Random Walk model provided better fit across all countries. - **Rainfall Data:** Autoregressive model outperformed Random Walk model. - **Forecasting Patterns:** - AR model forecasts indicate a **downward trend** in temperature for most countries. - RW model forecasts align with recent trends but have **wider credible intervals**, reflecting higher uncertainty. - **Implications:** Temperature is expected to decrease, while rainfall trends largely follow historical observations. --- ## 📁 Repository Contents - **Thesis Document:** Main PDF with full methodology, results, and discussion. - **Supplementary Materials:** - Datasets used in the analysis - R scripts for data processing and model implementation - Visualizations and charts --- ## 🛠️ Tools - **Language:** R - **Packages:** tidyverse, coda, rstan, ggplot2 --- ## ▶️ How to Run 1. Open the relevant R scripts in RStudio. 2. Execute scripts sequentially to reproduce data cleaning, model estimation, and visualization. 3. Output includes forecast plots, credible intervals, and model comparison metrics. --- ## 📌 Citation If you reference or use this work, please cite as: **Rebecca Efua Guisgo**, *"Trend analysis and forecasting of yearly temperature and rainfall in some African count …