🌧️ Acid Rain Predictive Analytics Pipeline Across India

Applied Data Science & Machine Learning Research
Developed during the Edunet Foundation data analysis internship, analyzing over 50+ environmental datasets (~300,000+ records) to model precipitation pH trends, identify industrial pollution hotspots, and predict acid rain vulnerabilities across India.

---

📌 Executive Summary

Atmospheric deposition of acidic precursors ($SO_2$, $NO_x$, $SO_4^{2-}$, $NO_3^-$) severely impacts soil chemistry, agricultural yields, aquatic ecosystems, and architectural heritage. This repository delivers an end-to-end data science pipeline that:

1. Aggregates & Cleans multi-year historical air quality and chemical precipitation metrics.
2. Performs Rigorous Feature Engineering incorporating meteorological factors, atmospheric ion ratios, and seasonal fluctuations.
3. Trains Supervised Predictive Regressors to forecast precipitation pH levels with high fidelity.
4. Applies Unsupervised Hotspot Clustering to classify geographical regions into high-, moderate-, and low-risk environmental degradation zones.

---

📊 Dataset Architecture & Features

The modeling pipeline ingests atmospheric telemetry and rainfall chemistry data containing over 30,000 cleaned records:

Feature Category	Variables	Description
Gaseous Precursors	$SO_2$, $NO_x$, $NO_2$	Industrial emissions driving sulfurous and nitric acid formation
Particulate Telemetry	$PM_{2.5}$, $PM_{10}$, RSPM, SPM	Fine particulate matter acting as airborne reaction surfaces
Chemical Ion Deposition	Sulfate ($SO_4^{2-}$), Nitrate ($NO_3^-$), Ammonium ($NH_4^+$), Calcium ($Ca^{2+}$), Chloride ($Cl^-$)	Dissolved ionic components dictating neutralizing and acidifying potential
Physical Water Quality	Electrical Conductivity, Total Dissolved Solids (TDS)	Aqueous ionic mobility and mineral contamination
Spatial & Temporal	Monitoring Station, City, State, Season, Month	Geographic coordinates and seasonal monsoon cycles
Target Variable	Precipitation pH	Continuous acidity scale ($pH < 5.6$ denotes acidic precipitation)

---

🔬 Modeling Architecture & Results

1. Exploratory Data Analysis & Chemical Balance

• Correlation heatmaps evaluating ion-pair affinities ($SO_4^{2-}$ vs. $NO_3^-$).
• Neutralization factor analysis assessing alkaline dust ($Ca^{2+}$, $Mg^{2+}$) buffering acid precipitation.
• Seasonal pH distribution boxplots identifying pre-monsoon vs. post-monsoon acidity swings.

2. Supervised Regression Models

• Linear Regression (Baseline): $R^2 \approx 0.70$
• Random Forest Regressor: $R^2 \approx 0.88$ (capturing non-linear interactions between atmospheric humidity, particulate loading, and ion concentrations).
• Feature Importance: Sulfate ($SO_4^{2-}$) and Nitrate ($NO_3^-$) identified as the primary drivers of low pH, counterbalanced by Calcium ion concentration.

3. Unsupervised Spatial Clustering

• K-Means Clustering ($k=4$): Segregated monitoring zones based on industrial emission profiles and resulting pH severity, identifying distinct geographical clusters requiring emission intervention.

---

🚀 Setup & Reproducible Execution

Prerequisites

• Python 3.9 or higher
• Git

1. Clone & Set Up Environment

# Clone the repository
git clone https://github.com/Knight6azer/Edunet.git
cd Edunet

# Create and activate virtual environment
python -m venv venv
# On Windows:
venv\Scripts\activate
# On Linux/macOS:
source venv/bin/activate

# Install dependencies
pip install -r requirements.txt

2. Launch Analysis Notebook

jupyter notebook Project.ipynb

Execute cells sequentially to reproduce the data preprocessing, exploratory statistical visualizations, regression modeling, and clustering analysis.

---

🛠️ Tech Stack & Tooling

• Data Processing: pandas, numpy, scipy
• Machine Learning: scikit-learn (Random Forest, Linear Regression, KMeans, StandardScaler)
• Data Visualization: matplotlib, seaborn
• Environment: jupyter, IPython

---

📄 License

This research and analytics project is distributed under the MIT License — see the LICENSE file for details.