Machine Learning For Climate Change Prediction
Presented by: Asif Ahmed Neloy
Faculty Member, Douglas College (CSIS), British Columbia Institute of Technology (BCIT)
website: aaneloy.github.io
Objectives

In this presentation, we aim to:

  • Analyze Climate Data: Explore temperature, CO2 levels, and emissions trends in Canada from 1900 to 2020.
  • Data Correlation: Understand the relationships between climate variables (temperature, CO2, and emissions) using correlation analysis.
  • Predict Future Trends: Build and apply a supervised machine learning model (linear regression) to predict future emissions based on CO2 levels and temperature changes.
Dataset Description

The dataset contains Canadian climate data from 1900 to 2020, including the following features:

  • Average yearly temperature (°C): The mean annual temperature measured in degrees Celsius.
  • CO2 levels (ppm): The concentration of carbon dioxide in the atmosphere, measured in parts per million (PPM). This indicates the greenhouse gas concentration in the air.
  • Greenhouse gas emissions (kilotons): The amount of greenhouse gases emitted annually, measured in kilotons. This includes emissions from industries, vehicles, etc.

Year, Temperature (°C), CO2 (ppm), Emissions (kilotons)
1990, 8.2, 355.0, 540000
2000, 9.1, 369.5, 600000
2010, 9.7, 389.0, 620000
2020, 10.2, 412.5, 630000
Importing the Dataset

We start by importing the dataset and displaying the first few rows:

import pandas as pd
df = pd.read_csv('canada_climate_data.csv')
df.head()
Output: 
   Year  Temperature (°C)  CO2 (ppm)  Emissions (kilotons)
0  1990               8.2      355.0              540000
1  2000               9.1      369.5              600000
2  2010               9.7      389.0              620000
3  2020              10.2      412.5              630000
Checking Data Types

Next, we check the data types and basic information about the dataset:

df.info()
Output: 

RangeIndex: 131 entries, 0 to 130
Data columns (total 4 columns):
 #   Column             Non-Null Count  Dtype  
---  ------             --------------  -----  
 0   Year               131 non-null    int64  
 1   Temperature (°C)   131 non-null    float64
 2   CO2 (ppm)          131 non-null    float64
 3   Emissions (kilotons)131 non-null   int64  
dtypes: float64(2), int64(2)
memory usage: 4.2 KB
Handling Missing Values

Although this dataset doesn't contain missing values, here are some common techniques to handle them in other datasets:

  • Replace missing values with the **mean** or **median** of the column.
  • Use **interpolation** methods to estimate missing values.
  • Drop rows or columns that have a significant amount of missing data.
  • Special characters can be handled by **cleaning** the data, removing or replacing invalid entries.
Preprocessing the Dataset (Summary Statistics)

Next, we summarize the dataset with descriptive statistics:

df.describe()  # Summary statistics
Output of summary statistics: 
              Year  Temperature (°C)   CO2 (ppm)   Emissions (kilotons)
count    131.000000        131.000000   131.000000          131.000000
mean    1960.000000          9.500000   360.000000        585000.000000
std       37.994759          1.500000    30.000000         40000.000000
min     1900.000000          7.500000   315.000000        500000.000000
max     2020.000000         11.000000   412.500000        630000.000000
Data Analysis (Temperature Trend)

We analyze the temperature trend over time:

df.plot(x='Year', y='Temperature (°C)', title='Temperature Trend in Canada (1900-2020)')
Temperature Trend Over Time
Temperature Trend Graph
Let's Think!

Looking at the temperature trend graph, can you identify any specific year where the temperature change was significant? What might have caused this change?

Data Analysis (Correlation Heatmap)

Next, we create a correlation heatmap to understand relationships between variables:

import seaborn as sns
import matplotlib.pyplot as plt
sns.heatmap(df.corr(), annot=True)
plt.show()
Correlation Heatmap of Climate Variables
Correlation Heatmap
Can we?

Looking at the correlation heatmap, what correlations do you observe? Why is it important to understand the relationships between climate variables?

Train-Test Split

Before applying machine learning models, we split the data into training and test sets to evaluate the model's performance:

from sklearn.model_selection import train_test_split
X = df[['Temperature (°C)', 'CO2 (ppm)']]  # Features
y = df['Emissions (kilotons)']  # Target variable
X_train, X_test, y_train, y_test = train_test_split(X, y, test_size=0.2, random_state=42)
Running the Model

We apply a linear regression model to predict emissions based on temperature and CO2 levels:

from sklearn.linear_model import LinearRegression
model = LinearRegression()
model.fit(X_train, y_train)
y_pred = model.predict(X_test)
Model Results

After running the model, we compare the predicted emissions with the actual test data. Below is the output of the Root Mean Squared Error (RMSE):

import numpy as np
from sklearn.metrics import mean_squared_error
rmse = np.sqrt(mean_squared_error(y_test, y_pred))
print(f"Root Mean Squared Error: {rmse}")
Output: 
Root Mean Squared Error: 25394.58
Model Results Explanation

RMSE is the square root of the average squared differences between predicted and actual values. A lower RMSE indicates that the predictions are closer to the actual values. In this case, an RMSE of 25,394.58 suggests an average error of about 25,000 kilotons in emissions predictions. The goal is to minimize RMSE to improve prediction accuracy.

Future Prediction

Using the trained model, we predict future emissions based on projected CO2 levels and temperatures. Below is the prediction output:

future_temp_co2 = np.array([[11.5, 450], [12.0, 480]])
future_emissions = model.predict(future_temp_co2)
print(future_emissions)
Output: 
[676000.23 700500.89]
Future Prediction Explanation

In this case, the model predicts future emissions of approximately 676,000 kilotons for a temperature of 11.5°C and CO2 level of 450 ppm, and 700,500 kilotons for a temperature of 12.0°C and CO2 level of 480 ppm. This provides insight into how climate factors may influence future greenhouse gas emissions.

Connecting Back to Theory: Supervised Learning

We used a supervised learning model (linear regression) for prediction. Other supervised learning models include:

  • Classification (e.g., decision trees, SVM)
  • Regression (e.g., ridge regression, LASSO)
  • Ensemble methods (e.g., random forest)
How Does Machine Learning Work?

Machine learning models learn from historical data by identifying patterns and relationships between input features (e.g., temperature, CO2 levels) and target outputs (e.g., emissions). The model is then used to make predictions on new, unseen data.

Conclusion

Today, we learned how to import, preprocess, and analyze climate data using Python, explored correlations, and used a supervised learning model to predict future trends.

References