If the World Would Limit Carbon Emissions Would It Prevent Any Future Global Warming

The image above shows the current prediction until 2100 of global warming without mitigating carbon emissions at all. Using Keras and an LSTM model I was able to make the forecast prediction.

Introduction: Since the 1840s, we’ve seen a steady increase in global temperatures and there have been many theories that look at what may be to contributing. More specifically how we can solve this Global Crisis that could make millions of people around the globe lose their homes. This is so important for humankind as a whole, the earth will be around no matter if we evolve to protect our environment or continue to ignore the inevitable and become another failed civilization. The data was collected by Berkely Earth (a independent california based nonprofit that focuses on data of global temperatures).

Image released by NASA back in 2016 showing reduction in polar ice caps leading to more Co2 in Earth’s atmosphere

Greenhouse Gasses: Around 95% of scientists agree that the current global warming is being caused by human intervention because in climate history we have never seen this great of an increase in this short amount of time. More importantly, they have targeted greenhouse gasses to cause this global warming and reducing the size of polar ice caps leading to more carbon dioxide in the atmosphere. This is concerning the increase not only has to produce rising sea levels but additional more violent weather patterns and violent droughts that have led to various forest fires around the globe. At the dust of the Industrial Revolution, the atmospheric co2 concentration measured (278 Parts Per Million). Today we are at a concentration (414 Parts Per Million) and reach (516 Parts Per Million) by 2080 at the current rate. That's about twice pre the Industrial Revolution concentration levels.

Slightly Symmetric Distribution

Distribution: When we dive deeper into the distribution of decade average temperatures we can see that it hovered around 8–8.5 Degrees Celsius between (1760–1920) and more recently has crept its way up to 9.5 recently in 2010.

Baseline Models: When producing the first model I used a linear regression model to get my first baseline and used r-squared scores to evaluate the model.

Baseline Model Two Xgboost(XGBRegressor): Which had a reasonable accuracy and was significantly better than the baseline model. Besides, the predictions for the test set were significantly off by more than .5 degrees Celcius.

Final Model RandomForestRegressor: Then by using a grid and hyperparameter turning by using RandomizedSearchCV I was able to produce a model that made predictions only off by 0.12 degrees Celcius. Using model I could accurately look at the Carbon Emissions Features Importance.

Evaluate Final Model: Using an input of a special row I could see how well it could predict my target(Decade Average Temperatures)

From a random row in 1770 August the predictor was only off by .04 degrees celsius

Using Shap we can see that all Carbon Emission Columns were near zero in this row so it predicted the temperature lower than the base value and accurately

Feature Importance: Using SHAP Summary we can see how important each feature is in predicting the temperature for that date. Moreover, we find columns that aren’t as important. The average temperature for that month column(Averahe_Temps) didn’t help figure out the decade average only carbon emissions columns did.

Conclusion: Now for seeing if mitigating carbon emissions would result in the future global cooling effect and reducing the greenhouse effect(the elivation of co2 in earths atomostphere). Using the function below I was able to make unique rows of data to input into the predictor to see if global temperatures would reside to the baseline(8.5).

Zero Carbon Emissions by 2100 would, in theory, reduce decade global temps to or below a baseline of (8.5).

Continuation of current carbon footprint by 2100 results in an increase of around a degree celsius.

My Notebook with datasets:

https://github.com/John-G-Thomas/Projects-data/blob/master/Project_2.ipynb