Abstract
Among the most pressing issues in the world today is the impact of globalization and energy consumption on the environment. Despite the growing regulatory framework to prevent ecological degradation, sustainability continues to be a problem. Machine learning can help with the transition toward a net-zero carbon society. Substantial work has been done in this direction. Changing electrical systems, transportation, buildings, industry, and land use are all necessary to reduce greenhouse gas emissions. Considering the carbon footprint aspect of sustainability, this chapter provides a detailed overview of how machine learning can be applied to forge a path to ecological sustainability in each of these areas. The chapter highlights how various machine learning algorithms are used to increase the use of renewable energy, efficient transportation, and waste management systems to reduce the carbon footprint. The authors summarize the findings from the current research literature and conclude by providing a few future directions.
TopIntroduction
Human existence is inextricably intertwined with nature. Ecology by its inherent trait supplies humanity with vast natural resources. It aids in the sustenance of the huge living population and compensates for the ecological imbalances that surface repeatedly. Due to the rising global population, the consumption of nature’s wealth is ever-increasing. The surge in production of energy, food, goods, services, etc., to meet the demand and supply gaps has led to a colossal depletion of raw materials worldwide over the years. Overexploitation of the flora and fauna has endangered numerous species while putting several others in the high-risk category threatening the ecosystem’s equilibrium.
A sustainability metric for measuring human impact on Earth's ecosystems called the “Ecological Footprint” was proposed in the early 1990s by two Ph.D. researchers at the University of British Columbia (Wackernagel & Rees, 1996). Considering the dependency of humanity on the biosphere, Ecological Footprint (EF) is defined as a measure of the area necessary to sustain any given population. In its broadest sense, it is a measure that incorporates all forms of water and energy use, infrastructure, forest management, and other material inputs required by humans to flourish day in and day out, as well as accounting for the land devoted to waste assimilation. Ecological Footprint per capita is one of the most widely recognized indicators of environmental sustainability. Human society becomes unsustainable when its Ecological Footprint surpasses its biocapacity. Considering the sustainability of natural resources becomes essential due to the burgeoning demands of the growing population. Ecological Footprint and sustainability are emerging research areas that have grabbed the attention of contemporary researchers and policymakers.
Figure 1. Components of Ecological Footprint
The carbon component of the Ecological Footprint, highlighted in red in Figure 1, is also an indication of the amount of forest land that will be required to absorb the Greenhouse Gas (GHG) emissions from the burning of fossil fuels. The carbon footprint measures the amount of greenhouse gas released due to the consumption of fossil fuels excluding the fraction absorbed by the oceans. The amount of greenhouse gas emitted into the atmosphere when fossil fuels are burned contributes directly to an ecological footprint. As more greenhouse gases are released into the atmosphere, there will be a need for more sea and forest areas to remove them. Lacking the requisite sea and forest areas will increase the carbon footprint. A larger carbon footprint implies a more substantial ecological footprint. In this chapter, we examine the use of Machine Learning to address the problem of increasing carbon footprint in the context of ecological footprints and sustainability.
Machine Learning (ML) is a progressive technology that has the potential to offer practical solutions for environmental sustainability. Machine Learning has much to offer in terms of monitoring, analyzing, and resolving sustainability issues. Even with exceptional advancements in the field, the area continues to have a lot of scope for improvement. The discipline’s ever-expanding horizons still hold plenty of opportunities for solving challenging real-world problems. Artificial Intelligence (AI) and ML handle complex data, enabling data scientists to make accurate forecasts. These technologies can recommend comprehensive rational solutions that help in achieving sustainable development across the globe.
Prior research on Machine Learning applications in the sustainability domain is promising, and we believe that Machine Learning can support the development of culturally tailored organizational processes and individual responsibilities to cut down natural resources and energy consumption. We believe that ecological sustainability is the key to balancing the rising Ecological Footprint. Eventually, AI/ML will be highly valuable in contributing to environmental governance and not just limited to minimizing society's energy, water, and land usage intensities. In this survey, we study the applications of Machine Learning to analyze and predict the impact of the Ecological Footprint and to strategize carbon footprint reduction.
TopMethodology
We started by framing the following research questions:
Key Terms in this Chapter
KNN: - K-nearest neighbors
NWP: - Numerical Weather Prediction system
EVTREE: - Tree Models using Genetic Algorithms
KNNReg: - K-nearest neighbors Regression
ARIMA–BPNN: - Autoregressive Integrated Moving Average - Back Propagation Neural Network
MSW: - Municipal Solid Waste
GDP: - Gross Domestic Product
MASE: - Mean Absolute Scaled Error
SMAPE: - Symmetric Mean Absolute Percentage Error
VIP: - Variable Importance for Projection
GHG: - Green House Gas
NRMSE: - Normalized Root-mean-squared Error
CNN: - Convolutional Neural Network
ECMWF: - European Center for Medium Weather Forecasts
GPPoly: - Gaussian Process with Polynomial Kernel
LR: - Logistic Regression
RNN: - Recurrent Neural Network
NN: - Neural Network
CUB: - Cubist
SMOGN: - Synthetic Minority Over-Sampling Technique for Regression with Gaussian Noise
EV: - Electric Vehicle
GBRT: - Gradient Boosting Regression Tree
GPS: - Global Positioning System
GBR: - Gradient Boosting Regression
MLP: - Multi-Layer Perceptron
MLR: - Multiple Linear Regression
ILSVRC: - Image-Net Large-Scale Visual Recognition Challenge
ANNqr: - Quantile Regression with ANN
PLS: - Partial Least Squares
IL: - Imitation Learning
MetaFA: - Metaheuristic Firefly Algorithm
RL: - Reinforcement Learning
ERT: - Extremely Randomized Trees
SVM: - Support Vector Machine
XGB: - Extreme Gradient Boosting
AV: - Autonomous Vehicle
SVR: - Support Vector Regression
BPNN: - Back Propagation Neural Network
RFR: - Random Forest Regression
SWM: - Solid Waste Management
DP: - Dynamic Programming
MAPE: - Mean Absolute Percentage Error
DRL: - Deep Reinforcement Learning
PSO: - Particle Swarm Optimization
AI/ML: - Artificial Intelligence/Machine Learning
GRU: - Gated Recurrent Unit
ANN ReLU: - ANN Rectified Linear Unit
RF: - Random Forest
SARIMA: - Seasonal Autoregressive Integrated Moving Average
ANN SPOCU: - ANN Scaled Polynomial Constant Unit
MetaFA-LSSVR: - Metaheuristic Firefly Algorithm-based Least Squares Support Vector Regression
MDP: - Markov Decision Process