IoT-based climate change prediction system. Case study: Masoro Industrial economic zone

Abstract

Climate change is one of the most significant challenges to every country's development, causing uncountable ravages in the life of all living around the globe. An increase in carbon dioxide emission, one of the main greenhouse gases has been gradually recorded from burning fuels and natural gases in generating power and heat, manufacturing processes, and deforestation in general. Numerous research and studies of strategies for tracking climate change have been raised by researchers. In Rwanda, the existing climate change tracking method uses a weather station model, in which numerous fixed weather stations are installed around the nation, however, due to its immobility; this process can not cover the whole country. With the lack of advanced methodologies and technology, the process of climate change tracking has become extremely expensive and suffered inaccuracies due to a lack of proper knowledge of which parameters to collect, knowledge of analyzing collected data, and the lack of specific accurate hardware. Throughout this research, with the use of the MQ-135 sensor and DHT11 sensor, ESP8266 collects carbon dioxide gas and temperature/humidity respectively and other component include a push button for detecting the current season. Data from these sensors and push button are serially connected to it. With ESP built-in Wi-Fi capability, ESP8266 is programmed to send data over MQTT protocol which relies on Wi-Fi capability to send data to MQTT Broker which is hereby referred to as MQTT Box. With the Publish/Subscribe criteria of the MQTT protocol, node-red subscribes to the topics defined in MQTT broker to get data, which is sent to MongoDB for permanent storage, and also fed to the machine learning model for prediction of climate change/warming, this model is built by using Jupyter notebook which is a good tool for python users. This model is evaluated by using different machine learning classification algorithms for optimizing the prediction accuracy. Random Forest approves itself to be the best model in evaluating the built model with 99% of training accuracy and 84% of testing accuracy. The study shows that the increase in carbon dioxide gas leads to the gradual increase in the environmental temperature. Finally, the prediction clarifies that if no measures are taken presently, the climate change in Masoro's industrial zone will be dominated by warming periods in 2023.