IOT-Based Water Quality Monitoring for the Tigris River: Addressing Pollution Challenges

The Tigris River is a second-largest river in southwestern Asia, flowing through Turkey, Syria, Iraq, and Iran. The river is facing lack of water from the source and increasing pollution from agricultural runoff, industrial wastewater, and sewage. Water quality in the basin is primarily threatened by increased salinity rates caused by intensive irrigated agriculture and high evaporation rates. This paper aims to develop an IoT based and data mining system for monitoring the water quality of the Tigris River. The proposed system achieves real-time monitoring depending on the Internet of Things (IoT) to deter-mine water pH level, total dissolved solids (TDS), temperature, and turbidity or conductivi-ty. The acquired data to be sent online to a hosting server (SQL server) for collecting, storing, and analysis as well as for displaying it on a website The SQL server receives data from the ESP8266 Wi-Fi module (Node MCU). The MUC serves as the central hub, interfacing with sensors and coordinating data transfers to a PC for more analysis. In real-time, quickly compiled data is forwarded to a PHP custom-built website and also displayed on a map using ArcGIS for Power BI to simplify communicate with interested persons. Data mining techniques are used to identify patterns in the data and to detect changes in water quality as well as to predict areas of pollution in order to take corrective action by environment and water resources authorities. Decision Tree, Random Forest, Support Vector Machine, and Support Vector Classifier algorithms are used in this paper. By Using Python code, these methods are compared based on their accuracy rates. It was found that the model using the Random Forest method had the highest accuracy of 0.92, making it the best algorithm for classifying the water quality of the Tigris River in Baghdad, Iraq. Furthermore, due to the nature of these algorithms, Decision Tree has also good classification accuracy due to the nature of these algorithms, where Random Forest is progressively improved from Decision Tree.