GCRO GIS Intern Earns Masters with Distinction

Phemelo Mahamuza, a dedicated GIS intern at the Gauteng City-Region Observatory (GCRO), joined the organization in April 2023. Since then, she has been actively involved in several impactful projects, including collaborating with the National Institute for Communicable Diseases (NICD) to monitor COVID-19 infections through human wastewater. Additionally, she has contributed to mining-related research within the Gauteng city region and led a personal project on the spatial validation of government infrastructure. Phemelo's passion for GIS and remote sensing and her commitment to research and environmental problem-solving positions her as a promising figure in her field.

During her internship, Phemelo also completed her Master of Science in GIS and Remote Sensing at the University of the Witwatersrand. Her master’s dissertation, titled "Assessing and Comparing the Performance of Different Machine Learning Regression Algorithms in Predicting the Concentration of Chlorophyll-a in the Vaal" earned her a distinction.

Read the abstract below:

In recent years, machine learning models have become increasingly prominent in river water quality prediction, utilizing techniques such as Ridge Regression, LASSO, Artificial Neural Networks (ANN), Extra Tree Regression, Decision Tree Regression, Random Forest Regression (RF), Support Vector Regression (SVR), and Linear Regression. However, as outlined by David Wolpert's "No Free Lunch Theorem" no single algorithm is universally optimal across all problems and datasets. This underscores the necessity of understanding the specific properties of the data and the problem domain when selecting an appropriate algorithm. Satellite imagery, which often presents non-linear and complex data structures, requires machine learning models that can effectively address these challenges. Studies have demonstrated that advanced models like SVR, RF, and Decision Trees frequently outperform simpler regression models, such as Multiple Linear Regression (MLR), in predicting algal bloom concentrations within non-linear datasets. Despite these advantages, the application of advanced machine learning methods, such as SVR and RF, for water quality assessment using remote sensing data remains relatively underexplored.

Congratulations to Phemelo on her outstanding achievement!