Clustering Similar Groundwater Sites (SOM)

Assessing groundwater site similarity via Self-Organizing Maps

To support a regional water council in assessing groundwater measurement site similarities, I developed a custom Self-Organizing Map (SOM) algorithm in Python. This advanced Machine Learning approach provided deeper insights into spatial and physical variabilities within the groundwater monitoring network, revealing patterns that might have gone unnoticed with traditional methods.

Unlike standard implementations, the custom SOM model was designed to handle both clustering and classification. This offered greater flexibility in analyzing groundwater data. By incorporating these enhancements, I provided the organization with a powerful tool to categorize monitoring sites based on key hydrogeological characteristics. This improved their ability to interpret groundwater behavior across the region.

Beyond technical development, I played a key role in communicating insights across various departments. Through tailored visualizations and targeted explanations, I helped colleagues from different backgrounds understand and apply the results. This ensured that the findings could support data-driven decision-making in groundwater management.