RazerArdi/SA-River-Quality-Forecaster

A machine learning solution to forecast water quality parameters (alkalinity, salinity, phosphorous) for South African rivers using geospatial and environmental data. South African River Water Quality & Spectral Archive # ECOHYDRO-INTELLIGENCE: GEOSPATIAL WATER QUALITY FORECASTING ### Advanced Spectro-Chemical Modeling & Spatio-Temporal Analysis **Architecture** • **Predictive Analytics** • **Geospatial Intelligence** • **Environmental Auditing** ## Executive Summary **EcoHydro-Intelligence** is a high-precision machine learning initiative designed to predict critical water quality parameters—**Total Alkalinity**, **Electrical Conductance**, and **Dissolved Reactive Phosphorus**—across South African river systems. Developed in response to the challenge of monitoring distinct river locations between 2011 and 2015, this project synthesizes remote sensing data (Landsat Level-2) and climatic variables (TerraClimate) with limited in-situ ground truth data (~200 samples). The core objective is to overcome the limitations of sparse environmental data through rigorous feature engineering, non-linear polynomial interaction modeling, and asymmetric hyperparameter tuning. The system achieved a **Pearson Correlation Coefficient of 0.40**, establishing a robust baseline for inferring chemical concentrations from moderate-resolution satellite imagery. --- ## Project Ontology & Workflow The system operates on a strictly defined ontology where spectral radiance serves as a proxy for chemical composition, modulated by temporal and spatial scalars. ```mermaid graph TD %% --- Data Ingestion Layer --- subgraph Data_Layer ["1. Data Acquisition & Preprocessing"] SAT["Landsat Level-2 & TerraClimate"] --> ETL["Snowflake ETL Pipeline"] GROUND["In-Situ River Samples"] --> CLEAN["Outlier Detection & Imputation"] ETL --> MERGE["Spatio-Temporal Join"] CLEAN --> MERGE end %% --- Feature Engineering --- subgraph Feature_Engineering ["2. Advanced Feature Synthesis"] MERGE --> POLY["Polynomial Interactions (Degree 2)"] MERGE --> SPECTRAL["Spectral Indices: NDWI, Turbidity, Salinity"] MERGE --> TEMPORAL["Cyclical Seasonality (Sin/Cos)"] POLY --> …