AnyiraRodney/spatial-analysis-stunting-kenya

An end-to-end geospatial data science pipeline linking high-resolution WorldClim satellite rainfall data to DHS child malnutrition metrics in Kenya to identify climate-driven stunting hotspots # Spatial Analysis of Climate-Driven Child Malnutrition in Kenya (2022) ** Live Interactive Dashboard:** Launch the Climate-Driven Malnutrition Predictor *(Built with Python, Scikit-Learn, and Streamlit based on the spatial data engineered in this repository).* ## Executive Summary This repository contains an end-to-end geospatial data science pipeline that links high-resolution satellite climate data with demographic health surveys to identify stunting hotspots in Kenya. By shifting the analytical focus from national annual averages to localized seasonal deficits , this project mathematically isolates the environmental triggers of chronic malnutrition. ## The Data Gap Traditional malnutrition research often relies on localized SMART surveys or generalized annual climate averages, which fail to capture the granular reality of rain-fed agricultural economies. To bridge this gap, this project utilizes **Geospatial Data Engineering** to fuse two massive, disparate data structures: 1. **Vector Data:** The 2022 Kenya Demographic and Health Survey (DHS), capturing the anthropometric Z-scores (stunting metrics) of over 19,000 children 2. **Raster Data:** WorldClim (CHIRPS) 30-second resolution satellite imagery, providing pixel-level historical precipitation data ## Methodology & Architecture The pipeline is entirely built in `R` using `sf` and `terra` for spatial extraction, bypassing standard tabular analysis to explicitly account for geographic dependencies. * **Cloud Deployment (Python Bridge):** The cleaned spatial matrices extracted from this R pipeline were ingested into a Python environment to train a Random Forest Classifier. The resulting AI model is hosted live via Streamlit Community Cloud to simulate real-time stunting risks based on localized climate inputs. 1. **Spatial Harmonization:** Extracted GPS coordinates from the DHS nested clusters and mapped 19,530 individual observations across the Kenyan landscape. 2. **The Seasonal Pivot:** Rather than rel …