OrbiShield AI
Explainable flood alerts from NASA data
Explainable alert system for floods and landslides, combining NASA satellite data with transparent artificial intelligence.
Overview
System developed during the NASA Space Apps Challenge that combines three NASA satellite data sources (rainfall, soil moisture and terrain slope) in a logic table to classify risk into three levels (moderate, high or critical) and generate automatic alerts with clear action recommendations for local authorities.
The problem
NASA climate data exists but doesn't reach municipalities in an accessible format. Traditional machine-learning systems work as black boxes, hiding the logic behind decisions. Local authorities need a tool that is scientifically robust and at the same time simple to operate in order to make decisions that can save lives.
Approach
Generates automatic alerts for municipal authorities and residents about flood and landslide risk, based on public satellite data. Every system decision is fully explainable: the user can see exactly which factors led to each risk classification, removing the opacity typical of machine-learning models.
Impact
Developed during the NASA Space Apps Challenge, with integration to three public satellite data sources and a focus on explainability. The architecture was designed to be adapted to different municipalities, with adjustment of local parameters.