Dataset Details
Title:
Ghana’s Indigenous Intel Challenge
Details:
Across Ghana’s Pra River Basin, local farmers have long relied on generations of knowledge- using the moon, wind patterns, bird and plant behavior, and even the stars - to anticipate rainfall. With limited access to modern meteorological tools and the inaccuracy of these modern methods in rural Ghana, indigenous knowledge has remained a vital component of agricultural planning and rural resilience.
Despite its importance, this form of forecasting has rarely been digitized, quantified, or evaluated systematically. Thanks to an innovative data collection initiative using the SIW Mobile App, for the first time we have structured data that captures forecasts based on indigenous knowledge alongside actual rain measurements. This opens a new frontier: merging traditional knowledge with AI modelling to build more inclusive and locally-grounded weather prediction models.
Your task is to build a classification model that predicts the type of rainfall—heavy, moderate, or small—expected in the next 12 to 24 hours, based solely on indigenous ecological indicators submitted by trained farmers.
The training dataset contains farmers' forecast submissions, the indicators they used (like sun, cloud, wind, moon), and the actual measured rainfall (via rain gauges). The test dataset will require you to make predictions on unseen examples using similar indicator data.
This challenge encourages you to explore the scientific potential of Indigenous Ecological Indicators (IEIs), while contributing to an effort that empowers local communities and enriches global meteorological understanding.
This challenge supports the development of accurate, hyper-local weather predictions where traditional models often fail and most importantly validates indigenous methods that helps bridge scientific and cultural knowledge systems, giving agency back to rural communities.