Page 233 - AI for Good Innovate for Impact

AI for Good Innovate for Impact

• Train a cohort of African data scientists in AIWP modeling and forecasting
• Source and curate data from public datasets (ERA5, WMO AWS) and unique local sources
(WASA, AfriNet, WNS)
• Support local organizations and government bodies to integrate AIWP tools into weather- Change
sensitive sectors 4.2-Climate

Approach:

• Use existing AIWP model weights from GraphCast/Aurora
• Incorporate station-level and satellite data
• Leverage low-cost infrastructure and local community knowledge
• Promote open access and reproducibility of tools and insights

Use Case Status: Research and Product Development Phase

Partners: African Institute for Mathematical Sciences (AIMS)

2�2 Benefits of use case

1. weather prediction and local climate model for Africa. Brings benefits to weather sensitive
sectors such as agriculture
2. Incorporating local community knowledge and supporting local organizations and
governments.
3. promoting open access, tools and insights.

2�3 Future Work

• Expand curated datasets across additional African regions and climatic zones
• Develop and validate proof-of-concept AIWP models for deployment in local sectors
(agriculture, disaster risk, energy)
• Publish open-source models and data pipelines

Host regional training workshops and establish mentorship programs for African AIWP
researchers

3 Use Case Requirements
• REQ-01: It is critical that models be fine-tuned or developed specifically for African
weather conditions
• REQ-02: It is expected that curated datasets include both reanalysis data and ground-
based AWS inputs
• REQ-03: It is of added value to incorporate low-cost sensors and community-sourced data
• REQ-04: It is expected that AIWP models can run efficiently on limited computational
infrastructure
• REQ-05: It is critical to establish a sustainable pipeline for local expert training and
community participation

197