AINPP Intercomparison

Intercomparison and evaluation of nowcasting products during WISER-EWSA Testbed 2

Background - AI for Nowcasting Pilot Project (AINPP) with application to the African continent

The Artificial Intelligence for Nowcasting Pilot Project (AINPP), envisioned as a WMO WIPPS pilot project, emerges in a critical era where high-impact weather events are increasingly affecting socio-economic developments worldwide. These events present significant challenges, especially in developing countries of the tropics, due to the inherent low predictability of convective storms, limitations in weather forecasting capabilities and resource constraints.

AINPP is a response to this global challenge, conceptualized under the auspices of the WMO. It aligns with the WMO's strategic focus on implementing new technologies, including Artificial Intelligence (AI), to enhance earth system observations and predictions. This initiative is part of the broader commitment to the Early Warning for All initiative, which aims to provide protection against extreme weather events for every individual on the planet.

The use of AI in weather nowcasting represents a paradigm shift in meteorological studies, offering an unprecedented opportunity to harness advanced technologies for enhancing severe weather nowcasting capabilities. AI-based nowcasting is particularly promising for developing countries as it can be developed and implemented more easily and with fewer barriers compared to traditional Numerical Weather Prediction methods.

AINPP's inception is driven by the recognition of the rapid deployment of AI-based nowcasting in various regions and the necessity for National Meteorological and Hydrological Services (NMHSs) in low- and middle-income countries to access effective and reliable nowcasting solutions. The project focuses on evaluating and demonstrating the potential of AI technology in improving severe weather nowcasting and strengthening the capabilities of targeted developing countries to provide early warning services.

The project convenes an international collaboration of experts, including AI-advanced participants and representatives from developing countries, to develop, evaluate, and implement AI-based nowcasting prototypes. Progress in AI-based nowcasting is coming in parallel from the public and private sectors: AINPP is working to bring these communities together. These efforts aim at increasing the accuracy of early warning systems, thus saving lives and mitigating weather-related losses.

The AINPP stands as a pioneering effort in the application of AI in the WMO community, with the potential to set a global benchmark in weather prediction and early warning systems, especially in regions vulnerable to extreme weather events to support the Early Warning for All Initiative.

 

AINPP Intercomparison Project Objectives and Goals

Aims of the intercomparison

• Document current capability for Africa, to inform operational application of the methods / products;
• Identify sources of error, e.g. challenging meteorological phenomena;
• Provide evidence needed for model development in future, e.g. the most promising current approaches and the phenomena which need more attention;
• Provide recommendations for the evolution of African-owned systems

Organisation of the intercomparison

Coordination and Contacts: The intercomparison initiative will be coordinated by Dr Abhilash Singh ([email protected]), with support from the AINPP Africa group. Collaborations will involve National Meteorological and Hydrological Services (NMHS) from the region, including Zambia Meteorological Department (ZMD) and South African Weather Service (SAWS), alongside international stakeholders providing nowcasting products. A centralised coordination mechanism will ensure streamlined communication, adherence to standards, and effective data sharing.

Dates: The intercomparison activity will focus on Southern Africa and span 7 months, from October 2024 to April 2025. This timeline has been selected to coincide with the period of the WISER-EWSA testbed (https://www.wiser-ewsa.org/testbed/), which enables the use of local observations for verification, as well as narratives of weather impacts from the user engagement. The period includes seasonal variability and capture diverse weather patterns, ensuring a robust evaluation of nowcasting products.

Deadline for submission of data: All contributing organisations are required to submit their datasets by 31^st May 2025. This deadline provides sufficient time for the compilation and preliminary analysis of data before the final evaluation phase.

Statement/Agreement of data submission principles:

1. Ensure transparent and collaborative intercomparison of nowcasting products.
2. Utilise standardised formats and methodologies for data submission and validation.

• Share results for scientific advancement and operational improvement.

Plans for publication: The outcomes of this intercomparison will be documented in a peer-reviewed publication, highlighting methodologies, key findings, and recommendations for operational application. The results will also be presented at relevant conferences (AGU, EGU, AMS, etc.) to share insights with the broader meteorological community. We will have the option to anonymise data in the published results.

Timeline

Milestone	Date
Kick-off and domain finalisation	February 2025
Intermediate analysis and feedback	March-April 2025
Data collection and submission	May 2025
Analysis and evaluation	June-August 2025
Publication and presentation	September 2025

 

Aims and Outcomes document