Extreme Weather Forecasting and Its Impacts in Togo Using Machine Learning: WeatherAI

Local Chapter Lome, Togo Chapter

Coordinated byTogo ,

Ekoue LOGOSU-TEKO

Status: Completed

Project Duration: 29 Jul 2023 - 20 Sep 2023

Open Source resources available from this project

Github Repository

Project background.

We are in a rainy period at Lome, and it is usual to have floods in many districts. A recent article from RFI depicts the situation (kindly find the article at https://www.rfi.fr/fr/afrique/20230423-togo-des-quartiers-de-lom%C3%A9-inond%C3%A9s-et-de-nombreuses-familles-sans-logement), where local habitants need to take refuge while forsaking their home as well as their goods and letting them be destroyed by the flood. Having the water come to the adult waist is common in some places. And sometimes, several retention basins didn’t withstand the shock, overflowing and causing flooding in many neighborhoods.

Paradoxically, there are places in the country where it is usual to not have water from the faucet for a long period of time during the dryness period. People need to walk for kilometers to find a well or other water source. Difficult access to clean water sources have a negative impact on people’s health, especially children.

Other than people’s lives, extreme weather is disrupting other sectors, such as the economy. For example, agriculture in Togo depends very much on the period of the year and the weather. So it is easy for farmers to incur loss when the climate becomes extreme and unpredictable.

The problem.

We want to help address damages caused by (extreme) rain and dryness.
Climate change is making the rains and drynesses more extreme in Togo. Many things are easily destroyed/disrupted during the rainy period. Accessing clean water during dryness periods is a challenge in some parts of the country.
Some problems:
– Homes are destroyed
– Life habits are perturbed: being late at rendez-vous/works/school, inability to open stores
– Transit becomes difficult, as it is usual to have water up to the adult waist, making the roads impracticable. 
The government and the officials, as well as NGOs, are doing many things to address the situation, but the results we are seeing during extreme weather periods show that citizens are still victims of the situation.

Project goals.

1. Collect and analyze data- Collect weather data - Collect demographic data of the country (we are more interested in the density of each zone) - Exploratory Data Analysis on the collected data: the target results should be mainly oriented toward bad weather (heavy rains, drought period) and normal weather. - Explanatory Data Analysis based on the finding of Exploratory Data Analysis2. Predict weather and its impactsWe are interested in models that can make the following predictions:  - In a given time frame, and for a specific zone, what will be the rainy period and the dryness period? (For the local officials and citizens) - In a given time frame: Which zones may be subject to heavy rain? Which zones may be subject to high dryness? (For the government and NGOs) - What may be the duration, violence, and impacts of an identified rain to come? (For the local officials and citizens) - For an identified dryness period, what may be its duration? Its severity? (For the local officials and citizens) - Providing weather data on selected locations and selected time frames. (For citizens) - Other models that Collaborators may deem useful for addressing the situationWe ought to distinguish extreme weather from normal one. The ideal solution should consist of two key deliverables in the case of each model:- A model ideally deployed on API Gateway, which will respond to HTTP requests (as any API) - A containerized web application deployed on AWS, to allow end-users to interact with each model’s application. Ideally, we should have a map to depict the model's predictions. And we can also allow sending alerts (mail, sms) when incoming extreme weather is detectedTime frame: following 07 days, following 30 days, following 12 month Key zones: district, city, region (there are 05 regions in the country)

Project plan.

  • Week 1

    Data collection: Weather data, Weather impacts

  • Week 2

    Data collection: demographic data (we are more interested in density)

  • Week 3

    Data Cleaning

  • Week 4

    Exploratory Data Analysis

  • Week 5

    Exploratory Data Analysis

  • Week 6

    Explanatory Data Analysis | Model Development

  • Week 7

    Model Development

  • Week 8

    Model Deployment

Learning outcomes.

1. Data Collection and Data Wrangling
2. Exploratory Data Analysis, Explanatory Data Analysis, Presentation to an audience
3. Machine Learning and Deep Learning
4. Model Deployment through a web application and an API, on the cloud
5. Containers, Dockers, API Gateways, AWS, CI/CD, MLOps

Share project on: