Lighthouse Proof of Concept

What kind of problems are we facing?

Humanity and our global ecosystem are under attack on all fronts, with increasingly more severe weather events that take more important human and economic tolls every year, or risks of biodiversity collapsing at all scales, with rainforests burning down and oceans attaining their maximum CO2 absorption threshold. How can we assess and predict the consequences of increasing CO2 concentrations, rising sea temperatures, severe weather events on marine wildlife and on the economy of coastal areas?

Our solution

Lighthouse is a machine learning powered data visualization tool that helps scientists and policymakers gain insight into the consequences that climate change causes on the economy, wildlife, and the ocean’s behaviour. How do rising sea temperatures affect the ocean’s current at a local scale? How will the changes in the ocean’s current affect the local wildlife’s population and migration? What are the potential consequences of having a part of the local marine wildlife moving away from the coast or to another coastal area? How can we predict the occurrences of the impact of severe weather events on a coastal area’s local economy and wildlife using data features such as the concentration of CO2 in the air, sea temperature, and humidity? Lighthouse plans to answer these questions.

Datasets and tools used

In order to build the part of Lighthouse that will assess the potential economic damage a severe weather event has on a coastal area, we are going to build a deep learning model trained on the FAIR1M dataset so that our model will be able to recognize harbors via satellite imagery and assign to them an economic value.

To predict how ocean temperature and carbon concentration can affect wildlife migration, we will use SeaDataNet’s datasets and tools as well as ICOS, Emodnet, and SOCAT datasets to predict sea salinity, temperature, carbon uptake and chlorophyll concentration to train a model via regression that will help us predict how climate change affects the chemical contents of the ocean as well as marine wildlife.

Finally, to predict severe weather events, we will build a model using features such as carbon concentration, ocean currents and sea temperatures. In order to do so, we will mix datasets such as NOAA’s dataset on hurricanes and tornadoes, another NOAA dataset on coastal temperatures, SeaDataNet’s data on sea temperature and ODIAC’s fossil fuel emissions dataset to train our model.

Feedback

Adding API documentation to the services might also be a good idea to make integration easier. Another way to make the data more accessible might be to release it on open data platforms.