Initially we looked at animal behaviour as a marker for cataclysms and we realized that one cataclysm that animals cannot foresee is toxic algae, which causes whale embankments, and is responsible for the death of seals, fish, shellfish and countless marine life. We searched on the economic impacts and realised that it effects tourism, fishing, recreation and a ton of other activities that leads to extensive money loss. So we asked ourselves what was the only way to minimise losses .

We studied the pictures that satellites took of the algae and realised that getting an overview of the conditions that cause algae and thus predicting HAB production in a given area could lead to treatment before HAB even occurs thus preventing any negative impact. Then we thought "THEN WHAT? " So now we had data and we had the locations on where HAB would occur, but how could we use it? We studied about the different sectors which would benefit with this information and companies like Bgtechs that have a cure to this problem. To improve pre-existing methods we also will track the effectiveness of the cure and correlate it with the conditions and the 300 species of HAB. All throughout we applied Lean Business tools and Design Thinking processes, along with Innovation techniques to ensure our proposal was not only feasible, but also desirable and viable, integrating good ideas with business acumen and innovation.

There are other institutions, apps, and websites that work in this arena of research on algae bloom growth, where it is very common for Senintel-3 data to be used, which has very poor spatial resolution (500+ meters). For our purpose, we want to offer our users fine spatial resolution for any interested parties who would like to take preventive action on the algae growth, for which they would need specific locations of HABs mapped. For this reason we are using Sentinel-2 imagery as our main source of data from the ESA Copernicus Program which gives us better spatial resolution (10 meters).

An interesting application of the satellite imagery is our machine learning model used to quickly produce a vegetation index composition of the Sentinel-2 image. For this, we use a U-Net Convolutional Neural Network architecture which is commonly used for image segmentation and classification. Using Sentinel-2 imagery with a corresponding Normalized Vegetation Difference Index (NDVI) mask calculated from Sentinel-2's red and near-infrared bands. By using this as the training data, we created a model that can quickly create a corresponding vegetation index image without the usual NDVI processing step. This will allow quick access to our user with information related to areas of HAB. We will do this for other indices also such as NDWI, FAI, etc.

With this database, we will go another step further using our historical vegetation indices and accompanying ancillary data to produce a time-series forecast giving predictions of algae growth. If the predicted algae bloom growth is estimated to surpass a limit soon, we will notify our subscribers of such activity and the potential hazard.

https://docs.google.com/presentation/d/e/2PACX-1vTyPZmnhR5CqLk2NqC6uVFrtVmeTPWinVdEFr_8Ljduzhc2R_O6JtVLRGxohv9ElER2mj6Tim9dRk1U/pub?start=false&loop=false&delayms=3000

• https://scihub.copernicus.eu/
• Google Earth Engine
• https://towardsdatascience.com/unet-line-by-line-explanation-9b191c76baf5
• https://github.com/h4k1m0u/snappy-scripts
• https://www.nature.com/articles/s41598-020-65600-1