THE PROBLEM

The eutrofization is an extremely long natural phonmenon that can happen on any water body, given enough time and ideal conditions. However, when this process is accelerated through the inffluence of human actions, it's consequences are heavily aggravated.

The way this happens is simple: the availability of raw materials and organic matter is increased in a lake, river or even in sections of the sea through either dumping of garbage, chemical and industrial disposal or ship ballast coming from international travels, creates a synthetic imbalance and a favorable environment for certain kinds of seaweed and bacteria to thrive in. The exponential rise of those opportunistic organisms creates a superficial cover that blocks light and consumes most of the oxygen available in the water - in some cases even producing neurotoxic and hepatotoxic substrates that start to accumulate in the local food chain.

The consequences of this phenomenon can vary, being the most simple one the visual alteration on sections of the studied system, but the hypoxia environment created by those organisms can even kill the local fauna and flora and generate a systemic imbalance, intoxicating the beings that feed on, live near or even come in contact with them.

THE ALGA+ PROJECT

The main objective of the ALGA+ Project is to automatize the identification and classification of eutrophication processes and feed that information to decision makers, using a Machine Learning Model divided into two main phases.

We constructed an initial phase of AI Training, in which we provide information about the phenomenon to our Artificial Intelligence (AI), teaching it to understand the input from NASA's satellites on the parameters that indicate the presence of potentially harmful organisms, recognizing what is an affected aquatic body and discerning it from a healthy, clean environment. The Machine Learning Technique of supervised learning will make the AI understand the patterns that will be sought by the program in the subsequent identification.

Using monitoring data from existing NASA satellites, we will collect visual information, temperature variations inputs and data on the presence of organic compounds (such as Nitrogen, Carbon and Phosphorus), which are alterations and substrates derived from the eutrophication of lakes, rivers and seas.

Fed with this data, the AI ​​will be trained in order to create the ability to identify occurrences of eutrophication and make predictions of the worsening of the phenomenon, which allows for phase two, described below.

On the next phase of the model, our AI starts to identify and predict the growth of the problem in a completely automated way, using a Reinforcement Learning model with Crowdsourcing, where the locals (be it the population, biologists, researchers or other interested parties) can contribute to the solution directly:

People will be able to choose to be informed about the possible occurrence of the phenomenon from the system's forecasts, while they are encouraged to provide feedback about the veracity of that forecast with images and texts through online instant messaging applications and short message services. This information will be added to the database and will be used for the validation and improvement of the system in a complex continuous feedback process.

Finally, we return this information to a simple software, which will correlate the geographic coordinates with the data learned and previously crossed by the AI, returning an interface that classifies in severity levels (using simple visual color markers in green, yellow, orange and red), the occurrences across the globe, directing the decision maker's gaze to the most serious points and providing all crucial input collected along the way about that specific place on request.

The creation of this machine learning model that detects the phenomenon, understands it and integrates the population interested in preserving it's ecosystem, decentralizing the information - being it currently on public power's hands - and brings society closer to scientific contribution as a whole, giving relevance and making it possible to call for measures to combat the problem, while feeding crucial information to decision makers who have the knowledge to work and better understand the impacts, as well as the power to act directly in solving them.