Yannis Papoutsis, assistant professor at NTUA, explains
In the prevention and mitigation of the effects of risks from natural disasters, it will be possible to contribute to Artificial Intelligence, which is the leading topic of discussion in the scientific community. The use of artificial intelligence will be a tool in the hands of scientists, business entities and technology companies in the management and response to the climate crisis.
Speaking to the Athenian-Macedonian News Agency, the assistant professor at the National Technical University of Athens and associate researcher at the IAADET of the National Observatory of Athens Yannis Papoutsispoints out that AI in relation to climate change comes to work on three different levels.
First, as mentioned by Mr. Papoutsis, in the prevention but also in the prediction of the occurrence of a natural disaster. “Here we are focusing in relation to climate change on extremes, extreme events, which is different than predicting an average risk. That is, in the event that has the potential to develop into some very big disaster like those we have experienced in Greece with the floods and the forest fires. Artificial intelligence will try to predict this increase in the frequency and intensity of these phenomena and the disasters they cause“, explains Mr. Papoutsis and adds that a second piece is explainable artificial intelligence (explainable artificial intelligence). “In other words, to make interpretable data so as to build trust in what we produce. For example, it is not enough to tell the operational agent that the risk of a fire or a flood will be high, but you should explain to him the reason why a model has reached such a conclusion”, notes Mr. Papoutsis.
A third field, as Mr. Papoutsis points out, will be the causality. “There are some algorithmic approaches that try to link cause and effect. That is to say that when we have a prolonged temperature above that, and rainfall less than that then we have a high risk for the start and spread of a catastrophic forest fire. But to try to do this in a way not empirical but through the data that there is this cause and effect relationship and not a correlation but a causation relationship which is different in scientific terms, correlation with causation. When one uses artificial intelligence for events related to climate change and natural disasters one is called to do earth system deep learning. In other words, in order to be able to predict what will happen in Greece two months from now, I need to obtain information about meteorological conditions for climate indicators from everywhere in the world. It’s a new direction that essentially tries to utilize observations on a global scale to be able to say what will happen locally,” Mr. Papoutsis underlines to APE-MPE.
As he emphasizes, all these scientific tools can mainly help in mitigation, prevention and planning.
“There is a very large wealth of data, such as observations from terrestrial networks, meteorological, flying, seismographs, scientific instruments, with the result that a lot of data has been collected, whether it is satellite data, whether it is data derived from computer models and simulations. “The wealth of data that we have, the very good computing power now with the line cards and with the supercomputers that are available and the algorithmic part of artificial intelligence that has blossomed in recent years, if you take all three together you can leverage this historical data file where we had observations from natural disasters”, notes Mr. Papoutsis and adds that they have collected the historical data and events surrounding a natural disaster, such as what extent it had taken, with what intensity and with what frequency it occurred.
“Essentially we have the opportunity to train with all this data, models, so that we can find all these correlations between the parameters that lead to a natural disaster in order to predict when a major disaster will come. This is the great contribution of artificial intelligence to be able to predict the risk of a series of disasters on various spatial and temporal scales”, Mr. Papoutsis underlines to APE-MPE.
In the part of climate change, as he emphasizes however, there are some issues that are indisputable and some to be studied and investigated. Nevertheless, in the scientific community, as Mr Papoutsis explains, there is a buzz about Chat GPT and what it can produce by harnessing a wealth of data and training models. “Accordingly, the algorithmic technology is the same, the data we have to draw from the satellites, the meteorological networks are there and available, we have the corresponding wealth of information, so there can also be a corresponding Chat GPT that models all these complex and the dynamic character of the earth system with the atmosphere, the sea to model them and to be able to go to the corresponding Chat GPT”, notes Mr. Papoutsis to APE-MPE.
However, alongside the scientific community, artificial intelligence has also entered the lives of business entities that are called upon to manage natural disasters and extreme phenomena. As Mr. Papoutsis points out, business people are now surrounded by scientists, but what is needed is “for these two worlds to come a little closer.” And this requires effort from both sides.”
“There is distance to be covered on both sides. A model that will be presented to an operator and will concern the flood or the possible initiation of a landslide must be explainable, first for scientific purposes but also to help the decision maker to trust it. Because the issue of trust is very important in general in artificial intelligence but in particular when it has to do with decisions that have to do with human life, with economic and environmental impact etc etc. So he has to build a model that is explainable and this explanation to be consistent with the empirical knowledge that the business person has from years of cumulative work and which is very important and very useful”, explains Mr. Papoutsis.
Since 2021, through innovative programs developed with the coordination of the National Observatory of Athens, Mr. Papoutsis has utilized artificial intelligence to prevent and deal with forest fires. These innovative programs that end on 31/12/2023 attempted to assess forest fire risk in different ways by leveraging artificial intelligence technologies and a wealth of satellite data. The “Deep Cube” program was developed in collaboration with the Fire Brigade and operated as a pilot for Greece, while it was also extended to the rest of the Mediterranean. “The Fire Department also participated in this effort from the first day. And with the provision of data and with the provision of expertise but also with the continuous participation of staff who answered questionnaires, the data and the maps that we produce on a daily basis. The Fire Brigade was very open-minded and very positive in helping us”, notes Mr. Papoutsis.
“We live in a fantastic time, this is the reality, all this has come a bit cataclysmic and from one moment to the next we will see things being used directly and operationally. From classic numerical models for weather forecasting we will go to models that utilize artificial intelligence. This is the future and this will include both natural disasters and the study of climate change. I think cataclysmic things will happen, however there should be a regulatory framework for how it can work”, concludes Mr. Papoutsis to APE-MPE.
