Original article was published by Andjela Djurovic on Artificial Intelligence on Medium
Role of AI in weather prediction: SeVaRA
Weather forecasting has witnessed a great progress in history. In the late 20th century, satellites went a step forward by providing an “eye to the sky” to monitor Earth systems that include tropical storms and hurricanes. Now, there is a huge volume of data, from producing simulations of long-term climate trends to predicting small weather events just a few minutes or hours in advance.
To adapt machine learning to weather prediction, there are certain criterias to be fulfilled for diverse geophysical domains. This consists of different areas in order to satisfy the goal: image processing, pattern recognition, data fusion, mapping and prediction capabilities.
ML could be explained as a “learning from data” approach. A trained network estimates an output from a set of inputs. It is about extracting information from large data sets and establishing complex relationships between data sets of different types.
Modern weather forecasting is based on the large data collection. Important thing is the use of high resolution remote sensors that can contribute to more accurate weather prediction, it also brings the difficulty of how to understand and maximize that data. AI and data-driven methods can fill this gap.
AI is not new to meteorology, it has its role in weather forecasting when neural networks were first introduced. With AI models’ power across a variety of industries in recent years, meteorological researchers are now applying the tech in weather forecasting and other business and environmental fields.
Innovative potential of our project SeVaRA
Distinctive feature of the SeVaRA project is its high innovative potential. Through the system, in fact, it is possible to generate an aggregate environmental risk index, based on the fusion and coherent integration of data of different nature and usually usable independently: weather data, in real time and historicized, from the ground or remotely sensed, deformation maps, based on interferometric techniques deriving from SAR technology, data from social networks useful for characterizing the risk from the point of view of the common user. It will also be possible to obtain information in real time and have a history of certain events that have already occurred.
These data will be processed using appropriate algorithms and artificial intelligence techniques, in order to produce and make available, for each user of the SeVaRA system, whatever the area in which it is located, the aggregate risk index.
This is a highly innovative aspect, as no services have yet been found that are able to operationally dispose of complex information, generated by the fusion of such heterogeneous data, and to apply such information patterns to ensure the protection of man and the environment. .
The added value exhibited by the SeVaRA project in terms of innovation is also linked to the absence of forecasting systems for weather consequences. In fact, although weather forecasting systems are widely used, there is no standard, based on numerical indices, for monitoring the risk associated with the effects of a weather event, regulated based not only on the data collected by the sensors but also on the characteristics. geophysical and anthropic of the territory.
Increasingly, risk-related information will be usable in different ways depending on the end user of the service: the outputs of the SeVaRA system may, in fact, be of interest to the common citizen in the form of alerts, or can be integrated into public administration web services, as well as interpretable by qualified personnel or, finally, usable by private companies interested in entering data in other processing chains.
Finally, it is possible to note the high quality expected in the interferometric processing capacity proposed by the SeVaRA project: currently there are few realities, in the national and international panorama, able to offer outputs of the same level.
The services associated with interferometric products are also significantly remunerative and this aspect, combined with the technological competitiveness and continuous research activity that characterizes the project, could certainly orient SeVaRA towards promising market opportunities.