The emergence of powerful new artificial intelligence methodologies must be matched by tools that enable the remote sensing community to exploit their full potential when applied to the anticipated increasing volumes, cadence and diversity of sensor data. As it has been proved in other fields where artificial intelligence is revolutionizing entire industries (e.g. the self-driving car, video processing, voice recognition, face recognition), the availability of large volumes of training data representative of the phenomenologies to be modeled is key to overcoming a fundamental obstacle. A related challenge is the issue of sensor interoperability.

In September 2015 the United Nations adopted the 2030 Agenda to achieve a better and sustainable future for all. We have only a decade to meet its Sustainable Development Goals (SDGs), the most ambitious agenda for global change in UN history. The measurement of sustainability has been a topic of debate among researchers, policy makers and other stakeholders. The 17 SDGs and related targets have been designed to be monitored through a set of global indicators, 93 of which are designed to measure the environmental dimensions of “sustainable development.” Alarmingly, nearly five years into the agenda the UN Environmental Programme has found that we have little to no data on 68 percent of them.