Terra-i is a near-real time monitoring system for natural vegetation conversion at pan-tropical scale which is now being scaled for application across the entire tropics. It uses satellite data and computational neural networks in order to detect anthropogenic changes in the vegetation cover every 16 days in near real time.
Terra-i uses computational neural networks to detect or ‘learn’ how the vegetation vigour (measured as NDVI) behaves at each site over a period of time in relation to observed rainfall. This knowledge of historic vegetation response to rainfall is then applied to current measurements of rainfall to predict what the vegetation response should be. This prediction is then compared with NDVI data taken by the satellite and if the observed response is significantly different from the historic responses given the pattern of rainfall then the pixel is marked as one that may have changed through anthropogenic means. If this change remains two 16 day periods in a row then the event is confirmed. Much work is carried out to account for and remove the effects of drought, flooding and cloud cover or other image ‘noise’ that may produce false positives.
Funding and implementation partners
Developed in collaboration with the Nature Conservancy, Haute Ecole d'Ingénierie et de Gestion du Canton de Vaud, based in Switzerland and Dr. Mark Mulligan of the Department5 of Geography, King's College London.