Considering the importance of ensuring accuracy and scientific validation in remote sensing data processing, MapBiomas Indonesia has completed an accuracy assessment and validation process for Collection 2.0. In addition to showing the overall classification accuracy, this analysis also shows the accuracy and error level for each classified class. MapBiomas evaluated the global and per-class classification accuracy for each year between 2000 - 2022.

The accuracy estimation is based on the evaluation of pixel samples, which we refer to as reference data consisting of ~12,957 samples. These samples were previously selected using statistical sampling techniques. This reference data validation process involved 10 universities in Indonesia: Syiah Kuala University, Bengkulu University, and Lampung University in Sumatra; Muhammadiyah University of Palangkaraya and Mulawarman University in Kalimantan; Tadulako University in Sulawesi; Papua University; and University of Indonesia, Gajah Mada University, and Bogor Agricultural Institute in Java. Accuracy was assessed using an error matrix comparing the mapped land-cover class with the validated sample land-cover class.