Period: 2001 – 2017
Sensor: MODIS Terra + Aqua
Frequency: Yearly
Resolutions: 500m
Extent: Global
Data can be accessed here.
Algorithm description: The product consists of 13 science data sets including tree layer classes based on FAO LCCS. The algorithm is based on supervised classification. Inter-annual variability is caused by classifier instability. To remove this instability ‘Hidden Markov Model’ is used to post-process maps for each year. The classes have been aggregated as per six different classification systems. Most of these systems have forest types such as evergreen, deciduous, broadleaf, needleleaf forest.
Accuracy: Version 6 product is new collection and hasn’t been evaluated yet. Hao et al (2014) assessed earlier versions (5, 5.1) over China and report some unreasonable changes in the overall landcover in the MODIS product series from 2001 to 2012. This problem was reduced in version 5.1. The uncertainty was very high in ecological transition zones and heterogeneous landscapes. Zeng et al (2015) report overall accuracy of about ~65% in China for collection 5.1.
Limitations:
- Global scale assessment has not been done.
- Coarse resolution.
Reference:
Hao, G. & Gen-Suo, J. Assessing MODIS Land Cover Products over China with Probability of Interannual Change. Atmospheric Ocean. Sci. Lett. 7, 564–570 (2014).
Zeng, T., Zhang, Z., Zhao, X., Wang, X. & Zuo, L. Evaluation of the 2010 MODIS Collection 5.1 Land Cover Type Product over China. Remote Sens. 7, 1981–2006 (2015).