Resumen

VALDERRAMA-LANDEROS, Luis et al. Capacity of phenological data derived from CYCLOPES LAI for the year 2000 to distinguish land cover types in the state of Michoacán, Mexico. Rev. Chapingo ser. cienc. for. ambient [online]. 2014, vol.20, n.2, pp.261-276. ISSN 2007-4018.  https://doi.org/10.5154/r.rchscfa.2013.08.025.

The capacity of the phenological data of the CYCLOPES project LAI series for the year 2000 to distinguish general vegetation types (evergreen forest, deciduous forest, crops and pasture-shrubland) in Michoacán, Mexico, was explored. Using the TIMESAT program, 11 phenological variables of each 1-km pixel of that series were extracted. The behavior of each variable was analyzed using the INF2000 map as a reference. The main differences relate to the deciduous or evergreen character of the vegetation. The 11 variables were reduced to five principal components (98% of the variance) to make an unsupervised classification of 250 phenological classes or groups. Each class was associated with one of the cover types, with a criterion of maximum area matched with the INF2000 reference map, in order to construct the CYCL2000 cover map. Comparing the two maps yielded modest results, with 63 % total accuracy. Deciduous forests were better identified (80.1 % of pixels correctly identified and 62.1 % correctly classified), followed by evergreen forests (74.1 %, 69.9 %), crops (62.9 %, 61.1 %) and pasture-shrubland (16.9 %, 52.3 %). The relatively good identification of forests shows that this approach could be used to estimate deforestation.

Palabras llave : Time series; phenology; biophysical variables; cover map; global data.