Monitoring spatial accuracy of oil palm cultivation mapping in southern Cameroon from Landsat series images
Abstract
Studying and mapping palm grove evolution allow understanding the impact related to its
cultivation. Our study aims to map industrial palm grove using Landsat series images and
measures the accuracy of the produced maps. It was carried out in SOCAPALM industrial
plantation, located in southern of Cameroon. For the mapping and assessment of accuracy, perpixel
classification and confusion matrix method were used, respectively. We obtained high
correlated maps (Kappa =0.92 in 2001 vs 0.86 in 2015). However, some confusions were
observed between vegetation and oil palm classes for the two periods, affecting the maps
accuracy. These confusions are caused by the presence of mixed pixels resulting from the
spatial and spectral characteristics of palm groves, the method used to map and validate the
map, and uncertainty related to dada. To increase the accuracy, we suggest (1) to use another
mapping method such as super-resolution mapping, (2) develop a classification system of
cartographic products.