A summary of the extent of area associated with each class for 2000
and 2010, and changes in area per class over the study period is provided
in Table 6. These are likely valid estimates of areal LCLUC given the
high individual date accuracy. In addition, the values shown in Table 6
were adjusted in order to balance commission and omission errors.
The methods of Stow et al. (2013) were used to determine the magnitude
of bias between user's (commission) and producer's (omission) accuracy.
User's minus producer's accuracy was calculated on a per-class and per-date basis, and area estimates were slightly adjusted according
to these biases (e.g., by increasing or decreasing the area based upon
omission/commission percentage difference). Results indicate that Secondary
Forest exhibited the greatest decrease in cover, totaling approximately
4500 ha,while savanna exhibited the next greatest loss of cover
(almost 2000 ha). Agriculture land cover increased by over 6700 ha,
while Built increased by approximately 900 ha. It is clear from the results
that Secondary Forest and Savanna are being replaced with Agriculture
and Built land. Approximately 450 ha of Forest were lost, and
increases in Commercial Agriculture and Mining also occurred. The increase
in Commercial Agriculture is likely underestimated, due to the
fact that small plantationswith tree cropswere not detected in the classification
products. In terms of percentages, a 13% decrease in Natural
Vegetation occurred, with a corresponding 12% increase in Agriculture
and 1% increase in Built.