Accuracy assessment was completed using two different spatial
scales of analysis. At the first scale, accuracy was analyzed within
square, homogenous land units of approximately 2.25 ha or greater,
and only at locations where reference data could be interpreted with
high confidence. Using this approach, overall classification accuracy
values for 2000 and 2010 LCLU maps of southern Ghana were 95% and
89%, respectively. The accuracy of the LCLUC map depicting no-change
and transitions between Built, Natural Vegetation, and Agriculture was
found to be 83% while simple detection of change/no-change was 88%.
When analyzing accuracy at the pixel-level andwith much lower confidence
in the reference class interpretation (using the secondary accuracy
assessment approach), overall classification accuracywas found to be
approximately 85% and 74% for 2000 and 2010, respectively. Further,
the accuracy of the LCLUCmap depicting no-change and transitions between
Built, Natural Vegetation, and Agriculture was found to be approximately
65% while simple detection of change/no-change was
approximately 72%. The combined results suggest that the maps may
be effectively used to study LCLUC occurring within the study area.
Accuracy assessment was completed using two different spatialscales of analysis. At the first scale, accuracy was analyzed withinsquare, homogenous land units of approximately 2.25 ha or greater,and only at locations where reference data could be interpreted withhigh confidence. Using this approach, overall classification accuracyvalues for 2000 and 2010 LCLU maps of southern Ghana were 95% and89%, respectively. The accuracy of the LCLUC map depicting no-changeand transitions between Built, Natural Vegetation, and Agriculture wasfound to be 83% while simple detection of change/no-change was 88%.When analyzing accuracy at the pixel-level andwith much lower confidencein the reference class interpretation (using the secondary accuracyassessment approach), overall classification accuracywas found to beapproximately 85% and 74% for 2000 and 2010, respectively. Further,the accuracy of the LCLUCmap depicting no-change and transitions betweenBuilt, Natural Vegetation, and Agriculture was found to be approximately65% while simple detection of change/no-change wasapproximately 72%. The combined results suggest that the maps maybe effectively used to study LCLUC occurring within the study area.
การแปล กรุณารอสักครู่..
