Accuracy assessment of thematic map is often performed using groung truth sample points, generated randomly. There are techniques to evaluate the appropriate number of sample points. However, since those observation are "local", I have the strong feeling this strategy might fail to detect severe spatial inaccuracies (i.e, checking the accuracy of detected contours of different land cover patches obtained as a result of supervised classification).
As such, I was wondering how transect lines can help here. Since they provide a continuous information along a line over the mapped area, I was wondering how thematic accuracy can be assessed this way ? I came accross an interesting read (Skidmore, A. K., & Turner, B. J. (1992). Map accuracy assessment using line intersect sampling. Photogrammetric Engineering and Remote Sensing, 58, 1453-1453), however I missed some important details and could not be able to apply the method described in the paper correctly.
Any thoughts ?
Thanks.
Have a look at line-intercept/transect sampling in our papers (Chromolaena) of 2006 in the South African Journal of Science and 2008 (forest/gorilla) in the African Journal of Ecology
Sounds like an interesting idea.
For one, should I create two series, one for patchs lengths measured on the map and another for corresponding patch lengths measured on the field and run hypothesis tests (t-tests for means, Fisher for variances) to assess homogeneity between the two series ?
- Badges
- Science topic
- Similar topics
- Engineering