- Any one help me procedure regarding Area Biased and Unbiased Accuracy Assessment in Landuse/Landcover Mapping?
Maybe this is of interest:
Article Making better use of accuracy data in land change studies: E...
Bias has to do with non radomness or with systematic in the design protocol for sample selection, while unbiased assessment has to do with radomness. There is a good document describing all such procedures in the link:
https://run.unl.pt/bitstream/10362/17875/1/D0020.pdf
Hi Rajashekar,
The area biased sampling doesn't account the land cover class area, However, unbiased accuracy assessment take account of area into consideration.
For example:
case 1: You have 4 land use classes, in which 3 classes occupies 30% area respectively and on class occupies 10%. If you want to place samples for each class lets assume 100 samples per each class, You may get the more accuracy towards class four with 10 % area. And you will end with high overall accuracy.
Case 2: You have 4 land use classes, in which 3 land use classes occupies 10% area each and the last class 4 occupies 70%. If you place 400 sample randomly by not distributing to land class. you may probably end more than 280 in the land use class 4. And you will end with high overall accuracy.
In convenience sampling we chose location of sample which are easy to us to access or due to other reasons. One should avoid this type of sampling methods.
Random sampling is a best practice. For unbiased area accuracy assessment we can use a stratified random sample approach.
Allocation of sample: If you want to allot 'n' number sample for accuracy assessment use following formula to calculate the number samples to allocate.
number of samples for land use class (x) =
(n/Total Area of image) * Area covered by land cover x.
All the above procedure can be accomplished by ArcGIS, by using Raster to Polygon --> dissolve --> Calculate area and sample using field calculator --> using the respective field to allocate the random points.
All the best
- Badges
- Science topic
- Similar topics
- Geoscience
- Cartography
- Mapping