Land cover/use classification is a prominent analysis in remote sensing.Now the question is if I generate a landcover from LANDSAT, do not perform ground survey for cross-checking and want to do a scientific analysis with this classification, is it scientifically invalid in academia?
If I do this cross-checking with some very high-resolution image (5m) and Google Earth, is it robust method to fit scientific standards, if yes then how much.
Please guide me, experienced.
By ground survey, I mean field survey in body with some instruments like dgps, densitometer etc.
I asked another question at the below link, that has some tone matches with this question but still feeling unhelped. I emailed Mr. Hansen about a month ago but no response yet.
http://gis.stackexchange.com/questions/216421/accuracy-of-the-mathew-hansen-dataset
In fact that`s an ideal target, to find an algorithm which be able to interpret Images Land covers, But scientists are in believe that we are a bit far from that Goal. There are some chances in interpreting Hyper Spectral images for some objects which their "spectral signatures" have been obtained.
Anyhow You need to validate your Idea. For this purpose you may interpret some images by your method for a reachable place, and compare your predicts vs your field observations. Then if the result was satisfactory, you will permit to use your method for other similar conditions.
Go On and let`s your Idea has the opportunity to be test
Be success
Yes, it is common practice nowadays in Landsat-based land cover classification research to use very high resolution data (VHR) data for sampling and validation. However, the validity of this approach greatly depends on the classes you are trying to classify and how distinguishable they are from the VHR data. Therefore, proper field surveys are still considered better but not the only way of carrying out good quality land cover classifications at the Landsat scale.
Hi
You can use from high resolution images like quick bird image that acquired from same time or near.
It is valid and its accuracy is high percent. It is used in several paper and its outcome is acceptable
I wrote an article on surface observations (ground truth) which I will share:
Johannsen, C.J. 2010 "Foreword: Lessons Learned in the Collection of Surface Reference Information," Journal of Terrestrial Observation: Vol. 2: Issue. 2, Article 3. Available at: http://docs.lib.purdue.edu/jto/vol2/iss2/art3
Preferably, a ground truth or verification is the best way of validating your mapping accuracy in remote sensing.
You cant use the word Invalid. But verification either with high resolution or with field data is always preferable.
All persons have answered Mr. S. Islam's question very elaborate way. I shall add only that remote sensing and GIS study is never 100 percent accurate, that's why scientist use Kappa method to evaluate the accuracy of mapping. Please go through several published paper in landuse/landcover.
Thank you everyone for his/her earnest endeavour! Could you please quote article(s) from highest impact factor journals that used un-ground-verified data.
- Badges
- Science topic
- Similar topics
- Indexes