We have done it in wetlands using an airborne sensor with only four spectral bands and have reached accuracies of over 70%. There are also very successful studies with Landsat and Spot as far as I know. Regards, Javier
To answer your question I need more information as this is a question of spatial context and scale. My short answer is therefore yes and no. I have used medium scale Landsat and high spatial resolution multi-spectral satellite imagery to map Para grass reliably using Object-based classifcation methods. Commision error ranged from about 1 to 5% for QuickBird and WorldView.
This species tends to form very dense cover so is easily discriminated in this situation. Regardless of spectral resolution of your sensor, sparse or small colonies (< 9 pixel) will less likely be discrimated because of spectral mixing issues in combination with the highly variable environmental conditions on wetlands over time.
Dear Victor,
Spatial resolution is definitely a limiting factor when you are trying to map individual species. However, as you are trying to discriminate trees, it should be a fairly easy process given that you use high or very high resolution multispectral imagery (NIR included) and image texture features.
As I see it the only way to detect invasives iwith multispectral imagery s to choose the time of year carefully when their spectra differ from background vegetation. Do you agree?
Dear Vic
I agree that time of acquisition is crucial. Often multi-seasonal images are also highly valuable, and in the specific case of Phragmites, I would recommend using an early spring image and a summer image. The old flowering stands of Phragmites are quite distinct during wintertime, and in combination with the summer images it is possible to get quite a strong classification using 4, or even 3 band, aerial imagery.
- Badges
- Science topic
- Similar topics
- Geoscience
- Remote Sensing