Open thiagomanel opened 6 years ago
This will not be necessary anymore, since all the landsat scenes will be merged to make a single safe landsat scene.
Uhm, I'm afraid we cannot simply merge all the landsat scenes (maybe I did not understand yours merge algorithm, is it just a sum?) because we have a peak during the time range. In other words, we there are multiple three days samples, one closer to the peak than other, we should favor the closer sampler).
Em ter, 16 de out de 2018 às 16:26, Felipe Mota notifications@github.com escreveu:
This will not be necessary anymore, since all the landsat scenes will be merged to make a single safe landsat scene.
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/simsab-ufcg/ndvi-finder/issues/4#issuecomment-430367551, or mute the thread https://github.com/notifications/unsubscribe-auth/AANQ0XzL6g8q4lzlPbozKVJRRGHXTiatks5uljL1gaJpZM4XfZ-N .
For all the cases that I tested the maximun was only 4 images in the desired period, so it's possible to use everything, if this facts becomes true for the whole region, but we can let the issue open for later discussion
ok. let's make the simpler algorithm. Later on we can make it more complicated
On Tue, Oct 16, 2018 at 4:47 PM Felipe Mota notifications@github.com wrote:
For all the cases that I tested the maximun was only 4 images in the desired period, so it's possible to use everything, if this facts becomes true for the whole region, but we can let the issue open for later discussion
— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub https://github.com/simsab-ufcg/ndvi-finder/issues/4#issuecomment-430374836, or mute the thread https://github.com/notifications/unsubscribe-auth/AANQ0RuwVxplUq22JVrfd1aubpmXyNEoks5uljfSgaJpZM4XfZ-N .
Each sub-region of the Brazilian semi-arid has a target period in which we should analyse the landsat dataset (to generate the safe ndvi).
we should design and implement the transverse algorithm. e.g Should we start by analysing the "optimal" data and move in both temporal directions in the time range?