Dear Ladies & Gents,
I’m on the search for some global high resolution land cover type datasets. So far my search turned up the MODIS Land Cover Type Product (MCD12Q1) which is almost perfect except for the 500m resolution.
Do you know a higher resolution dataset?
Thank you for your help!
All the best,
Kristof
(Not exactly a GMT question but as there are some very knowledgeable people hanging around these forums I figured it might be worth a shot.)
Try exploring https://sentinel.esa.int/web/sentinel/thematic-areas/land-monitoring/land-cover-use-and-change-detection-mapping
Problem with ESA products is that they are horrible complicated to find what we are looking for. I almost have to change mouse batteries after a search.
Hi @Joaquim thank you for the link. Now I understand what you mean with “complicated to find”. This really is horrible. After traversing many ESA sites I found a 100m resolution data set.
Thank you again!
@KristofKoch, thanks for the mining work. Often I really don’t understand where the ESA web designers have their mind when they do their job.
After reading the Copernicus Global Land Service Product User Manual (pdf, 2,1 MB) I must report that - despite the name - the cover isn’t global. It stretches from 80°N to 60°S only. So the search continues. Maybe I have to merge it with lower resolution MODIS data.
Another possibility: Landsat8
Try this, they have a “Land Cover” cathegory
Thank you @joaquim another nice source. If I read this correctly AVHRR data has a resolution of 1km.
Those files eat my RAM for breakfast …
Get a bigger computer.
Well played gentlemen, well played …
If money no object then a $50k fully loaded Mac Pro would be nice… Otherwise, if max CPU/cores and price is important (!) then I would lean towards having as many cores as possible - that is what saves my time when I compile… I am sure a custom Linux box is the cheapest option.
Thank you for your thoughts – a maxed out Mac Pro really would be nice. While many cores are useful for compiling don’t most GMT programs run single thread? Maybe I’m compiling it wrong … I thought of as much single thread performance and RAM as I can afford.
GMT uses limited OpenMP support so some tasks are run on multiple cores if compiled with OpenMP. It comes down to use cases. In my case I probably compile GMT and run the 1000 tests many times a day, and with my now old 2013 Mac Pro and 24 cores the tests take < 3 minutes. Still, that is a long time to wait while your entire computer is humming. In your case you are probably running jobs to process data and make plots, and unless you run many at the same time then maybe single-CPU performance makes more sense. Well, any high-performance CPU these days will have many cores, but looks like as cores go up, frequency goes down for thermal reasons (?).
From what I’ve read here and there Macs are not good option to for performant computers. Specially if GPU or graphics (they will drop opengl support) are needed.
GMT does not have special needs. Multi-cores are not particularly used (parallelizing is an extremely difficult/laborious work). GMT itself is not very efficient in I/O.
My old-age discussion with Paul is that I don’t believe the answer is always “bigger computers” because data grows faster then chips.
Try some GDAL tricks with your breakfast eating grids. Namely pyramid layers (gdalado)
Working with those files on a laptop is a pain. I like working on Macs (easy GUI and UNIX under the hood) but the price point is a bit high. As it turns out hosters don’t like it if you saturate the rented virtual server for a couple of hours. So time for a machine under my desk. It will be a Linux box for most bang for the buck. AMD just unveiled some nice new CPUs …
Thank you for bringing up gdaladdo – it is completely new to me. Time for some reading and experimenting! If this holds what it promises the file size won’t work as an excuse to get a new box under my desk … dang.