LAI time series

HI all, SNAP newbie here :pleading_face: please be gentle!
i’m interested in using LAI values as an input variable for a combined canopy/hydrology model I’m developing, and having searched this forum i’m a little confused.
I find many posts about calculating LAI, but there is an available product for LAI at 300m resolution, why is everyone trying to calculate it? Is the available product not good?
my main question is, how do i identify a single pixel in the .tiff file i download as a point ‘on the ground’ so that I can use it to validate field measurements, and also create a time series of LAI for that point on the ground?

Many thanks in advance :slight_smile:

I expect you will find these forums supportive of people coming to remote sensing from other disciplines.

As you know, LAI is used for many different purposes. If you have a priori information about the nature of the vegetation in your region of interest and ground truth data, you should be able to improve on a generic LAI algorithm. If you find the standard LAI is not working for your purposes, you might want to look for ways to more directly apply optical remote sensing in your hydrology model.

The TIFF files should include geolocation metadata (e.g., they use GeoTIFF tags) so when you load the image in SNAP you should be able to view pixel values and coordinates (Pixel Value Tutorial). If you have a list of locations, SNAP provides a “pixel extraction” tool, which can used in the GUI or for batch processing. You can also run “gpt PixEx” in a terminal.

1 Like

Thanks @gnwiii , that worked perfectly!
Now i just need to overlay an Ordnance survey map in SNAP so i can locate a pixel on the ground…

Overlays are a bit of kludge in SNAP. You need to create a new “product” with both LAI and the ordinance map on the same projection and image dimensions. Particularly if your map is vector data overlays are more in the wheelhouse of GIS packages.

If you already use a GIS package you may find easier to reproject and save the LAI image (NetCDF4-CF is good if your GIS supports it) so you can do the overlay in the GIS package.

I’ve been involved in projects where the GIS users provided a .csv file with positions and I just added a columns with values of some remotely sensed data.

I think I’ve solved it the old fashioned way, using Digimap to find the co-ordinates of a point on the OS map, and then finding those coordinates on the layer when opened in SNAP gives me the index number for the pixel in the file. I can then use python to extract that index number form the array of the .tiff file and so make a time series for that pixel.
Lucky i’m only going to be using four ground sites!

Thanks for your help @gnwiii :+1:

Hi Matt,

I was involved with the development and testing of the SL2P algorithm used to make LAI products in the SNAP plugin.

I have just recently ported the SL2P algorithm to Google Earth Engin:
It allows you to process any subarea in the world without having to run SNAP etc. as long as L2 data exist. You can download results as geotiff to a google drive. I am awaiting review of the manual but the interface is pretty self explanatory.

You also wondered about performance of SL2P. There have been some validation studies that indicate performance varies with land cover. We found that is was generally acceptable for crops and pasture EXCEPT for extremely planophile crops Others have generally found it does not account for clumping in forests.




Thanks Richard,
I tried to use the link you sent, but I have to confess that I’ve never heard of EarthEngine! I certainly don’t have access to it :frowning:
thanks also for the paper, validating the data is going to be quite important for me as i will be using the LAI as a major input in my canopy model, and accounting for sub#pixel heterogenity will be a factor


@rfernand387 Hi Richard,
I’ve now got access to the Earth Engine, and tried to run the script. I think i’ll wait for the manual as i seem to be having difficulty with the interface. I’ve input the coordinates for my ROI and tried to export the result, it’s drawn a much larger area than i thought i’d specified and is still running after 25 minutes so i guess i got something wrong!!