Edit Raster Extent with NULL values

Question

Last week I posted a question in This Post but it remains unanswered. If anyone is up for the challenge, I would love to explore some creative solutions to my problem.

Some details are provided in my previous post, but basically, I would like to add a single row of NULL cells to a raster.

I have explored options such as creating a larger raster and merging the two, but am worried about this because of potential imprecision I do NOT want any resampling. Please let me know if there are other methods to try! Thank you!

E_haus · Answer

It is a 1-km^2 resolution global raster. There are 187 million land area cells, likely 510ish million cells total as you said.

To try to explain what I mean by "configurations", I am providing the below rough graphic, where I show a spatial subset of my global data and 40-km^2 windows overlays. The configurations are the different 1-km^2 cells that are aggregated within each "shifted" 40-km^2 window. I would like to perform this shift by adding rows of padding to the 1-km^2 raster, pre-aggregation. Does this make logical sense to you?

I would like to create a loop which: 1) adds a row of padding around my 1-km^2 raster, 2) aggregates this raster to 40-km^2 resolution, 3) exports the 40-km^2 raster values to points for further analysis, and then repeats by adding another row of padding to the 1-km^2 raster. Do you know of any way that this is possible? Thank you! 🙂

danretired · Answer

1 km or  km^2 global raster? is that 510ish million cells? is the unit width, height or area?

40 km or 40 km^2  (aka, 12.7 and 3.2 million cells

I am not sure what you mean by "configurations"

In any event, the 40 unit block is the only one that needs to be padded and the padding would be dependent whether you are doing a sliding-window or a block-window view (eg increment the subarray by 1 or by a block size like 10*10)

E_haus · Answer

Dear Dan,

Thank you again for your response. I really appreciate your time and effort in helping me.

My overall goal is to aggregate my 1-km global raster to 40-km resolution at a variety (>100) of different configurations. Then, I will use these aggregated rasters in a regression tree analysis and average out the results spatially to achieve a 1-km resolution resulting raster. The reason that I wanted to incorporate this padding with Null values is to create a different configuration of 1-km cells within the 40-km block, as the aggregation will take place from the bottom-left corner. Does this make sense? Please let me know if you would like further clarification on this.

As for the chunks (of the 1-km raster that is too large to be converted into a NumPy Array), ideally I would like to convert them back into rasters and mosaic them in such a manner that I end up with a single global raster with one row of Null values around the edges, not inbetween the chunks. Also, I would like this to align with the previous not-extended raster. Is this possible? Thank you so much for any other advice. I would be happy to DM you and share my data if that would be helpful! Thanks!