Add CMS Global Mangrove Canopy dataset

[nilsleh]

This PR adds the CMS Global Mangrove Canopy dataset. It's a dataset of mangrove areas around the world at 30m resolution, where the raster displays pixel-wise measurements of either aboveground biomass, basal area weighted height, or maximum canopy height. An example for agb can be found below.

Since there are 116 different countries, I think it is desirable to let the user choose for which countries data should be retrieved. Likewise this should be possible for choosing one of the three measurements. However, I am not sure how to best do this with inheriting the Raster dataset class, where I need to only select possible filenames matching country and measurement to populate the index.

Example AGB:

[adamjstewart]

Since there are 116 different countries, I think it is desirable to let the user choose for which countries data should be retrieved. Likewise this should be possible for choosing one of the three measurements. However, I am not sure how to best do this with inheriting the Raster dataset class, where I need to only select possible filenames matching country and measurement to populate the index.

The easiest way to do this would be to always load all countries and allow the user to select a region by passing an ROI to the GeoSampler. It'll be a bit harder to let them select the countries.

[nilsleh]

Okay, I just thought that having a list of country names as an argument would be the most user friendly thing to do. With respect to selecting one of three measurements, should I then overwrite the getitem method that filters for the selected measurement?

[adamjstewart]

If you do want to support a country name argument, you should filter by filename during __init__ (after super) so that the sampler doesn't even know about regions that it can't sample from. Alternatively, if we only supported a single country at a time, you could have a filename_glob property set dynamically based on the value of self.country but it sounds like you'd prefer to be able to choose multiple countries.

[nilsleh]

After second thought, supporting only a single country at a time might actually be very reasonable, since individual country tiles can be very large and a user could still use the Union to combine several countries.

However, I have come across an inconsistency for three country names in the dataset file naming:

• there is NewZealand and Newzealand
• FrenchGuiana and FrenchGuyana
• CoteDivoire and CotedIvoire

Do you have a suggestion on how to deal with this? Write a warning in the documentation?

[adamjstewart]

What do you mean by inconsistency? Inconsistency between what and what?

[ashnair1]

Inconsistency in naming. For case 1 and 3, you could lowercase them via .lower() before doing comparisons right?

[nilsleh]

Sorry, inconsistency of naming a country for one of the three measurements (agb, hba95 or hmax95):

• agb_FrenchGuyana.tif, hmax95_FrenchGuyana.tif, but hba95_FrenchGuiana.tif
• agb_Newzealand.tif, hmax95_Newzealand.tif, but hba_95_NewZealand.tif
• agb_CoteDivoire.tif, hmax95_CoteDivoire.tif, but hba_95_CotedIvoire.tif

[adamjstewart]

Yeah, that's a bit harder. It's easy to do if you override all of __init__/__getitem__ but we would like to avoid that if we can because it's a lot of duplicated code and it's something we continue to optimize.

[nilsleh]

Yes that's what I thought as well. Maybe then just add a warning for those three cases and wait for them to correct the dataset? I have emailed the data center there.

[adamjstewart]

It looks like you have filename_glob set to select only a single country/measurement combo so there's no need to warn now.

[calebrob6]

(just cleaned up a little bit)

[calebrob6]

@adamjstewart, are your comments resolved here?