Sitting in a meeting Friday, I was struck with a problem that needed solved quickly: we have a land cover classification being done by a third party for which we will have a team of more than a dozen entities doing some validation of the data. We could print out paper maps for them, but that seems labor intensive and expensive. So, I proposed we put together a site that allows each entity to go through each of the validation points one-by-one and populate the human-interpreted classification for each.
The twist in the problem space is this– we have less than a week to deploy the solution, it needs to have authentication baked in, and it needs to be easy to use.
Enter geojson.io, “A fast, simple tool to create, change, and publish maps”. I’ll say. Let me walk you through. First, we can choose to login using github, so authentication is baked in.
After adding the points, I can add fields to my table. That should probably be in quotes– this is JSON, so there are only key value pairs, no table per se, but it’s quite nice that the interface treats it as a table with nulls when a column is added.
Now, we’ll populate that id field. So far I’m manually constructing the dataset, but what will happen in the final version is that I’ll generate a GRTS drop for the even “random” sampling goodness that comes with that approach. I’ll do that in R, output as shape, translate to GeoJSON, and upload… . But in the mean time, let’s add a column for identifying land cover:
Now I can send my GeoJSON to my end user, and have her click in each of the points, zoom in to identify what she sees, and code it in the land cover column, and then share it back with me.
Now I just need to clone, host, and change out the base layer in the project with a leaf-on aerial… . Hopefully that’s as smooth a process.