If you're a crew leader, you're probably very accustomed to the tedious task of creating a number of polygons on a block before planting, then calculating how many trees are needed for each polygon, then creating pins for caches with data attached to the pins to show the size of the piece and the predicted number of trees. It's a fun task, until you've done a few hundred blocks. Then it gets to be tedious and boring.
What if you could automate this task? And even better, have KML files generated for you that match your traditional style for cutting up blocks?
This is my suggestion (I haven't tried this yet):
1. First, collect block maps from about two dozen different sources. Don't collect them all from a single contract/source, because having a variety of underlying map designs will help improve the skill of your model when running into new designs.
2. Next, decide on your "parameters" or preferences & limits. For example, I might decide: (a) No polygons smaller than 1200 trees; (b) No polygons larger than 4600 trees; (c) Make sure that 60% of polygons are between 1800 and 3600 trees, with 20% between 1200 and 1800 and 20% between 3600 and 4600; (d) If the map has a prescription that indicates multiple species, ignore that and simply calculate the number of trees for each piece, based upon the size of the polygon in hectares multiplied by the target density in stems/hectare.
3. Spend a few hours creating sets of polygons which roughly follow the above rules, going through one block map at a time and manually creating polygons.
4. Create caches based on where you think the best locations are. Use some sort of standard naming convention for cache pins, ie. the Aviation alphabet, a numbering system, etc. For each pin, include the metadata in the label which includes the size of the piece (polygon), the target density, and the trees required for the piece.
Once you have all of these set up, go into your AI (such as Work if you're subscribed to OpenAI, Co-Work if you're subscribed to Claude, or Spark if you're subscribed to Gemini). Tell it that you want it to learn your preferences for generating polygon and cache pin data from block maps, and that you're going to feed it a few dozen examples of KML files that you created based upon studying specific block maps. Tell the AI that you don't want it to analyze yet, but you're going to feed in all the maps and KML's first. Make sure that the file names are designed in a way to ensure that the AI correlates them properly, ie. "Block 01 map.pdf" and "Block 01 layout.kml"
After Work has all of the files, ask it to correlate all block maps with the relevant KML's that you created, and analyze your patterns in creating polygon layers, because you want it to be able to do the same thing in the future.
Once it has analyzed all of your files, give it a fresh map and ask it to create a block layout kml for that block, with a specific set of preferences for that exact block (which may vary slightly from what you were mentally aiming for above). As an example, use this prompt:
Now that you see the patterns that I use when creating Planting Piece Layout KML's, I want to give you a new block map and have you create a layout for me. Keep the following rules and preferences in mind:
1. Rule - polygons must cover the entire block, and must not overlap each other.
2. Rule - the target density for this block is XXXX stems per hectare.
3. Rule - no pieces/polygons can be smaller in size than that which would create a 1200-tree piece, based upon the target density that I just provided.
4. Rule - no pieces/polygons can be larger in size than that which would create a 4400-tree piece, based upon the target density that I just provided.
5. Rule - for each polygon that you create, name the polygon according to a series starting with "Piece 01," then "Piece 02" and so on. In addition, the size of the piece (in hectares) and the number of trees required (at the target density) should be included in the name of each polygon.
6. Rule - for each Cache that you create, add a name using the aviation alphabet going through in sequence, ie. the first cache should be called Alpha, the second cache should be called Bravo, etc. In addition to the Cache name, the name of the pin should include which polygons it services, and the total number of trees required, ie. "Alpha, services pieces 07 and 11, needs 7,941 trees."
7. Rule - cache pins can only be created along the border of a piece. In addition, cache pieces should always be located on a road or trail.
8. Preference - I'd like approximately 20% of the polygons to range between 1200 and 1800 trees in size, approximately 60% of the polygons to range between 1801 and 3600 trees in size, and approximately 20% of the polygons to range between 3601 and 4400 trees in size.
Now this is starting to get complicated. I'm approaching this concept theoretically, without having tried it yet.
Hold these thoughts. I'm going to have to create a set of block maps and piece layers myself, and see how the AI model responds.
I'll revisit this post as soon as I get this done.
