I am seeking an repeatable remote sensing workflow/algorithm to solve for irrigable area per residential real estate parcel. Irrigable area is the area that a homeowner waters such as grass and shrubbery. I know that this process could be more accurate using high quality and expensive aerial imagery but I need to use imagery that is either free or cheap such as Google Maps or NAIP (National Agriculture Imagery Program). Google maps imagery is higher quality than NAIP but doesn't have infrared. NAIP has relatively low resolution, one meter, but it does have infrared, which is useful to discern different coverage types. I'm open to any solution you could come up with that would be cheap to repeat. By cheap to repeat, I'm thinking between $.02 and $.07 per parcel. Accuracy here is less important than being able to scale this process in a cost effective manner. I'm open to any technique including using a combination of aerial imagery. I would like this algorithm to produce results where 80% are within 10% of actual measured irrigable area or best possible.
Attached are images of similar analysis performed using high quality imagery.
I have data of actual irrigable area measurements by parcel for ~1,200 parcels in Contra Costa and Alameda counties in Northern California along with boundary files for those parcels. The irrigable area measurements are in an excel file and the boundaries are in a .gbd file format. I can provide NAIP imagery for both those counties as well. I will provide measurements for irrigable area of ~600 of the parcels and would like to see your algorithm's results for the other 600 to test it out.
You can download the files described above here:
[url removed, login to view]
This project is time sensitive and I would like to get it completed ASAP.
Thanks for reading and I can find somebody to work with on Freelancer.