pcplantdb

[Richard Morris <webmaster@pfaf.org>]

Sean Maley wrote:
> Zip codes come with latitude and longitude, but many
> international postal codes don't.  However, the
> formula calculating distance is fairly simple.  Even
> better is if we can get GPS savvy users.  The GPS not
> only gives better accuracy to lat/long than postal
> codes, but also includes altitude.  Throw in some
> weather gadgets and an accurate micro-climate picture
> can be developed over time for any given locality.  On
> the exreme end of sensor automation, we get pretty far
> out of scope for PIW, which I'm still not sure where
> the real scope ends anyway.
I've been mulling this over a bit and heres a few ideas about how
i'd do it. Assumes superb data resources, infinite coding time etc.

First we need to seperate positions from properties. We really have two 
things:
        1) physical locations:
                lat-long coordinates,
                post code/zipcode regions
                names of states, countries regions
                bio regions
                native ranges of a particular plant
        2) Properties about these regions:
                max/min temp
                altitude
                plants who's native habitat is in specific place
                properties of a plant in that region (ie when it typically 
flowers)

Physical locations have two different types
        spot coordinates, define a single location (lat/long).
        regions, define an area.

Spot coordinates are easy to represent, just use two
coordinates, latitude and longitude seem obvious.

Regions can be represented as a closed polygon whos boundary
is the boundary of the region

        * --------------*
        |                \
        |                 \
        |                  *
        *                 /
         \               /
          * ------------*

the corners (or verticies) of the polygon are just spot coordinates.

Hopefully we can get data from the web for most of the regions
we're interested in (countries, postcodes).

There is a lot of fun stuff we can do with this data
        find which regions are inside which other region
        find overlapping regions
        find which regions contain a particular spot
                cordinate
theres some well know algorithms for doing the above.

In terms of searching a user could
        input the name of a particular region
        find all the regions which are contained
        within that region
        find all the plants in the region
or
        input a particular spot coordinate
        find regions containing that coordinate
        find plants which could grow in that position

Once we have regions sorted is then working out the types of 
relationships we might have with the regions. Trickiest
is features which depend on both the plant and the region
i.e. max height of plant in that position.

The above does make mean that we need to find a lot of data giving the 
boundaries of regions, but it seems like there is some open content data 
on this about.

Reactions?

Rich