Education: Collaborative tree mapping

Discussion in 'Plants: Conservation' started by Jean Weber, Sep 1, 2011.

  1. Jean Weber

    Jean Weber Member

    Messages:
    2
    Likes Received:
    0
    Location:
    Malaysia
    Most of the lowland forests are now managed by man. If many introduced trees are chosen for their proven adaptability and benefits, management for the conservation of local species education and cultural identity need to be improved.

    Collaborative tree mapping would give a visibility of the knowledge of trees in urban and sub-urban areas in order to bring biodiversity in the focus for decision makers. Mainstreaming biodiversity across the society is a goal of the Nagoya convention. I believe that we can only conserve what we know and even only if we can see what we know!

    To develop this idea, I would like to invite you mapping trees you know in your gardens and cities. It doesn’t matter if the trees are rare or invasive. Trees are recorded on a map with botanical name and can be edited. It is also possible to record a given area using polygons. Tree can be found by searching their names.

    I put a video-tutorial online http://pericopsis.org/trees/videotutorial.php. For each tree you map you can, if you desire, record a link to your own webpages.

    Thanks for your interest,

    Jean
     
  2. sgbotsford

    sgbotsford Active Member 10 Years

    Messages:
    167
    Likes Received:
    8
    Location:
    Rural Edmonton Alberta area, Canada
    A variation on this occurred to me.

    At the forestry level statistical class info is more important. Ok, you know that there is a sugar maple at 114.001225 W 54.002349 N. But more significantly you want to have an idea of how many sugar maples there are in the 114 54 grid square, how large they are.

    Here's my idea:

    Program your GPS for a transect -- a more or less straight path across a representative chunk of your landscape. Put in a waypoint every 100 meters.

    Now you go for a walk with your brand new spiffy digital camera with zooooom lens.

    At every waypoint you take a series of pictures.
    Set one. 8 pix starting from north, and going around your point parallel to the ground. Zoom is set so the pix just overlap.
    Set two. 4 pix at a 45 degree angle into the canopy.
    Set three. 4 pix at a 45 degree angle down toward the forest floor.
    Set four. Pix of anything of interest within 20 feet. Blooms, bugs, tracks, scat.

    If set one was done with a filter that was a survey prism, you could quickly extract basal stem area.

    I would add also that the walker add waypoints at any edge. Creek, forest edge, right of way, abrupt change of slope.

    Analysis. Most GPS can be set to record a breadcrumb trail with time stamps. So you know exactly where you are every 60 seconds. All digital cameras now record a time stamp on the image. So set your camera to GPS time (accurate to better than a microsecond...) I know of at least one photo package that will match the breadcrumb data to the picture, and write GPS info into the picture file as metadata.


    While computers are not up to it just yet, I bet within a few years that automated analysis of the photos could identify 95% of the visible vegetation. I suspect that facial recognition software would work for this with some modification.

    This would be a fast, relatively cheap way to do surveys to monitor environmental change. It would also be a way to inventory sites pre and post reclamation.

    Let's see: 30 pix per waypoint. With auto focus, auto exposure 1-3 minutes depending how many pictures you take in the last set.

    100 meters at 2 mph is about 2 minutes. Yes, you will go faster on grassland. Yes you will go slower in alder thickets. So 3-5 minutes per waypoint. 12-20 waypoints per hour. With edges and whatnot call it 14 waypoints per kilometer. So a transect of 6-10 km could be done in a day.

    at 30 pictures per waypoint, and 14 waypoints per kilometer, you get 400 pix per km. So 4000 per day. At an average of 6 MB each (Assumes 6 megapixels, raw encoding) you get 24 GB per day of activity. This is well within the capacity of a single flash card. You will likely have to take at least one spare battery pack with you.

    Long term storage: Hard drives cost $80 per terabyte, or about 12 cents per gigabyte. To put this onto serious, redundant backup will approximately triple this. (Figures for running 12 disks in ZFS Raid Z2 with hot spare. This cuts your storage in half, but the system is robust enough to lose any 3 of a set of 6 disks, and rebuild onto the hot spare. Even at a buck a gigabyte, storage is on par with the gas to get you to the trail head.

    For logistics reasons, it may make sense to do triangular routes, so you end up back at your transportation. Or you have a van full of people, and you drop them off in the morning, and pick them up in the evening.


    For repeated surveys to monitor change, I'd suggest timing to a stable ecological event. E.g. X days after last hard frost, after poplar bud break, after ice breakup, after first chorus frog activity. This increases the chances of getting comparable development of the forest floor, and of forbs in general.

    For this to be of maximal use, some group needs to define a standard.
     
  3. Jean Weber

    Jean Weber Member

    Messages:
    2
    Likes Received:
    0
    Location:
    Malaysia
    While computers are not up to it just yet, I bet within a few years that automated analysis of the photos could identify 95% of the visible vegetation. I suspect that facial recognition software would work for this with some modification

    Do you have any example where this works for a few simple cases?

    Your proposal of making a transect is good.
    I would suggest a kind of android application that draws a virtual square (ex 8x8 m) with one of the corner fitting with your own position. From your place you would shoot on trees, dead wood, etc using a range finder coupled with a compass and a hypsometer http://www.forestry-suppliers.com/product_pages/view_Catalog_Page.asp?mi=3872
    The rangefinder will transmit the positions to the application using blue tooth and voice recognition app would associate names and features (tree names, height, diameter etc) to the recorded points

    Jean
     
  4. sgbotsford

    sgbotsford Active Member 10 Years

    Messages:
    167
    Likes Received:
    8
    Location:
    Rural Edmonton Alberta area, Canada
    Examples of computer recognition/scene analysis:

    The military is really big on IDing things like tanks hidden in brush.

    Property tax authorities (cities & counties mostly) are using image analysis software coupled with aerial photography to check for building permit compliance

    Faces are awfully similar especially with family faces. Most of it is done with ratios and angles.

    My bet is that for tree identification that bark would give them away 95% of the time on mature trees. Forbs that weren't in bloom would be tougher. I remember bringing a yellow flower into class. I had been unsuccessful at keying it. "That? It's a DYC!" "DYC?" "Damned Yellow Composite!" or bird watchers with their LBJ's -- Little Brown Jobbies.

    For climate change shifts, it's going to be the annuals with fast moving seeds that tell us first. Trees like oaks that drop their acorns directly under the tree don't move to new territory fast. Things like thistle, fireweed, that float, or aster, burdock that cling will move further.

    If I were doing this as a reclamation program part, rather than doing a grid within a kilometer, I'd do a path that crossed from disturbed to undistrubed land. I'd try to set up a path that would allow me to cover adjacent path chunks at long intervals.

    Suppose I had the summer job of doing a photo survey on a 10 km grid square. Now one way to do it would be to do one 1 km2 each day. 100 days later I'm done. But south east km may be very different from north west km. So instead, I start in one corner, do a straight run north 10 km. Probably camp overnight. Next day move 5 km over and come straight south.

    Day 3, and I pick as spot 2.5 km east of my first start.

    In a 10 km grid on 100 m spacing there are 100 tracks 100 m apart. To save typing assume there were only 16 tracks.

    1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
    A E C G B F D H A' E' C' G' B' F' D' H'

    The idea is that by the time I get to E, it's enough later in the season that things have changed. we'll spot things we missed on the A path. This also gives you more variety in a day. Plus you don't have days with a 10 km walk each way just to commute to where you are working.

    Can I point to a specific example in botany? Not really. However Delta and Intkey with their analysis by multiple key characteristics would certainly form a building block.

    The system you propose would be far more accurate, but MUCH harder to teach someone to use, as they have to be proficient at the local botany. They also have to be good at estimated percentages, something that takes some training and practice.

    I suspect that it would also be much slower. I'm betting that 10 minutes at each way point would be common.

    Your way would do a better job of finding WHAT was there. I think it would do a poorer job of HOW MUCH was there without a lot of counting and estimating. That's really hard to do consistently in late afternoon when the horseflies and the mosquitoes have found you.

    Which means that both approaches have merit.

    From my time setting orienteering courses, having too many gadgets is a real pain. I carried a sighting compass, a GPS, a notebook, tags, nails for the tags, a hammer for the nails. Spare batteries for the GPS. Lunch. The notebook had a copy of the map, blank pages for new tags, a database printout of the old tags for verification, My compass was leashed to the notebook. My GPS was leashed to my jacket through a button hole. I carried a lot of the misc stuff in a satchel at my side.

    One of the advantages of doing it entirely photographically is that given someone who is comfortable with a digital camera and a GPS, I think you could train them in half a day. It would also mean that in reasonable terrain (I don't need flippers or pitons) I could shoot a set of pix within 71 meters of every spot in a square kilometer in a day. (13 kilometers) This would be the cats pajamas for doing 'before' and 'after' for things like habitat restoration after mining.

    But it still means playing with two gadgets all the time. Using an android or iPad or equivalent brings it up to 3. Yes GPS is built into most phones, but it's not very good GPS. They don't work worth a damn under forest canopy, nor in ravines. And most phones are also cameras, but they are awful cameras. So you can reduce the gadget count by making compromises.

    If you want more accurate 3d spacing, repeat the imaging process, retaking each shot for the horizontal plane 2 meters to the left of where it was shot the first time. (Would require carrying a stick to do the measurement each time.) So the first image set is at a given point. the 3d plotting set is from a circle 2 meters away from your start point. So the North facing pic is taken 2 m west of the center, the east facing pic is taken 2 meters north of the center.

    If you wanted the best of both worlds, use teams of 2. The photographer would do the every hundred meter thing. The plotter would do it every 500 meters. This keeps them withing 300 meters of each other, which gives you a safety margin if they have radios (ANOTHER )(*^(*&*( gadget!!)

    Or they would do the transect in opposite directions which could simplify the transport logistics.

    We now have 3 approaches:

    The first one was in effect "Inventory all of the significant trees"
    The second was in effect "sample and shoot now, identify later."
    The third was in in effect "sample, measure, and identify now."
     

Share This Page