Greetings ODK folks!
I have an experimental crowd-sourced map that could benefit from your
participation (instructions on this page):
http://gnomapix.appspot.com/
The idea is for users running ODK Collect to get a few training points
about land cover. These points get added to a pool of globally distributed
training data that are passed to Earth Engine to classify a running median
of Landsat 8 imagery. The motivation was the lack of a consistent, science
quality, 30 meter land cover product. The premise that I started under was
that with a high dimension feature space and a high variance classifier,
addition of training data will increase overall map accuracy until the
learning curve (accuracy vs. number of training samples) flattens out.
What better way to get training data than to let the people on the ground
collect it? (Previous efforts used "visual interpretation" of the imagery,
with the attendant error implicit to that method). Assuming the map is
locally inaccurate, could a user go into their study area, collect a few
points and "fix" the map in their area of interest? That is one of the
questions the experiment is addressing. (I'm assuming that this will only
work up to a point. When the size of the training dataset becomes
sufficiently large, the marginal contribution of adding an additional data
point will decrease. It's not obvious where that threshold is.)
So whoever is feeling curious, or adventurous, or restless, or helpful ;),
feel free to grab a few points. In fact, it doesn't even really require a
special trip if, in your travels, you encounter a nice homogeneous area
that is representative of a land cover category. A description of the land
cover categories is reproduced below (from Friedl et al. 2002), for
convenience. My only requests are that you be IN in the polygon (at least
47 meters radius) when you collect a point, that you only upload the points
once, and that you not upload "test" points. (Earlier, I asked someone to
get a few points and then they informed me that they had not gone into the
polygon, but stood on the highway, looking at the polygon. That doesn't
work. Also, I found a few "test" selfies had been uploaded, which I then
had to clean out manually. Obviously, when/if the dataset gets large, that
won't work and some better filtering methods will be needed.)
I still need to add a learning curve to the website, a confusion matrix,
perhaps some other diagnostics and download capability. For now, it is
very preliminary. But it is fun to watch the map evolve as the "crowd"
makes it.
Feel free to try it out! I welcome any questions, comments, concerns,
issues, suggestions. Thanks!
-Nick
Land cover classes:
Evergreen needleleaf forests: Lands dominated by needleleaf woody
vegetation with a percent cover >60% and height exceeding 2 m. Almost all
trees remain green all year. Canopy is never without green foliage.
*Evergreen broadleaf forests: *Lands dominated by broadleaf woody
vegetation with a percent cover >60% and height exceeding 2 m. Almost all
trees and shrubs remain green year round. Canopy is never without green
foliage.
*Deciduous needleleaf forests: *Lands dominated by woody vegetation with a
percent cover >60% and height exceeding 2 m. Consists of seasonal
needleleaf tree communities with an annual cycle of leaf-on and leaf-off
periods.
*Deciduous broadleaf forests: *Lands dominated by woody vegetation with a
percent cover >60% and height exceeding 2 m. Consists of broadleaf tree
communities with an annual cycle of leaf-on and leaf-off periods.
Mixed forests: Lands dominated by trees with a percent cover >60% and
height exceeding 2 m. Consists of tree communities with interspersed
mixtures or mosaics of the other four forest types. None of the forest
types exceeds 60% of landscape.
*Closed shrublands: *Lands with woody vegetation less than 2 m tall and
with shrub canopy cover >60%. The shrub foliage can be either evergreen or
deciduous.
Open shrublands: Lands with woody vegetation less than 2 m tall and with
shrub canopy cover between 10% and 60%. The shrub foliage can be either
evergreen or deciduous.
*Woody savannas: *Lands with herbaceous and other understory systems, and
with forest canopy cover between 30% and 60%. The forest cover height
exceeds 2 m.
Savannas: Lands with herbaceous and other understory systems, and with
forest canopy cover between 10% and 30%. The forest cover height exceeds 2
m.
*Grasslands: *Lands with herbaceous types of cover. Tree and shrub cover is
less than 10%.
Permanent wetlands: Lands with a permanent mixture of water and
herbaceous or woody vegetation. The vegetation can be present either in
salt, brackish, or fresh water.
Croplands: Lands covered with temporary crops followed by harvest and a
bare soil period (e.g., single and multiple cropping systems). Note that
perennial woody crops will be classified as the appropriate forest or shrub
land cover type.
Urban and built-up lands: Land covered by buildings and other man-made
structures.
*Cropland/natural vegetation mosaics: *Lands with a mosaic of croplands,
forests, shrubland, and grasslands in which no one component comprises more
than 60% of the landscape.
Snow and ice: Lands under snow/ice cover throughout the year.
Barren: Lands with exposed soil, sand, rocks, or snow and never have more
than 10% vegetated cover during any time of the year.
Water bodies: Oceans, seas, lakes, reservoirs, and rivers. Can be either
fresh or salt– water bodies.