search

PalEON-Project / stepps-calibration

STEPPS pollen-veg calibration model code and paper.
0 stars 1 forks source link

Potential domain size #10

Closed andydawson closed 9 years ago

andydawson commented 9 years ago

In the model we define a potential domain, which defines the area that pollen can disperse to/from. This may or may not catch all the rare extra-long dispersal events, but should be large enough to catch almost everything. Right now we define a rectangle that is larger than the area defined by the UMW states. We can think about each core site as being centered in this potential domain (imagine a source/sink area for each deposition site), and for sites near the boundaries, we will have a larger number of cells for which we do not have vegetation data.

I think since our dispersal kernel is symmetric, we may want to choose a square (or circular-like) potential domain. We talked about going a fixed distance from the edge of our state boundaries, but I'm not sure this makes sense (since pollen doesn't know anything about geographical borders).

What do you think @paciorek ?

This is related to #3 , which I'm closing.

paciorek commented 9 years ago

I thought the idea of the fixed distance from the edge is that for each core site, we want the potential domain to include all locations within some minimal distance. Since all sites are within the state boundaries, this ensures this.

Regardless of square, circle or rectangle, we always have that the source area for different sites is different because sites vary in how far they are from the boundary. Right?

On Thu, Jun 11, 2015 at 9:05 AM, Andria Dawson notifications@github.com wrote:

In the model we define a potential domain, which defines the area that pollen can disperse to/from. This may or may not catch all the rare extra-long dispersal events, but should be large enough to catch almost everything. Right now we define a rectangle that is larger than the area defined by the UMW states. We can think about each core site as being centered in this potential domain (imagine a source/sink area for each deposition site), and for sites near the boundaries, we will have a larger number of cells for which we do not have vegetation data.

I think since our dispersal kernel is symmetric, we may want to choose a square (or circular-like) potential domain. We talked about going a fixed distance from the edge of our state boundaries, but I'm not sure this makes sense (since pollen doesn't know anything about geographical borders).

What do you think @paciorek https://github.com/paciorek ?

This is related to #3 https://github.com/PalEON-Project/stepps-calibration/issues/3 , which I'm closing.

— Reply to this email directly or view it on GitHub https://github.com/PalEON-Project/stepps-calibration/issues/10.

andydawson commented 9 years ago

The potential source area does not change from site to site, but the portion of this potential source area for which we have data changes from site to site (maybe this is something like the measured source area?). For each core site, we imagine a source area centered at that site. So the constant C that is equal to the sum of the weights for the potential source area is fixed (constant across all sites), given a parameterized kernel.

Does this make sense?

paciorek commented 9 years ago

I think we're going to have to do this by talking. Let me know if this should happen before our call next week.

My main confusion here is that if we have a circle centered at each site and calculations are done based on the portion of the circle for which we have veg data, what is the role of the large rectangle? Sorry, I know we've discussed this multiple times.

On Thu, Jun 11, 2015 at 11:17 AM, Andria Dawson notifications@github.com wrote:

The potential source area does not change from site to site, but the portion of this potential source area for which we have data changes from site to site (maybe this is something like the measured source area?). For each core site, we imagine a source area centered at that site. So the constant C that is equal to the sum of the weights for the potential source area is fixed (constant across all sites), given a parameterized kernel.

Does this make sense?

— Reply to this email directly or view it on GitHub https://github.com/PalEON-Project/stepps-calibration/issues/10#issuecomment-111229929 .

andydawson commented 9 years ago

Following up after our previous conversation (for the record). We decided to have the potential domain be circular with a radius of 700 km.