Manifold Community Site: Calculating population within circle

RE: Calculating population within circle

sitesatlas — Fri, 06 Jun 2008 19:33:49 GMT

I haven't been able to figure out how to solve a simple GIS problem with Manifold: how can I calculate the population within a circle? I have one layer containing the circle and another layer containing counties with population data. Of course, the problem is how to calculate the population when only part of a county is located within the circle.

Attachments:
Nappanee2.map

RE: Calculating population within circle

jburn — Fri, 06 Jun 2008 21:35:54 GMT

There are a bunch of steps to this, but once you do it, it should be pretty straight forward:

first, copy the area of your counties into a new field in the county drawing
"clip" your counties to your circle. If you look at your data, the counties that were on the edge should have an area which is smaller then your original that you copied into the field in step 1.
copy this "new" area to a new field (so you should have the original area in one field, the new area in another, and these should be the same for everything except for the counties that were on the edge).
divide your "new" area by the original. This should set a bunch to 1, and the ones on the edge to less then one.
multiply the population by this new number. As you can see, a bunch of the countie's populations should have remained the same, while some (the ones on the edge) should now have their appropriate portion of the population being shown.
create a field in your circle layer that is exactly the same as your population field in your counties (same name, type etc.)
right click on this population field in both drawings, and set the transfer rules to 1-N copy, N-1 sum.
open your map, and click "drawing - spatial overlay"
from here, you can select your from, to etc. and the total population will be placed.

Of course, this assumes that the population for your counties is evenly distributed;-) Also, you may want to change the projection (at least temporarily) of your counties to TM, just so that the areas are reported in meters so its a little easier to read.

As an aside, I got an answer of 21867353. I have attached your project back with the above done.

Hope that helps.

Attachments:
Nappanee3.map

RE: Calculating population within circle

Graeme — Fri, 06 Jun 2008 22:21:36 GMT

I initially thought this would be a straight-forward spatial overlay, just needing transfer rules correctly set. Basis of this was counties straddling perimeter need population proportional to area (same assumption of course), so transfer 1:n should be proportional using Area(I). n:1 should be the sum of all counties pop entirely within perimeter. However Spatial overlay using "Areas to touching areas" returns the sum of all the counties, apparently without computing the proportion of the marginal ones. Can't find explicit reference that area / area combinations don't honour proprtional transfer rules?

A workaround version 2. Create a boundary for your circle (transform / boundaries) put it in another drawing if you want, but no need. Making sure transfer rules are set as in the paragraph above, use the split transform to split your counties (make a copy and do it on that) using the boundary line. If you thematically format your counties beforehand using Population, you'll be able to visually see and verify the results. Now you can successfully use Spatial Overlay with the "Areas to containing areas" option. That gives a valid result.

As far as the initial method is concerned, I thought it would work anyone got any definitive insight or reference - can't pick anything from the forum or doco - most examples use a combination of areas and points or lines?

RE: Calculating population within circle

Graeme — Fri, 06 Jun 2008 23:40:18 GMT

Just reviewed the doco and in particular the topic "Intersection Overlays" '..This is a small, but sophisticated example that uses Transfer Rules, Split with, Clip with (Intersect) and Spatial Overlay commands'. I suspect using the "split with" step may be necessary in area-area spatial overlays. From the Transfer Rules topic in help: Spatial overlays are a different concept than the idea of using transfer rules to aggregate or to allot values when creating objects with transforms. However, it is a related concept in that aggregations or allotments are used to transfer field values between objects. It would be nice to see the required interactive "split" step built in and computed on the fly for Spatial Overlay, so maybe time for an enhancement request.

Omitted from your first post to mention that both your drawing's tables MUST have an identical field (name, type, structure) with transfer rules set the same in each. Think one of your drawings didn't have a population field.

RE: Calculating population within circle

danb — Fri, 06 Jun 2008 23:43:09 GMT

I would also have thought of the sparial overlay method, but recently noticed this with transfer rules set to proportional/sum [Area (I)] in the current build.

I am sure that I have used this method in the past as in my work I need to calculate quite a few spatial statistics where I proportionally assign a value based on the area of intersection. I sent a report to tech the other day in relation to this so if there is an issue I am sure they will look into it.

RE: Calculating population within circle

adamw — Sat, 07 Jun 2008 06:59:53 GMT

I don't think there is an issue.

Transfer rules affect values transferred from one or more source objects to one or more target objects. For the proportional transfer rule to split the accumulated value, you have to have more than one target object which is not the case in this example.

RE: Calculating population within circle

danb — Sat, 07 Jun 2008 12:58:10 GMT

Thanks Adam, I hadn't properly understood the question.

RE: Calculating population within circle

danb — Sun, 08 Jun 2008 13:11:50 GMT

Adam, I have just seen the response from tech. Thanks to you or whoever sent this for taking the time to explain how the proportional/sum [Area (I)] attribute transfer works. I was evidently barking up completely the wrong tree so just in case there are others who are as confused as me here is an official explanation ...

The fields are indeed being transferred by proportional assignment, but this works slightly differently than how you might expect it to work in this case.

The proportional transfer rule takes original value that it has to redistribute from the source object and splits it proportionally to values in the guiding field between target objects. If there is only one target object, it gets the entire value from the source object regardless of the value of the guiding field. There are never any comparisons between the values of the guiding field in the source object and any of the target objects.

We agree it might be desired to split values differently, using the proportion between the value of the guiding field in the source object and in each target object. We have inserted this into the wishlist for future versions of Manifold.

... thanks also for the insertion into the wishlist. This would be a really handy feature for me at least.

RE: Calculating population within circle

sitesatlas — Sat, 07 Jun 2008 05:49:04 GMT

Thanks a lot, everyone. The solutions are straightforward enough, but involve more steps than I expected -- I believe it can be done in one step in Maptitude.

RE: Calculating population within circle

adamw — Sat, 07 Jun 2008 06:55:01 GMT

One step? Here you go:

--SQL

UPDATE (

SELECT

[POPESTIMATE2007],

Area(ClipIntersect(c.[Geom (I)], r.[Geom (I)])) / c.[Area (I)] f

FROM [Counties] c, [150 mi Radius] r

) SET [POPESTIMATE2007] = [POPESTIMATE2007] * f

This reduces population values that you have to what they would be after the counties are clipped with a circle. To retain both the original values and the new values, create a new field called "POPREDUCED" and use the following variation of the query:

--SQL

UPDATE (

SELECT

[POPESTIMATE2007],

[POPREDUCED],

Area(ClipIntersect(c.[Geom (I)], r.[Geom (I)])) / c.[Area (I)] f

FROM [Counties] c, [150 mi Radius] r

) SET [POPREDUCED] = [POPESTIMATE2007] * f

RE: Calculating population within circle

RonHendrickson — Sat, 07 Jun 2008 10:17:24 GMT

Here is another method perhaps a little easier, but still requiring more than one step, but not too bad, especially if done within the GUI.

1. Check the transfer rule of your population column within Counties to make sure it is checked for transfer and the 1:N is Proportional according to Area(I).

2. Click on Drawing - Topology Overlay. The data dropdown should be set to Counties, the Overlay to the 150 mile Radius, and the method to Identity. Click OK.

3. The resultant County 2 drawing should be inserted into the map. Select the 150 mile Radius drawing circle, then click on Drawing - Transfer Selection. The Transfer Selection dialog should be set to Modify - Counties 2 Using the 150 Radius. Click OK.

4. The appropriate Counties and their fragments within the circle should be selected. Open a viewbot and ask it to give you the sum of the population column. Answer: 21,866,044.

A copy of your file with this done is attached. Hope this helps.

Attachments:
Nappanee2 topology overlay.map

RE: Calculating population within circle

sitesatlas — Sat, 07 Jun 2008 22:36:15 GMT

Thanks so much for going out of your way to help, everyone. This forum is great!

RE: Calculating population within circle

tomasfa — Fri, 04 Jul 2008 12:39:01 GMT

This thread was excellent for me. I Tested all the approaches and they all worked great. Adamw said in "One Step". I did a query to obtain the final result. Its the following (very simple). Note I used other table names, with some of my data. Its all the same as the examples above. DivNacT3 is the Counties drawing and ZI3 is the radius drawing.

SELECT Sum([PobReduc]) as TotalPob FROM [DivNacT3] d, [ZI3] z

WHERE Touches(d.[Geom (I)], z.[Geom (I)])

The result is the same as in allthe other approaches. My question is how can I instead of using 2 separate queries (first Adams query and then this one to finish the job) how can I merge both queries into ONE single query. I have the doubt, because Adams query is a Action Query. Is this an obstacle? Can we merge them into one query? Can an Action query and a Select query be merged?

Cheers, Tomas

RE: Calculating population within circle

adamw — Sat, 05 Jul 2008 03:15:44 GMT

No, you can not currently merge an action query and a select query. You can launch one then another using a script though.

RE: Calculating population within circle

tomasfa — Sat, 12 Jul 2008 17:17:41 GMT

Can you give me an example of a script joining 2 queries, an update and a simple select, please.

cheers. Tom

RE: Calculating population within circle

adamw — Mon, 14 Jul 2008 08:18:23 GMT

Here you go:

'VBScript

Sub Main

Document.ComponentSet("Query1").Run

Document.ComponentSet("Query2").Table.Open

End Sub

Query1:

--SQL

UPDATE [Drawing] SET [Selection (I)]=([ID]=5)

Query2:

--SQL

SELECT * FROM [Drawing] WHERE [Selection (I)]

RE: Calculating population within circle

AR_Rick — Sat, 05 Jul 2008 13:26:37 GMT

Another option is to use point-based population data. The US Census Bureau releases population data at the block-level and the points are distributed densely enough that you could probably do accurate population at the square mile in an urban or suburban area.

RE: Calculating population within circle

tomasfa — Sun, 06 Jul 2008 08:53:55 GMT

Thank you for the tips. I'll keep working on it.

Have a nice day!

Tomas

RE: Calculating population within circle

rheitzman — Mon, 14 Jul 2008 10:49:05 GMT

Pop Data: http://www.census.gov/popest/datasets.html Subcounty, All States

Populated Places: National Atlas: http://www-atlas.usgs.gov/atlasftp.html?openChapters=%2Cchpref#chpref

Place FIPS points on your map (and clip unnecessary points if desired.)

Build relation between table using FIPS Places.

This method yields 20.8m if you just look at the 'C' FIPS CLass codes (incorporated places - cities)

While it leaves out non-City populations it provides a finer resolution then using Counties.

This select query:

Select [ID] from [All_fips55 3]

where [FIPS_class] in ("C1","U6")

refines the selection to include non-incorporated places and yields 22.9m. Further refinement may be necessary for more accuracy.

FIPS Class definitions:

http://geonames.usgs.gov/domestic/fips55codedef.html#classu