Cost Weighted - Raster Question

[JerryGarcia1]

Hello,

I am doing an accessibility study and I am using roadways as my "Friction" raster.  I am then using the cost weighted tool in SA to create a walkable area around set points. I reclassed the results so there is a higher value for being closer to the point and it tapers down to the edges. So if one lived next to the point they would get 10 while if they lived on the edge of the walkshed they would get 1. The problem I am having is that I end up with just the roadways scored and I need a raster that is "filled in" with the scores  (not really sure how to articulate this, kind of like the polygon that is generated in network analyst). I can think of some ways to do this in SA but it would be in Euclidean distance and that is not accurate when measuring walkable access. 

My other idea is to use network analyst to create a walkable polygon, create negative buffer layers, merge them together, convert to raster and reclass with the scoring system I created.

Any help will be much appreciated!

[JerryGarcia1]

Thanks for the advice Bill. I enjoy reading your posts.

I tried that method yesterday and if I start including off road walking it becomes more or less an euclidean distance. I also tried playing with the impedance ratio between on road and off road and got something that is close to what I wanted but really it is not completely accurate. The size of the walkshed needs to be based on the road ways. This is for a dense city where almost all walking will be done along streets. I thought there may be a tool or some process that could fill in the raster scores based on the scoring along the roadway. Do you think if I used the impedance ratio of off road and on road that I thought was somewhat accurate and then used a walkshed created by network analyst as an analysis mask it would be an accurate representation of walkability?

I should point out that I intend to allow walking across some parks and open space and make certain highways, railways, and water features impassable.

[JerryGarcia1]

Happy New Year Bill.  Thanks for your help. I plan on incorporating your ideas into this study.  The reason for the �??filled�?� in raster was so I could create a walkability score for each census block group. My boss and I want to apply the scores from the cost distance algorithm into the CBGs beneath each walkshed so we could create a demographic make up of those with  and without walkable access to healthy foods, health centers, ect�?� I plan on using this method for several studies that involve walkability. There will be some studies where these scores will need to be aggregated to council districts, zip codes, ect. Applying just the scores from the roads into the CBGs is just not accurate, areas with more roads will always score higher which is not always true. 

I just tried the Euclidean Allocation and it worked great! Exactly the output I wanted.  Thanks!