The GIS Forum
The GIS Forum is a community site for Spatial users to connect and grow

Selecting Raster Cell Size

rated by 0 users
Not Answered This post has 0 verified answers | 2 Replies | 2 Followers

Top 500 Contributor
Male
1 Posts
Oll posted on 08-29-2009 11:10 AM

Hi,

I have a question regarding cell size of Rasters.  I have a contour shape file and from this I've made a TIN.  I'm now making a raster from the TIN but was just wondering if there are any hard and fast rules about choosing raster cell size.  As a general rule I've heard that keeping columns and rows beneath 1000 each is a good rule of thumb.  Can anyone corroborate this or are there methods that can be employed to help you choose?

Any help, comments or suggestions greatly appreciated.

 

 

All Replies

Top 10 Contributor
Male
165 Posts

I am in no way a raster expert, but I do have a few things to consider.

1. Remember that the smaller the cell, the larger the file.  Keep in mind your storage space and the amount of processing you plan to do.

2. Take into account the contour divisions.  If you have 100 foot contours then it wouldn't make much sense having a 2 foot cell.  This would create a ton of unnecessary data. 

3. Determine your end purpose.  How detailed do you want your analysis.  If you are doing analysis at the parcel level then you would need a small cell; however if you are doing it at the county level then your cell size can be larger.

In all, it all comes down to a subjective decision.  I don't think there is a concrete answer.

Timothy Hales - The GIS Forum

Website - GIS Pathway

Blog - GIS Pathway Blog

Top 50 Contributor
Male
7 Posts

1000 x 1000 x 1-layer (aka band) is very small. For an unsigned 8-bit uncompressed raster this would be around 1MB and an unsigned 16-bit uncompressed raster will be 2MB, signed versions will be larger.  ERDAS and ESRI use the same dynamic range run-length compression, so integer data will be compressed in a lossless fashion. Floating-point data have very few repeating values so the run-length compression will not help much. (I discuss this in my blog: http://field-guide.blogspot.com/2008/05/native-and-fast-lossless-compression.html)

The key is not so much the file size, but to protect your accuracy and precision; but do not over protect what the data cannot support.  

If you have 20 foot contours based on NMAS, then your true vertical accuracy is a combination of the horizontal accuracy and the vertical. Do realize that contours are a derivative dataset, not an original source.

Look at the steepest slopes and determine what you need to protect. If the contours needing protecting are horizontally 20 meters apart, then you need a 10 meter resolution.

Page 1 of 1 (3 items) | RSS
Bookmark and Share
Creative Commons License

PLEASE SEE THE THE LICENSE FROM THE ORIGINAL SOURCE FOR ALL POSTS PULLED INTO THE COMMUNITY MASHUP BLOG SECTION