{"id":121641,"date":"2007-07-03T19:36:00","date_gmt":"2007-07-03T19:36:00","guid":{"rendered":"http:\/\/www.esri.com\/arcgis-blog\/products\/product\/uncategorized\/choosing-an-appropriate-cell-size-when-interpolating-raster-data\/"},"modified":"2007-07-03T19:36:00","modified_gmt":"2007-07-03T19:36:00","slug":"choosing-an-appropriate-cell-size-when-interpolating-raster-data","status":"publish","type":"blog","link":"https:\/\/www.esri.com\/arcgis-blog\/products\/product\/mapping\/choosing-an-appropriate-cell-size-when-interpolating-raster-data","title":{"rendered":"Choosing an appropriate cell size when interpolating raster data"},"author":4351,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"open","ping_status":"closed","template":"","format":"standard","meta":{"_acf_changed":false,"_searchwp_excluded":""},"categories":[22941],"tags":[23191,23201,38471],"industry":[],"product":[],"class_list":["post-121641","blog","type-blog","status-publish","format-standard","hentry","category-mapping","tag-cartographic-concepts","tag-cartographic-design","tag-data-modeling"],"acf":{"short_description":"By Charlie Frye, Esri Chief Cartographer\n\n\n\n\u201cIf I want to create a digital elevation model (DEM) by interpolating from a point shapefil...","flexible_content":[{"acf_fc_layout":"content","content":"

By Charlie Frye, Esri Chief Cartographer<\/strong><\/p>\n

\"Choosing<\/p>\n

\u201cIf I want to create a digital elevation model (DEM) by interpolating from a point shapefile (digital terrain model or DTM), how can I select the most correct cell size given the original separation of the point file?\u201d We recently received this question via the Ask a Cartographer<\/a> Web page, so we checked with a couple of the product engineers on our Spatial Analyst development team and found that we didn\u2019t have a help topic addressing the subject. Moreover, the answer was not simple\u2014it varied depending upon why the DEM was needed.<\/p>\n

The consensus was that the cell size should be suited to what the data will be used for. However, another dimension of the Ask a Cartographer question was directed toward the distribution of the points that would be used to produce the DEM. So, we determined that before producing the DEM based solely on the use requirements some checking needed to be done to ensure the input points were suitable for producing the desired output. We also thought the approaches we would use in addressing this problem could be applied more broadly than just producing DEMs. This discussion can be extended to any continuous raster data \u2013 that is, raster data that vary smoothly across the landscape, like rainfall or elevation or statistical surfaces, like hot spot or density surface. It does not pertain to discrete raster data like land cover.<\/p>\n

Using the map scale of the final product to guide cell size<\/em><\/strong><\/p>\n

There are many uses for interpolated raster surfaces; however, We’re going to keep this discussion to maps and give a few of scenarios for map products that can be generated using DEMs hoping this provides some guidance for how to think about the solution required for other map products not described here.<\/p>\n

    \n
  1. Paper map at 1:25,000 scale, printed at 300 DPI (dots per inch)<\/strong>. We\u2019ve found that 10 meter DEMs barely work for these types of maps (larger map scales should definitely not use 10 meter DEMs). Although, if the main purpose of the map is to show terrain, We would recommend finding a DEM with a smaller cell size, perhaps 3 or 5 meters.<\/li>\n
  2. On-screen map at 1:12,000 scale, displayed at 96 PPI (pixels per inch)<\/strong>. The same 10 meter DEM can be used to a scale of 1:12,000. A smaller cell size would be better if the terrain is integral to the purpose of the map.<\/li>\n
  3. On-screen map at 1:5,000, displayed at 96 PPI<\/strong>. This requires that the cell size be no more than 3 meters.<\/li>\n<\/ol>\n

    Bilinear Interpolation Resampling Method<\/em><\/strong><\/p>\n

    All of the methods above require that the raster layer\u2019s display resampling be set to use Bilinear Interpolation (access this through the raster layer properties on the Display tab). When displaying continuous raster data on-screen or on paper, use the Bilinear Interpolation resampling method \u2013 this smoothes out the pixellated appearance of the data but only for display purposes \u2013 it does not change the data. This is a great way to get your data to look smooth but it should not be used to mask inappropriate data resolution.<\/p>\n

    You can alternate the display between Bilinear Interpolation and Nearest Neighbor resampling to easily see the pixilation. This is what we did for In scenarios #2 and #3 above.<\/p>\n

    \"Cell
    \nThis raster hillshade has an appropriate resolution (cell size is 3 meters for a map scale of 1:4,000) but it appears pixilated because it is shown using the Nearest Neighbor resampling method.<\/p>\n

    \"Cell
    \nThis raster hillshade has an appropriate resolution (cell size is 3 meters for a map scale of 1:4,000) but it appears pixilated because it is shown using the Nearest Neighbor resampling method.<\/p>\n

    It is inappropriate to use the resampling method to mask inadequate resolution of your raster data.<\/p>\n\n\n\n
    \"Cell
    \nThis map shows an appropriate level of detail for the raster data based on visual comparison with the base data used to show geographic reference (in this case two classes of roads). That the resolution of the raster data (hot spots) matches the resolution of the base data (roads) is obvious because identifiable cold or other spots can be seen along identifiable roads.<\/em><\/td>\n    		\"Cell
    \nThis raster, shown with the Nearest Neighbor resampling method, has a cell size that is too large in comparison to the base data. The mismatch in resolution is obvious because there is no ability of make obvious connections between the roads and the raster cells – the level of detail in the base data is much greater than the level of detail in the raster data.<\/em><\/td>\n    		\"Cell
    \nThis is the same raster as in the previous figure \u2013 it has a cell size that is too large. But because it is shown with Bilinear interpolation, the raster data might appear to be at the same resolution as the base geographic data. This is an inappropriate use of the resampling method to mask the coarse resolution of the raster data.<\/em><\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n

    Checking for an appropriate cell size<\/em><\/strong><\/p>\n

    It is important to verify that the point distribution of the input data is adequate for the output raster resolution that you want to use. This relates back to the requirements for your product \u2013 it could be for display purposes only, this but also holds for analysis.<\/p>\n

    In order to check the cell size, we suggest producing a DEM at the cell size you anticipate needing, then evaluating the appropriateness of this raster by checking to see whether there are enough nonzero values. To do this, you can use the Point to Raster tool<\/a> (available with ArcInfo license or with ArcView or ArcEditor licenses when either the Spatial Analyst or 3D Analyst are available).<\/p>\n

    To do this, first set the resolution (cell size) of the output raster to what you think is adequate for your purposes, then use the Point to Raster tool to check the output to see how many pixels have \u201cno data\u201d values.<\/p>\n