representation_rasterdb2v - Maps as Numbers Maps...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Maps as Numbers Maps Geographical Features Geographical Features s A GIS dataset (often referred to as a GIS layer) consists of Spatial (geometrical data capturing location Spatial and form of a geographical feature) and s Attribute data (textual information describing Attribute key characteristics of associated geographical feature) geographical s Maps as Numbers Maps s s s The data are stored in data structures and as files on storage devices (e.g. hard drives, USB drives, DVDs). Files can be written in binary or as ASCII text. Binary is faster to read, requires less disk space, and is typically more efficient (software allows us to read and edit binary), ASCII can be read by humans and a variety of software packages (e.g., word processing software) but uses more space. ASCII- American Standard Code for Information Interchange “Machine code” This is a significant challenge! Two Main Data Structures in GIS Two GISs have traditionally used either GISs raster or vector data structure for raster vector capturing geographical features. capturing s LEARN RASTER and VECTOR! s “Reality” is in Reality” the middle and alternative representations are to the left and right… and When selecting a data storage technique you should try to: 1. Maximize accuracy 2. Minimize storage requirements 3. Minimize processing time 4. Facilitate analyses Some objectives conflict When striking a compromise among competing objectives you must consider the particular problem you are trying to solve. What level of data accuracy is required? What are your available resources? Can you solve your problem given required accuracy with available resources? The raster data structure is based on a simple grid. simple s s s s One grid cell holds one attribute (usually). One Every cell has a value, even if it is “missing.” Every A cell can hold a number or an index value cell standing for an attribute (e.g., 300m above msl, 4 = residential development). msl, A cell has a resolution, that is- the cell size in cell resolution that ground units. All cells in a raster dataset have the same resolution. the Rasters are conceptually simple and often fast Rasters (computationally)... (computationally)... A grid or raster maps directly onto a grid programming computer memory structure called an array. structure s Grids are poor at representing points, Grids lines and areas, but good at surfaces. surfaces s Surfaces? What are these? Surface (fields) vs. objects Rasters are conceptually simple and often fast (computationally)... (computationally)... s Grids are good only at very localized topology, and Grids topology and weak otherwise. weak s s Grids are a natural for scanned or remotely sensed Grids data. data. Grids suffer from the “mixed pixel” problem. s Grids must often include redundant or missing data. Grids redundant s Grid compression techniques are often used in GIS Grid to reduce memory requirements (e.g., run-length encoding and quad trees). encoding Generic structure for a grid Generic Grid extent min x, min y Max y, max y Rows Grid cell Resolution Columns Figure 3.1 Generic structure for a grid. Binary (0,1 or present, absent) grid or using a matrix representation Redundancy Forest corn Urban 35 cells to capture THE feature “forest” A multi-valued (n=3) grid in matrix form multi-valued This is a “non-matrix” representation, more complex, not very efficient- but more flexible if additional variables need to be added. added. One grid cell holds one attribute (usually- but not always- you will see a refinement of this later). …jjust add ust columns to add data for other features… features… General idea- more efficient techniques exist. Problems with representing objects defined by points lines and polygons Not really a point “Line” has width Not stored as “features”, rather unconnected cells Cell Resolution Is Important Cell Which size do you pick? s Too large and you miss intra-cell Too variability variability s Too small and you record too much Too redundant information redundant s Same data, different cell sizes and very different results. Intra-cell heterogeneity is the problem. problem. Problems occur when assigning values to cells values s Different Different methods for assigning values can be used values s Again, problem is: different Again, methods yield different results methods The mixed pixel problem The Water dominates Winner takes all Edges separate WW G WG G WE G WW G WW G WE G WW G WG G E G E 5 dominates- e.g., a cover type of particular importance Winner takes all Different method = Different Different results Problems can occur is you need to change cell resolution resolution When might you use this? How about this? False connections Mostly water Mostly land Lost elements Vector map to raster cells: the effect encoding strategy (assuming “winner takes all” strategy). all” Storage considerations: When you increase resolution (or double the extent), you increase data storage requirements exponentially 2 n Double the resolution- four times the space 1 TB 10MB The 23rd cell counting rowwise from upper left to lower right row column Raster in matrix form is inefficient because of redundancy. Compression used. used. Value Count Variable 2 Variable 3 9 6 13 .23 6 19 5 .65 … … … … For integer grids, additional attribute can be maintained. Example? Why just integer? Bolstad Quadtree is a hierarchical (tree) representation of space. representation Quadtree 111 111 A 111 111 22B 2 B2 B BFG E GD 1 1 1 1 B D D G 22 2 A2 CD DC DC CC GF GD AB CC 33 3D 3 BA FG DG DG 0 A A D 1234 2 1 1234 0 3 4 A B C C From Bolstad Given a diagram like this, could you draw the associated quadtree? Each cell is 4km2, 12 cells= 48km2 Value Count Variable 2 Variable 3 9 6 13 .23 6 19 5 Raster measurement: area = .65 (cells * Σ … … … resolution2) resolution… 6 * 4 = 24 x 2 km/side = km perimeter 64 64 km perimeter Perimeter = Σ (external edges* resolution) Raster analysis using multiple layers Raster Integration requires all datasets (layers) to use a consistent raster framework i.e., cells are geographically “coregistered” (e.g same projection, same datum) and have the same resolution. A quick Review What are the benefits of the raster data structure? What are some of the disadvantages of the raster data structure? What is the mixed pixel/cell problem? What is the relationship between resolution and data storage requirements? What techniques are available to reduce storage requirements? Examples of the kinds of geographical features are best represented by the raster data structure are? The Triangulated Irregular Network (TIN) Bolstad 2002 Bolstad 2002 The Delaunay Triangulation Bolstad 2002 ...
View Full Document

This note was uploaded on 04/01/2012 for the course 044 005 taught by Professor Davidbennett during the Fall '11 term at University of Iowa.

Ask a homework question - tutors are online