attractiveness mapping
DESCRIPTION
Attractiveness Mapping. Modeling Land Use Preference. Outline. General Concepts in Attractiveness Modeling Refresher on Basic Raster Analysis Technical Implementation Issues. General Concepts & Methods in Attractiveness Modeling. Identify abstract best/worst conditions - PowerPoint PPT PresentationTRANSCRIPT
Attractiveness Mapping
Modeling Land Use Preference
Outline
General Concepts in Attractiveness Modeling Refresher on Basic Raster Analysis Technical Implementation Issues
General Concepts & Methods in Attractiveness Modeling
Identify abstract best/worst conditions Find geographical correlates for key factors Develop Factor Maps “Weight and rate” to Generate Single Output
Best Case / Worst Case
Identify abstract best/worst conditions Important perspectives
Legal Physical (natural amenities or dis-amenities) Fiscal Social services
Roleplay developers potential customers citizens
Finding Geographic Correlates Often, data you might most want are not available
Example: we have no land cost data layer
Two options A) Ignore the factor entirely B) Generate a reasonable spatial approximation
If Option B, how? Generally, use qualitative and relative (versus
quantitative or absolute) factors E.g. likely land cost = low, medium or high
vs. land cost <= $133,456.34/ha Use proximity when appropriate
All other things being equalNear existing expensive might likely be expensive
Developing Factor Maps
Factor Maps Express the main decision criteria spatially
Example: distance to nearest school, land price, travel time to employment
Should be in common vocabulary/units/scale Here, since output is given 1-9 scale, use that
Depth versus Breadth and Spatial Autocorrelation Better 3-5 spatially un-correlated factors than more As in statistical regression, better to have few but solid
explanatory variables If sub-factors are needed, organize hierarchically
Example: Good Views = Ocean Views or Mountain Views, Ocean View = …
Raster GIS in ArcGIS
A Spatial Analyst refresher
Raster Data Model
Rasters are conceptually similar to pixels Instead of coding visual appearance as
red/green/blue, encode spatial data Common Types of Rasters
Categorical e.g. land use code where 1 = urban, 2 = suburban
Continuous and representing measured data e.g. elevation, where 1 = 1 meter above sea level, 2
= 2 meters, etc. Continuous and representing preference
e.g “attractiveness to urban development” where 1 = least and 9 = most attractive
Operating on Rasters
Quick and easy for the computer Generally, a set of raster GIS layers are
designed to “line up” Same overall spatial extent Same raster grid cell size
Most operations involve simple algebra Known as “map algebra” (Tomlin) Just as 1 + 1 = 2 111 + 111 = 222
111 111 222
Operating on Rasters 2
“NoData” critically important concept to understand in raster
analysis *despite* the name, “NoData” in many cases
represents areas which the user wishes to treat as transparent, empty or background
Example In creating raster layers from vector features, the areas
on the map between features are coded as “NoData” In this case, “absent”, “empty” or “background” are
more appropriate conceptual meanings A systematic measurement *was* made and the
mapped feature was *not* found
Why Worry About “NoData”?
In ArcGIS map algebra, NoData is a special value NoData + anything = NoData
Adding two maps in Spatial Analyst Get only areas which didn’t include NoData in either
map For order-dependent overlay, need “Merge” command
In this case, “NoData” in top layer = transparent Spread command (for distance calculations) only
expands into NoData areas In this instance, treated as “background”
Spatial Analyst Review
Enable the Extension Show Toolbar Adjust Options
Working directory – local writeable Set Common Extent (bases civitas nova/study
area) Cell Size (25m to start) Optionally, set mask to study_area_25m as
well
Spatial Analyst Basics
Conversions
Reclassification “Selection” in raster Aggregation
Vector to Raster Vector files, including CAD files, can be directly converted into
raster
Only selected features are converted Remember to clear selection first Useful for converting only features meeting particular criteria
(select first, then convert)
A Raster Value Column must be specified These can be numeric (leading to predictable result) Can also be text (leading to automatic generation of raster
code values based on sequential position) If you don’t have an appropriate pre-existing column, can
simply create one Example: create integer column named “one” with value
calculated to equal “1” for all features
Conversion
Already have spatial geography you want, just need to extract & reformat
Example: Have existing urban areas mapped from CAD, want all
of these as a grid of “most attractive” with value of 1 Create a new column and calculate an appropriate
code value for it e.g. create new integer column named “attractiveness” Calculate value of “attractiveness” = 1 Convert vector to raster using new column
Basic Spatial Relationships
Overlay Sites inside “rural” zoning Sites outside of city limits
Proximity Adjacency Near Far
Arithmetic Overlays
Can Use “Raster Calculator” NoData +-/* Anything = NoData If NoData causing “dropouts”
Reclass NoData to “0” Alternatively, can use “isnull()”
Example Map Algebra Overlays
Which urban areas are over 5% slope? Have urban areas vectors, percent slope from
base GIS Strategy:
isolate desired components into “mask” maps (desired = 1, background = NoData)
Add maps in Map Calculator
Proximity Relationships in Raster GIS Proximity
Adjacency hard in raster – usually better to develop
appropriate “near” criteria Near/Far
Can be absolute or relative Within 100m = near? For relative, prior analysis can calculate distribution
(Relatively) near primary schools could be based on standard deviations of existing distances to primary schools
Simple Proximity
To start, we’ll use Euclidean Distance Aka “as the crow flies”
Later, Cost Distance Requires transit data grooming Much more time consuming
Points to remember Euc Distance spreads only into “NoData” cells Objects are “0” distance from themselves Buffers in raster usually a 2-step
Distance / Reclass
Technical Implementation Issues
Summarizing existing conditions Categorical Variables Continuous Variables
Expressing factors along equal scales Using reclass or slice
Weighted overlays in ModelBuilder General operation Special cases
Summarizing existing conditions
Categorical Variables Usually can use zonal statistics run on land use
mask Code land use as “1” Run zonal stats against land use Careful with “area” column sums – often wrong
Continuous Variables Table summary stats ok Can do in interface and record manually Or can run with output to tables
Expressing factors along equal scales
Generally need to convert arbitrary and mixed units into evaluation units
Dealing with ranges First, exclude unreasonable values Then scale range of reasonable values Flip if necessary (distance to water = good or bad?)
Dealing with absolutes Usually can use reclassify operation
Example: if being adjacent to airport is a dealbreaker then recode distances to airport within “tooclose” range to “1”
Weighted overlays in ModelBuilder
In raster, could simply add factor maps Example:
“closeness to school” rated 1..9 “closeness to work” rated 1..9 Map Calculator sum
Value 2 = furthest from school and work Value 18 = closest to both school and work In between = equally weighted index
Weighted overlay expresses two additional concepts
Some factors are more important than others Some factors are “dealbreakers”
Weighted Overlay Demo
Imagine a “tourist restaurant” land use Want to be visible to tourists Don’t want to pay more than necessary for
land Best / Worst
Relatively low cost but highly visible location Factor Maps
Factor 1 = Resort & port accessibility Factor 2 = Land Cost
Tourist Restaurant
Travel time versus Traffic Travel time
Can be across existing roads network But since attractiveness models have roads as input,
can also accommodate future road changes
Local & global accessibility measures in “road accessibility lines polyline” shape file
Accessibl = local (c. walking distance of 1.6m) Accessib0 = regional Values can be treated as an approximation of trips
Tourist Restaurant Accessibility
Subfactor 1: Concept: Busy street Metric: Scaled Global accessibility Implementation:
Use the natural log (ln) to massage highly skewed data distribution of global accessibility
Take 9 equal interval slices Higher values = busier = more attractive Busiest sites at intersections, so use focal mean
to summarize busyness in 3x3 cell area
Legend
Raw Values of Local Accessibility
<VALUE>
0 - 8,073,901
8,073,902 - 16,147,803
16,147,804 - 24,221,704
24,221,705 - 32,295,605
32,295,606 - 40,369,507
40,369,508 - 48,443,408
48,443,409 - 56,517,309
56,517,310 - 64,591,211
64,591,212 - 72,665,112
Resorts
Legend
Resorts
Value
1 - Least Accessible
2
3
4
5
6
7
8
9 - Most Accessible
Tourist Restaurant Accessibility
Subfactor 2: Travel time to nearest resort (Not ideal because better might be average
distance to all resorts within a theshold) Implementation
Cost distance From resorts Over transit time surface
Base walking time = 4 miles/hour Walking slope penalty = pcnt_slope^2
Results ‘manually’ reclassified within MB
Legend
Resort Travel Time in Minutes
<VALUE>
0 - 1
1.1 - 5
5.1 - 10
10.1 - 15
15.1 - 20
20.1 - 25
25.1 - 30
30.1 - 45
45.1 - 1,838,485.3
Resorts
Legend
Resorts
restattract
Value
1
1.000000001 - 2
2.000000001 - 3
3.000000001 - 4
4.000000001 - 5
5.000000001 - 6
6.000000001 - 7
7.000000001 - 8
8.000000001 - 9
General Concepts in Urban Simulation
Basic Modeling Options Endogenous
Attempt to simulate & predict market functions Based on “bid-rent” theory and transportation cost
Exogenous Attempt to predict distribution (but not amount) of
given types of development Form-based Models
Gravity Models Diffusion-limited Aggregation