- Raster Analysis
- Review of Structure
- grid of values
- spreadsheet with "spatial sense"
- terms
- pixel, cell
- extents
- resolution
- Boolean/ logic layer
- House Keeping
- copy data
- create geodatabase (.gdb) - in ArcCatalog
- start ArcGIS
- create OR open existing map document (.mxd)
- create/set default geodatabase & relative paths
- Spatial Analyst
- activate the extension
- add the toolbar
- analytical mask
- defines area of analysis
- raster mask
- cells with values define area
- cells with NoData are excluded
- use Reclassify to set NoData
- feature data layer (vector map) can be used
- Reclassify
- creates a new layer
- simplifies/ generalizes
- used to
- convert raw data into classes
- young, immature, OG
- create Boolean (logic) layers
- flood, clay
- bring layers to a common scale
- suitability: 0-10
- in vector ...
- classify in Symbology tab - temporary colour display
- dissolve - to create a new map layer
-
- conduct math on maps ... map algebra
- single layer
- a formula is applied to each pixel
- e.g. used to convert elevation to temperature; feet to metres ... ha --> $$
- multi-layer
- we use this alot to combine layers
- math overlay
- e.g. difference (land value, stand vol., ave temperature)
- Boolean (and, or, not)
- e.g. [Flood] & [Clay]
- relational (<, >, = ...)
- e.g. elev < 9m & soils = clay
- Spatial Interpolation
- what is interpolation?
- est. value based on known values nearby
- Surfaces from point values: elevation
- closer values carry more weight
- checker board - hoola hoop simulation
- methods
- IDW - inverse distance weighting
- 1/distance provides a basis for weights
- eg est. elev. based on 2 values
- #1 is 200m and is 2 m away
- # 2 is 300 m and is 4 m away
- should the pixel value be closer to 200 or 300?
- closer (i.e. sample #1) should have more weighting, but ...
- 2m < 4m ...
- ... so we take the inverse ... 1/2 and 1/4 ... now value #1 has a larger weighting
- the math ...
- it is RELATIVE weighting
- 1/2 = 0.5, 1/4 = 0.25 ... the sum is 0.75
- #1 contributed 0.5 to the sum ...
- #1 weighting is 0.5 / 0.75 = ... 0.67
- thus #2 contributed ... 0.33
- pixel value = (0.67 * 200m) + (0.33 * 300m)
- pixel value = 134m + 99m = 233m
- often taken to a power
- e.g. squared (1/d)^2
- this emphasizes local values
- consider
- 2m and 4 m
- simple inverse (1/d)
- 1/2 and 1/4
- 0.5 and 0.25
- weighting is a proportion of 0.75 ...
- 0.67 and 0.33
- if squared ...
- 1/4 and 1/16
- 0.25 and 0.0625
- weighting is a proportion of 0.3125 ...
- 0.80 and 0.20
- this really emphasizes local values
- worked example
- spot ht of 200m is 10m away
- spot ht of 300m is 20 m away
- intuitively X should be closer to 200 than 300
- the math ...
- 1/10 = 0.1 and 1/20 = 0.05
- so the closer value is "worth" 2x's more
- Relatively it is worth 2/3's
- value = (0.67*200m) + (0.33*300m)
- value = (134m) + (99m)
- value = 233m
- issue: max/min values not honoured
- IDW uses averages ...
- no derived value can > spot ht's
-
- can produce max & min that exceed data points
- occurs if 2 slopes are converging ..
- ... and no peak value is provided
- aims for minimum curvature (general trend)
- surface is very smooth (no abrupt ridges/peaks/valleys)
- Kriging
- "preprocessing of data" to look for trends (directional bias recognized)
- provides a statistical estimation - error and certainty values generated
- more complex process ... more difficult to understand
- user needs to provide 'sophisticated inputs'
- other factors to consider
- local or global
- local uses only nearby values
- global uses all values (but weights closer values stronger)
- global provides a smoother surface ... shows "overall trend" minimizes local variation
- question for local ... how many points to use (defined by # closest OR search radius)
- exact or approximate
- exact - the surface passes thru' measured points
- approx. - "smooths out" sample points - shows trends
- IDW, Kriging & spline are exact
- other methods are approx.
- gradual or abrupt
- gradual good for "rolling terrain" (phenomena)
- abrupt recognizes "boudaries/ barriers"
- ridges/cliffs & river bottoms
- fault lines
- cultural boundaries
- doesn't use values from "the other side"
- Surface Analysis
- create contours
- slope
- rate of change - how fast "elevation" changes
- can be other phenomena (temp., pollution, salinity)
- aspect
- "downhill direction" that a slope faces
- direction of change
- hillshade
- aethetics
- provides a realistic look ... light/ shadow texture
- analysis
- assess level of illumination (heat) on surface
- alter direction for time of day
- alter angle ("height" of sun) for season
- viewshed
- single or multiple viewpoints
- applications
- viewable area - aesthetics
- Proximity
- distance surface
- buffers
- cost distance (below)
- Thiessen polygons
- Density
- also starts with point values ... ends with a surface
- thus creates a surface from discrete data points
- density is value / m2 (ha, km2, etc.)
- compare 3 "point maps"
- spot heights
- point A = 500 and point B = 400
- a location (pixel) located midway is likely 450
- animal sightings (crime locations)
- each point = 1 incident
- closely spaced points indicates "high activity"
- what if I stand at a location that is not a "mapped point"
- what is the likely "activity level"?
- city pop'ns
- city A = 1 million and city B = 2 million
- a location (pixel) located midway = ??
- "simple point" method
- define search radius
- tally point values (= sum) in search area (hoola hoop)
- density = sum/ search area (value / m2)
- convert to /ha or /km2 or whatever
- good for "single incident" points
- kernel method
- the value at each point is "distributed" before the tally is done
- point value (city pop'n) = "tube of sand"
- distributed value = "pile of sand" after tube is removed
- radius defines outer limit of spread
- tension defines "amount of spread"
- value for a pixel = tally "sand contributions" from surrounding points / search area
- diagram for kernel
- can convert to /ha or /km2 or whatever
- good for "artificial pop'n" points
- Cost Surface/ Cheapest Path
- concept
- lazy person walking from A to B
- shorest distance ...
- ... but cost of movement (aka friction)
- big "hill" btwn A and B
- a "ridge" btwn A and B
- a "cliff" btwn A and B
- need a cost surface that "blends" distance with cost of movement
- examples to consider
- pebble in pond
- generates circles of distance
- oil slick would slow ripples (= friction or cost)
- break water would be an absolute barrier - ripples would have to 'bend' to get around
- hiker with mountain/cliff btwn A and B
- road building through populated, hilly terrain
- method
- create a cost/ friction layer
- cost to move thru that pixel
- steepness of hill
- cost of land
- if more than one parameter
- bring to a common scale (i.e. reclass to 0-10)
- consider the relative weighting for each
- create a weighted sum of cost (i.e. raster calc to add the layers)
- create a cost distance layer
- combines concepts of cost (friction) and simple distance
- start at point A ... get distance from pixel back to A
- distance from A * cost (from cost layer) = cost distance
- thus each pixel = cumm. cost from point "A"
- determine path btwn "B" to "A"
