geol 588 gis for geoscientists ii -...

Geol 588 -

GIS for Geoscientists II

Lecture 1, Jan. 14. 08

1

Today

• (Day/Time for lecture / lab)

• Syllabus

• Introductions

• Using the GIS lab’s Delphi server

• WebCT

• Making screenshots

2

Syllabus

• Full syllabus on WebCT

• Format: Lecture + Lab (exercises)

• Lecture may include “follow along” activities

• Thu.: 1 hr lecture, pause,1 hr lab

• Tue. 2 hr lab (with pause)

• No text book (Arc pdf manuals are online)

• Start with chapter 15 of Mastering ArcGIS *

3

• Lab: class exercises (together)

• No regular home work assignments

• Graded exercises: during lab (I’ll be present to help) + on your own time (if needed)

• Documentation-only exercise (~ 10 pts)

• describe GIS methods and results of a class exercise (text + screenshots)

• Grade: effectiveness of your description

• good explanation > good understanding

• Time: 1 week

4

• Variation exercise (20 - 30 pts)

• work on a variation of a class exercise

• same data, use more/different methods

• document the results (screenshots)

• differences? explain/speculate

• Grade: results + clarity of documentation

• Time: 1 week

5

• Mini-project (30 - 50 pts)

• new data + series of tasks

• use many different methods as you see fit

• document methods used + results (text + screenshots)

• Time: 1-2 weeks

• Your advantage to keep good exercise notes!

•All graded exercise: may use Word or Powerpoint

•Additional points for intelligent thoughts about methods and results: good/bad aspects, caveats, alternatives, real-world connections,

6

• Quiz:

• 2 quizzes: open book (notes), test theory

• 20 min to answer 3 of 4 questions (write text)

• Midterm, Final exam (open book)

• Multiple Choice part (15 min.)

• Practical part (90 min. - should only need 60 min.)

• Class (final) project (last 3 weeks)

• poster (only electronic form needed!)

• 10 min. oral presentation (10 powerpoint slides)

• Material up to you (your research?), start early!

7

• Lecture feedback (WebCT) (class participation)

• Best/Worst of last lecture

• Clearest/Muddiest point of last lecture

• “What I learned today”

• Intelligent* question about material

• Total Grade:

• quizzes: 10%

• midterm: 20%, Final Exam: 20%

• graded exercises 20%

• class project (includes oral presentation): 20%

• class participation 10%

8

Topics

• raster data concepts (general)

• types of rasters: ERSI Info grid, images, DEMs, satellite img., int vs. float vs. categories

• images: color-indexed, RGB vs greyscale(luminance), statistics, histograms,

• raster analysis setup (temp/perm rasters, mask, env., raster information)

• using the ModelBuilder interface (ArcToolbox)

• raster symbolization (color-ramp, stretching, classified, histogram)

• raster projection issues

9

• raster import/export/conversion: vector data, TIN, raster file types (geotiff, grid, ASCII)

• map algebra/Raster Calculator (boolean overlay, mask, NoData)

• cell statistics (map stack math)

• neighborhood stats (moving kernels)

• zonal (summary) statistics

• distance analysis (straight line, least-cost-path)Interpolation (IDW, natural neighbors, Spline, (((kriging)))

10

• Terrain (surface) analysis:

• DEM, TINs (3D analyst), slope, azimuth, contour, cut-fill, viewshed, hillshade

• (hydrological analysis: flow direction, accumulation, watershed delineation)

• Suitability analysis

• Volume/thickness analysis

• Data sources for rasters (Iowa DNR, USGS seamless)

• Geo referenceing

11

Instructor – Chris Harding

•Masters Degree in Geology (Geoinformatics)

•Worked for Lynx Geosystems (Vancouver, Canada) writing GIS software

• Ph.D. in Geology from University of Houston

•Assistant Professor Geoscience Department.

•Contact info: [email protected], tel. 4-4868,

Office: 2635 Howe Hall, Rm. 2635

Introductions

12

Your turn, tell us:

• Name ?

• Major ?

• Advisor ? Thesis project ?

• Prior GIS involvement ?

• Why are you taking this course ?

13

Wrap-up

• Week of March 17: Midterm

• Week of May 5: Final Exam

• Until next session: read p. 523 - 532 in Chapter 15, Mastering ArcGIS

• WebCT: GIS materials folder

• Delphi: geol588\data\MGIS_chpt_15.pdf

• Try WebCT, delphi

14

Geol 588 -



1

Today

• Assignment: making screenshots (clear?)

• Examples of using raster (Spatial Analyst)

• Raster concepts (intro)

• Pause

• go through first part of MGIS Ch. 15 tutorial

2

• Derive new information

• Visualize/analyze properties

• Find area most suitable for an objective

• Suitability analysis

• Identify the best path between locations

Examples of Raster GIS tasks

3

• Perform distance and cost-of-travel analyses

• Perform statistical analysis for predetermined zones

• Interpolate data values for a study area based on samples

4

• Clean up a variety of data for further analysis

• Generalization (“image processing”)

• Has anybody used any of these fucntions?

5

• cells (pixels) in rows (vertical) and columns (horizontal)

• 2D matrix, 2D array

• cell: width hight (same?)

• resolution

• Cartesian (index)

• georeferenced (UTM)

• Extend

Inside a raster

6

• Surface (elevation)

• Continuous Data

• no clear boundary

• each cell:

• real lnumber, floating point value

• at center of cell

7

• thematic data

• nominal (categorical)

• class (name, “concept”)

• cell filled with whole number

• ordinal (ranked) data

• integer data (value) as index

Value Count Landcover

1 6 Forest

2 8 Lake

3 10 Urban

8

• image = raster (*)

• satellite image

• four ways of displaying raster/image

• colormap = 255 color ramp

• multiband: overlap 3 images

9

• NoData “values” (raster)

• “transparent” (image)

• no data (elevation) measured

• ArcGIS: word (NoData)

• internally” “magic” number

• any operation with NoData yields NoData

• set Cells to NoData

10

• Math with binary rasters

• presence = 1 (True, Yes)

• absence = 0 (False, No)

• Boolean function: AND, OR ... ; Math: +, -, ...

A AND B

A XOR B A NOT B

A OR B

A B

0

11

0

110

11

0

11

0

11

0

1

2

11

• Suitability Analysis

• Best site for new School

• define desired rules

• slope < 10 perc.

• close to recreation

• not close to other schoole

• Accept simplification

• closeness = straight line distance

• reclass inputs to 1 - 10

• weighted average of all inputs

• 10 = “perfect spot”

12

• Getting help:

• Public (P:) drive - ESRI folder

• Build-in help: F1, ArcToolbox Tools: right side panel

• Webhelp for ArcGIS 9.2: http://webhelp.esri.com/arcgisdesktop/9.2/

• Tutorials (flash video), including ArcGIS extensions:http://webhelp.esri.com/arcgisdesktop/9.2/index.cfm?TopicName=Tutorials

13

Class Exercise 1• model (predict) snail habitat via “rules”

• snails like (only):

• elevation 1200 m - 1600 m

• limestone

• coniferous trees

• snails hate:

• too-close-to-roads (squish!)

• Task: use boolean maps (like/hate)

• copy \\delphi\Geol588\data\MGIS_ch15 folder to your student folder (pdf with instructions)

• load MapDocuments/ex_15a.mxd

• follow along - make notes (graded exercise later)

15

Geol 588 -



1

Today

• Topographic functions (surface analysis)

• Slope, Aspect, Hillshade, Viewshed

• (connect to delphi, start ArcMap)

• Pause

• Image/Raster manipulation via Python

2

• multiband: overlap 3 images

• Red, Green, Blue

• 3 intensities:

• 0 (black) - 255 (Red, Green, Blue)

3

Digital Elevation Model (DEM)

• Digital Terrain Model (DTM)

• topography: Elevation (ft. or meter)

• US Geological Survey:

• 7.5’ maps: 10 m, 30 m; 1degr.: 90 m

• National Elevation Dataset (NED): 10 m (?)

• Shuttle Radar Topography Mission (SRTM): 90 m

• raster values: large integers (1243 ft) or float (543.2 m)

4

Slope & Azimuth

5

Slope

• Based on elevation difference of cell to elevation of surrounding cells

• Gradient at cell (center)

• Gotcha: X/Y are in meters, elevation (Z) is in feet

• Solution: multiply elevation by 3.280 (Z-factor)

• Spatial Analyst -> Surface Analysis or ArcToolbox - Spatial Analyst tools (NOT TIN slope!)

• geol588\data\mgisdata\BlackHills\rasters\dem2

6

Slope

• How to symbolize?

• Properties - Symbology: Classified vs Stretch

• Apply a color ramp from elevation A (1230 ft) to elevation B (1423 ft)

• Gotcha: A and B are not always min and max !

• Types of Stretch: Standard deviations (2 sigma default), user sets min & max, none (data min & max)

• Select color ramp by name (right-click - Graphic View)

7

Aspect

• direction of cell center, when looking down towards max. (!) slope

• in degrees (0 - 360) from North, circular data type

• Problem: How to effectively convey direction?

• Looks like “3D” slopes (similar to hill shading)

• Spatial Analyst - Surface Analysis, ArctoolBox - Spatial Analysis Tools(!)- Aspect

8

Hillshading• Trick to fool our eyes to see 3D relief

• Simulate Sun from direction (0-360) and angle (0-90)

• Gotcha: Sun must be in the North (270 - 90), otherwise relief looks inverted (human evolution: sun from above)

• 10-50% transparent, drape over DEM

• graphical only(Symbolization) - less options

• Spatial Analyst - Surface Analysis, ArctoolBox - Spatial Analysis Tools(!)

9

Viewshed

• needs a shapefile with points (Sturgis -summits)

• Which cells can be seen from these points?

• Shoots rays from point to each cell, check for intersect with terrain

• Can be more complex - add more data to points: height, view angle, distance (more next lab)

• Spatial Analyst - Surface Analysis, ArctoolBox - Spatial Analysis Tools(!)

10

5 min pause

11

Python Exercise

• copy geol588/data/588_python_img_exercises to your folder (E:)

• on python code (.py), snake-icon: right-click - Edit with Idle)

• Idle: editor window (write code), shell window (output)

• Run code: click on editor, Press F5 (yes to save)

• will print output into shell window

• Abort (“hanging”): Press Control-F6

• Let’s look at 1_show_raster_true_elev_in_ft.py

http://www.monty-python.com/mp/cover.gif

12

Geol 588 -



1

Today

• Ex. 2 solution

• Neighborhood & zonal functions

• (ch. 15 tut. pp 547 - 551)

• also: Regions vs. Zones

• Pause

• More about the raster calculator

2

• 3 binary rasters (Red, Green, Blue):

• Elev: (raster, int or float), Elevation > 1400

• Veg: (vector, cat.):TREE_SZ96 = ‘O’ OR TREE_SZ96 = ‘L’

• Soil: (vector, cat.) NAME = "Madison Limestone"

• SnailHab_1 = Elev * Veg * Soil or: Elev AND Veg AND Soil

• Min: 0 * 0 * 0 = 0, Max: 1 * 1 *1 = 1

• Area: 16886 cells, each cell is 50 m x 50 m, 16886 * 2500 = 42.2 Mio m2

3

• more levels of “happiness”

• SnailHab_1 = Elev + Veg + Soil

• Min: 0, Max: 3, Range: 4

• Caveat?

• Good colors?

• % for each level?

• 100 % is raster width x raster height

• 410 x 540 cells or ~540 Mio. m2

• 0: 40%, 1: 30%, 2: 26%, 3: 5%

4

Neighborhood functions

• each new cell: result of “using” cells “around” it

• typical: 3 x 3 square

• other shapes

• ArgMap menu, Toolbox and raster calculator

5

• Ex: 5 x 5 Majority ofSlope > 10 deg

• Effect of 2. pass?

• Majority: needs int or categorical raster!

• Gotcha for Majority for 3 or more possible values: tie results in NoData!

•

6

Block Functions:

• Similar to focal but fill area with same value

8

Zonal functions

• Needs 2 inputs:

• a zone layer:

non-float raster

or polygon

• a value raster

• Zones don’t need

to be continuous!

(see: regions)

9

Zonal functions

• for each zone do:

• get stat for value raster cells that are inside the zone

• Creates many types of stats per zone “id”

• creates a standalone table and a chart

• Table can be joined to zone layer

• example: get

• mean (min, max, etc.)

• elevation

• for each watershed

11

• more zone related functions in Toolbox and Raster Calculator

• for later: make regions (REGIONGROUP function)

• ex_15b.mxd:

• zonal layer - watershed

• (convert to raster)

• value raster: slope

• join result to watershed polygons

• (can also use points, lines as zones!)

12

More about the raster calculator

• About Building Expressions button

• so far we used mainly binary tests: [raster] <logical expression> ; result: 1 or 0, examples: [slope] > 0.2 , [landuse] == 3

• [Mygrid] = [input-raster] * 2 ( not == !)

• [Mygrid] will be in temp folder!!

• functions: ex: Int( [r] ) - set values of [r] to integers

• Usage: right-click - Usage (better: ArcGIS help)

1

• if-else function: (3 parts)IF <logic. expr.> THEN <yes-case> ELSE <no-case>

• CON( <logic. expr>, <Yes>, <No>)

• CON( [slope] > 10, 10, [slope] )

• test for NoData: IsNULL([raster]), returns 1 if NoData, else 0

• “convert” NoData to 1000: CON( isNULL( [r] ), 1000, [r] )

• binary NoData raster: CON( isNull([raster]), 0, 1)

• setting NoData: setNull( <logic. expr>, <no-case> )setNull( [r] < 1000, [r] ) setNull ( [r] = -9999, [r]( there is no explicit yes case, yes means: set to NoData )

2

• Masking: ignoring areas you don’t want to deal with

• Mask raster: Nodata for “outside”, any valid value (1, 0.32, -45, ..., even 0!) for “inside”

• Spatial Analyst - Options - General - Analysis mask

• sets result of any raster operations (raster calc., slope, ....) to NoData for the masks “outside” cells

• Visual-only mask: 2 x click on Data Frame (“Layers”)

• Data Frame tab - Clip to shape (enable)

• give shapefile (polygons) or extend

Masking

3

Geol 588 -


Lecture 5, Feb. 7. 08

1

Today

• Ex. 3 solution

• 2D spatial interpolation (theory)

• Pause

• Effects toolbar

• Interpolation in ArcGIS (examples)

• No lab Feb 26 (and Mar 11)

2

Ex. 3 - Many Peaks

• Where to build a lodge?

• flat ground (< 20 deg. slope)

• has a view of as many peaks as possible (13 of 18)

• viewshed analysis uses canopy (elevation + trees)

• possible cell values in initial viewshed: 0 to 13 (# of peaks)

• convert slope < 20 to 0,1 (binary), thresholding operation), good_slope

3

• Trick: filter viewshed with good_slope

• any 0 value cell in good_slope“flags” the same cell in viewshed as bad for lodge location

• (even if cell has lots of peaks!)

• viewshed (0 - 13) * good_slope (0,1) = lodgesite (0 -13) (0 has double meaning)

• viewshed: integer raster, attribute table, how many cells for each view?

• 5 or more views: add COUNT value for 5 to 13: 4233 cells

4

• Final result: one cell with 13 peaks and slope < 20

• hunt for cell: turn 13 red, rest ( < 13) black

• More data about cell (information tool, select correct layer!)

• Elevation:1728 feet? 1716 feet! (not canopy, not meters)

• Slope: 1.58 degrees (very flat)

• Aspect: 174.8 degrees (approximately due south)

• Viz tips: smooth rasters, good contrast for points

5

Spatial Interpolation

• point samples (x,y, “value”)

• fill each cell (center) in raster with an “appropriate” value

• Principle: the closer together points are, the more similar their value (should be)

• (depends on many factors: type of underlying phenomenon, etc.)

• Examples?

• general: distance and sample value matter

6

Thiessen (Voronoi) polygons

• Interpolation? Space division scheme?

• Assign each sample “its fair share” of space around it

• Raster: fill this space (polygon) with same value

• Problem?

7

(Thiessen polygons: not in spatial analyst but keep in mind for later)

8

Moving “Averaging”• Simple form: grab all

the point samples within the sample radius

• Cell value is Sum of these points’ values divided by their number (mean)

• problem?

• (interpolated value at sample location?)

9

Related topic: point density

• NOT a value interpolation scheme!

• How dense are the samples together?

• count all the point samples within the “circle”, divide by circle’s area

• “kernel” method: different math, smoother

10

Inverse Distance Weighted (IDW) interpolation

• grab all the point samples inside a radius

• When calculation the cell’s value, take the sample distances into account (“weight”)

• Weight of each sample point is an inverse proportion to its distance to the cell

• The further away the point, the less the weight it gets (contributes less)

• effect of exponent (1,2,3, ..)?

Zi is value of known pointDij is distance to known pointZj is the cell value (unknown)

n is a user selected exponent or power (often 1,2 or 3)

exponent n is 1 here !

11

“linear”: exponent = 1“squared”: exponent = 2

• Higher exponents: less weight to distant points (point that are farther away)

• (closer point are more important)

• Use more samples: “smoother” distribution

• interpolated value at sample location?

Simple Averaging

12

IDW: search radius

• variable search radius: grab n (12) closest points, up to a distance of d (default 0, no limit)

• fixed search radius:grab ALL points within d (250) units distanceuse at least n points (default 0, use all)

• isotropic search (more options via geostatistical analyst)

13

IDW: using break lines

• line features to limit (block) point “grabbing”

• samples “from the other side” of the break line will NOT be used

• use for: ridges, faults, ...

• Caveats: much slower, some NoData values (extrapolation issues?)

14

Splines• smooth surface, non-exact

• can over-shoot / under-shoot

• ArgGIS: 2 types - regularized and tension

• ArcGIS: Weight - smoothness “tweak” factor

• gotcha: both have different meaning of weight

• tension (0.1 - 15): low = smooth, high = coarse)

IDW: squared,

12 nearest pointsSpline

15

5 min Pause

17

Interpolation class exercise

• copy geol588\data\interpolation_excercise folder

• new ArcMap, add elev_pts_samples_100 only!

• IDW fixed: fixed search radius 5000 m

• IDW variable: variable search radius (12 pts)

• Tension spline, weight 0.1

• Tension spline, weight 10

• Color all with Temperature (spectral) color ramp (smooth)

• Activate Effects tool - compare results

• play with 10000 points, compare to “true” data (dem_steep)

18

Geol 588 -



1

Today

• Ex. 4 solution

• Dealing with distance (cost) maps

• Pause

• Practical examples

2

Ex. 4 - Ski Areas• What values do the ski areas (zones) have?

• Look at elevation, slope, distance to roads

• dem_steep is a float raster: properties -source - statistics (last section)

• elevation:1301.556 m - 2277.573 m (avg.:1670.188 m)

• focal (neighborhood) statistics: 3x3 mean vs. 3x3 median: small differences only:

• mean: 1302.74 m to 2263.37 m, median: 1302.42 to 2264. 73 m

• Both: move extreme (min/max) towards “middle” (smoothing)

• Median: sort high-to-low, pick “middle” value

• Median: more robust to outliers

• Honest color comparison: same min/max

3

• Zonal statistics for

• Mean Slope: 5.5 deg. to 25.10 deg.

• Area: 265,000 sq. m. to 32,675,000 sq. m. (not pixels!)

• Absolute Min/Max Elevation: 1393.71 m - 2144.53 m

• Assign a color to each polygon according to a field (e.g., ski_area.RANGE)

5

• Coloring: 5 colors - 5 suitability ratings (rating 1 - rating 5)

• Can we compare ratings across maps?

• Simple approach: Average of the three ratings

• What is a “good” area?

6

• Lines as zones for zonal statistics:

• COUNT? AREA?

7

Dealing with distances

• Two types: Straight line (euclidean) distance + cost distance

• straight line distance needs: points or lines (many points)

• each cell knows: “what is my distance to the closest point?”

• NOT know: which (ID) or direction is that closest point?

• Allocation raster: Which point is the closest?

• Similar to ? (last lecture)

8

• Direction grid

• Each cell encodes: “which direction do I need to go to get to the closest point”

• Same scheme as aspect map (azimuth)

9

Real cost to traverse a cell

• Need to define “to traverse this cell, one needs to pay X”

• X on a scale of A to B

• Examples:

• slope: 0 (%) to 50 (%)

• land use (1 - 10):

• roads (city) cost 1

• fields cost 6

• water cost 10

10

Cost weighted distance with roads as sources

11

more outputs of cost weighted distance

operation

• Cost direction raster

• direction optional - but needed later for shortest path operation

• 8 possible choices (directions)

• “On this cell, which direction do I need to go to the next cell”

• next cell: closer to the source (feature), here grey pixel

• Repeat until hit the source

• (roads are not directly involved here)

• Also: Cost allocation raster (not needed for shortest path)

12

• Input 1: path to, “end point(s)”

• Gotcha : you may need to covert “end point” to a raster (?)

• Inputs 2 & 3: Cost distance raster, my “source point(s)”

• result: shortest (cost) line to go from any source point(s) to end point

Shortest path operation - School

Start point(s)

End point(s)

13

Start point (path to)

Source points (lines)

Shortest path operation - point to nearest road

14

Distance class exercise• copy geol588\data\distance_exercise folder

• load dist_ex.mxd, set extend to landuse cost raster!

• Spatial Analyst - Distance - Cost Weighted ...

• Weighted Distance to: schools (or schools_grid), Cost Raster: landuse cost, create direction:YES, create allocation: YES

• Distance - Shortest Path:

• Path to: (think: “Where do want to go?”): destination point (star)

• Cost distance and direction of source: .... to school (created in last step)

• (More? Make cost distance raster for roads (end), shortest path - use schools as start)

15

Geol 588 -



1

Today• Ex. 5 solution (interpolation)

• How to use the Model builder

• Notes:

• No lab Tues. next week - but lecture

• Midterm March 13 (no lab Mar. 11)

• Final: Fri., May 9 !9:45-11:45 ?

2

Ex. 5 - Interpolation from points

• Inital data: bedrock elevation points

• visualize elevation (same color scheme as interpolation for comparison)

• shorten labels: use int( [elev]) in label expression (not in attribute table!)

• Where are samples “bunched up”?

• kernel density - histogram equalization for smooth color: bring out extremes

• no actual point at max. density!

• What’s the average distance between samples? ~ 1050 m (max: 4500!)

• Good interpolation up to 1050 m only?

3

IDW with fixed radius (~2100 m)

NoData assigned for cells beyond

2100 m away from any sample!

(“our comfort zone”)

4

IDW (12 closest points) power 2 vs power 3

5

IDW with break line vs.Tension spline

Close (duplicate?) points with different elevation!

IDW just “averages”, spline tries to “optimize” - and “freaks out”

6

TSpline - IDW (simple subtraction)

Better: absolute difference abs( Tspline - IDW) or Root-Mean-Squared (RMS)

How to get total difference (sum of all cell values)?

Shows smooth (spline) vs IDWs “linear patters” (why?)

Real measure of interpolation quality: subtract from true elevation (dem_steep)

7

5 min pause

8

Model Builder• Sequence of geoprocessing tools

• Flowchart story of geoprocessing operations

• Allows parameters for flexible models

• Allows to to show a process (on a poster)

9

Creating a new toolbox and

Model

drag

tool

Add data

drag layer from TOC

11

Validate & Run

13

Add to display and TOC

Text

Connect graphically

14

Model Builder: practical case• copy data/modelbuilder_ex

• change location for storing your tools in your folder (Tools > Options > Geoprocessing > My Toolboxes)!

• check: “overwrite output of geoprocessing operations”

• new ArcToolbox, (MyTools)

• new Model (Name: Interp_tool, Label: My Interpolation Tool, Store Relative path names!)

• Interpolate from bedrock_elevation points (different schemes and parameters) (call: samples

• Spatial Analysis Tools > Interpolation > IDW

• Connect Samples to IDW

• configure IDW tool, make sure output is saved locally (call IDW_result)

15

• output > Add to display (show changes). Save Tool and run model

• play with different IDW parameters(will overwrite!)

• Sp. A. Tools > Math > Minus Tool

• connect to IDW result

• add dem_steep, subtract IDW result from it

• check order: 2. input - 1. input

• Add result of subtraction to display

• Next: zonal statistics: sum of differences ( = thickness)

16

Geol 588 -


Lecture 8, Feb. 28, 2008

1

Today

• Suitability modeling in GIS

• Reclassification, Weighted Overlay tool

• Spatial Analyst Tutorial Excercise 2

• Notes:

• No lab Mar. 11 (in two weeks)

• Midterm March 27 (prep March 25)

• Final: Fri., May 9 !9:45-11:45 ?

2

Data Types

• continuos (floating point, floats)

• discrete

• whole numbers (integers, ints)

• names/categories (IDs, ZIP), encoded as integers with a lookup table

• What type is suitability?

3

• “Raw” Data sets

• Derived data sets

• Deriving new data is more than visualizing data in different ways!

• Ask: has the data base changed?(attributes, cells)

4

• Goal: combine derived and raw data

• Need: common measurement scale (1-10)

• Reclass(ify) into integers (!)

• (Tip: use the same color scheme, to compare unlike here!)

5

Reclassify• Numerical Input

• Classification Method (Equal interval, Defined interval, Std. deviations, etc.)

• “Direction”: high(low) input value to high or low reclass (scale) value?

• Slope (0 - 45.5) for engineering ( 1 - 10) ?

• categorical input (unique):

• assign a reclass value for each (or use ranges)

• reclass LU 1-3 to 1, LU 4-5 to 2, LU 7 to 3

6

The “restricted” value

• Used only in Weighted Overlay tool

• Means: cell not allowed (even if some derived values are very good!)

• Smallest suitability value - 1, here: suitability 1 - 9, therefore restricted = 1 - 1 = 0

• effect: result (overall suitability) is 0 (“restricted”) whenever it is involved

• Why use restricted value? Allows NoData for other meanings (i.e. true data errors)

7

Combining and weighing data sets

• Need weight factors for each reclassed data set

• A (worth 30%) + B (worth 20%) + ... = 100%

• use weighted overlay tool

• Raster calculator:A * 0.3 + B * 0.2 ...

8

Spatial Analyis Tutorial

• Data and tutorial instructions (pdf) data\spatial (copy spatial)

• New ArcMap, add data (pg. 2), look at data, go to page 14

• Shortcut: create slope (use 0.3048 as Z-factor, degrees), distance to recreation sites (EuD_rsites), dist to schools (EuD-Schools), all need to end up in the Results folder

• Add to model (canvas), toggle on “display got to pg. 24

• reclass slope only, Land use is already in integers (categories) 1 - 7 !

• Use Weighted overlay tool (p. 33) for slope and landuse

• Note the difference between NoData and Restricted! (p. 33), restricted will result in 0 here

9

• make sure to use the Landuse attribute as input field for land use

• Influence: reclassed slope: 30%, reclassed landuse: 70%

• Result: suite_areas (0: restricted, 9: optimal)

• Note: result are integers!

• Raster calculator: 0.3 * recl. slope + 0.7 * recl. landuse

• Call is suite_area2 - What’s its numeric type?

• Check if your new tool is saved in your student folder before you quit

10

Geol 588 -


Lecture 9, Mar. 6, 2008

1

Today• Surface Hydrology analysis (intro)

• solution to ex 6 (anchor hill) on Mar. 13

• Notes:

• No lab Mar. 11 (in two weeks)

• Midterm March 27 15 mult. choice questions, 60 min practical part

• prep session March 13 - need 3 questions from everybody (until Mar. 12)

• Final: Fri., May 9 !9:45-11:45

2

Surface Hydrology

• Disclaimer: I’m not an expert in this

• Basic principle: Water from above (rain), flows downhill

• Deals with sub-areas of terrain (DEM): basin, watershed

• Basics:

• Surface flow direction - how (where) does a drop of water move between cells (depends on what?)

• accumulation: each cell keeps count how many drops move through it

• get the geometry and size of streams

• delineate each stream “area of influx” basin)

3

Calculating flow direction

• Uses elevation - do NOT use a slope raster. Output is NOT an aspect raster

• go through all cells ...

• water will move from each cell “down hill” cell

• compares cell’s elevation to the elevation of the 8 adjacent cells

• result (flow direction) is coded with 1, 1 + 2 = 3, 1 + 128 = 129 (6? 7? 256?)

7

• what if a cell has the same elevation as one or more adjacent cells?

• could flow into 2 (or more) cells!

• expressed as sum of directions: 1 + 2 = 3, 1 + 128 = 129 (6? 7? 256?)

• Gotcha: Sink - all adjacent cells are higher

• Sinks need to be filled!

• also: elevation drop map:

• what’s the vertical difference to the downhill cell?

8

Finding & filling sinks

• local minimum - all adjacent cells are higher

• problem with accumulation:

• for each cell: run a drop of water downhill (using the flow direction)

• But: each cell need to be able to reach the raster’s edge!

• Sinks will simply “suck in” water (i.e. there is no “overflowing”)

• Sink map (Sink tool) Sink ID, rest NoData

• fill sinks (Fill tool) - will change the DEM - need to re-do flow direction

12

Flow accumulation • for each cell:

• run a drop of water downhill (using the flow direction) until the map ends (mini-stream), no losses (evapotranspiration, groundwater)

• each cell the drop touches on its way gets +1 added to its internal count

• effect: if a cell get run over by many mini streams, it will have a high count

• (optional weight raster - “rainfall intensity”)

• use stretched colors with good contrast and 2+ std. dev to bring out the “major streams”

13

Min: 0, Max: 20250

14

Problem with unfilled sinks - stream disappear suddenly

Good stream (reaches end of raster)

Sinks (black)

Same DEM but without filling sinks

15

Stream Channels - stream order

• on accumulation raster: use a threshold (cutoff) to define actual stream channels

• Raster calc: Flowacc > 140

• stream order tool: define magnitude (“size”) based on stream joins not on accumulation

• 2 methods: for defining magnitude:

• Strahler (low numbers), Shreve (large numbers)

16

Basins • ArcGIS: Basins are different from watersheds!

• Basin: completely define by flow direction raster

• Basin tool, creates zones

• Problem here: lots of small basins (COUNT < 65)

17

Geol 588 -


Lecture (Lab) 10b, Mar. 25, 2008

1

Today• solution to ex 6 (Road to anchor hill)

• lab: new exercise (Surface Hydrology)

• some pointers about layout of maps

• go around: class project ideas (group proj.?)

• IowaDNR data on P-drive (IA_State, IA_Regions, Basins, support, etc.)

• Midterm this Thursday

• class project presentations (10 min): Apr. 24

2

Exercise 6 solution

• How (where) should we build a road from the highway to Anchor Hill?

• Note: we don’t care where the new roads starts along the highway

• Straight Line distance: no effects on distance (terrain type, slope, etc.)

• The higher a cell value (yellow - blue) the farther away (yellow: “valley”)

• Straight Line Distance from the highway to AH is 13,211 m, to FM is 14,131 m

• How to get the precise value?

• Could you use the dist. measure tool?

3

• For more complex road, weighted cost distance

• need traversal cost raster

• need slope and forest cover

• need to “invert” rock %

• need rock% everywhere

• Stretched: “Display NoData as” red

4

• Assume that NoData cells mean 100% rock

• How to turn NoData into 100? rock_percent => rock_pct_all

• IF < is this cell’s value NoData? >THEN < make it 100 >ELSE < don’t change the value >

• IF < question >THEN < yes answer to question>ELSE < no answer to question >

• IF A THEN B ELSE C

• in raster calc: CON(A, B, C)

• isNull( [r] ) function - is the value of the current cell in [r] NoData (Null)?

• CON (isNull([rock_percent], 100, [rock_percent])

100% bare rock

0 -100% bare rock

How could we solve the NoData problem around

the forest fringes?

5

• create a traversal cost raster

• cost for traversing each cell (in isolation) depending on slope and density of forest

• Forest density: 100.0 - [rock_pct_all]

• (100.0 - [rock_pct_all]) + slope

• Min: 0.1, Max 211.6

• Calculate cost-weighted distance raster

• Need: traversal cost and point/line feature (“distance to”)

• Gotcha Alert: what is the feature here?

• Also creates a direction raster

• Again: creates a “valley” (Unit?)

0 -100% bare rock

6

• perform a shortest path operation to AH and FM

• we supply the path-to point

• also: uses weighted-cost distance raster and direction raster

• But: how does it know where we want to start?

• weighted-cost distance raster <= “distance-to” feature (badly named)

• Problem: path takes (unfair) advantage of the NoData gap around the forest

7

• Solution: constrain the weighted-cost distance raster and direction raster

• Use Mask raster (NoData vs. None-Nodata)

• How much does each new road cost?

• Use path a zone in zonal analysis (SUM)

8

• Cost for each road?

• Adding up the cost from highway to AH along the path

• But: need to use the cost raster, not the weighted-cost distance raster (my mistake)

• Reason: each cell in a weighted-cost distance raster is already “added-up”

• (same principle as the straight line distance operation “adds-up”)

• Unconstrained path: 25 km, $24 Mio. ($128 Mio. )

• Constrained (mask) path: 19 km, $19 Mio. ( $170 Mio.)

9

Layout mode

• Page setup: deactivate “Use printer settings, chose page size and orientation

• Multiple Data frames? Switch to layout mode, Copy/paste Data frame

• Wait until you have all layers before making multiple data frames (saves layer copying)

• Activating different data frames: Active (TOC), click in layout mode

• Scale: manually set to “nice” numbers

• zoom/pan data “inside” data frame

• Use guides to line up boxes (click inside ruler)

10

Geol 588 -


Lecture 11, April 3, 2008

1

Today• solution to ex 7 (Suitability)

• go around: class project ideas (group proj.?)

• Final Test: Take home (due May 9), related to your class project (so get started soon!)

• Exercise 8 solution next Tuesday (Midterm solutions?)

• Tuesday / Thursday 2 - 4: I’m available to help with project

• Rest of semester: no lecture/lab, work on class project

• I’m away next Thursday (April 10) and May 1

2

Final project:

• Put everything to learned together in a bigger

exercise

• OK to explore some new tools

• Should (mostly) use raster GIS data

• suggestion: suitability analysis (prediction map)

• OK to make up an “interesting” story

• Before you start: tell me a project outline

3

• Don’t get too hung up on finding the “perfect” data

• Look at: P-drive (ftp://pub.gis.iastate.edu/) DNR Iowa data (environmental, people, transport, commercial locations, water, soils, ...)

• Deliverables:

• Poster (36“ x 27” ) as 96 dpi jpg

• Poster’s Layout & ability to “tell the story” count!

• Powerpoint version (no more than 10 slides!)

• Poster due + 10 min presentation: May 9, 9 am

4

Exercise 7 solution

• Perform a suitability analysis

for building a new school at a

certain site with 4 factors:

A. Distance to recreational facility (small is good, 50%)

B. Slope of Terrain (small is good, 13%)

C. Distance to existing school

(small is bad, 25%)D. Type of land use (some good,

some bad, 12%)

• Reclassify each factor to go from 1

(bad) to good (10)

5

• Assign importance (weights) for each factor:

Suitability = A * 0.5 + B * 0.13 + ... + D * 0.12

=> cell value: 0 (bad) to 10 (good)

• Result: Overall suitability (reality: 1-9, no perfect 10)

• But: restrict “good” sites to very high suitability values ( cyan = 9)

• Experience with Model Builder (good/bad) ?

6

• Start: cells where Suitability = 9

• Earlier: removed very small (1 cell) sites (How?)

• But: Site needs to be “islands” (regions ! zones!)

• Use RegionGroup tool

• 4 way connectivity vs. 8 way connectivity (here?)

• Regions > 10 cells needed

• Use Extract by Attribute Tool (uses a SQL query)

• 4 connectivity:56 regions, 13 regions >10 cells

• 8 connectivity:36 regions, 14 regions >10 cells

• (Many different results given)

7

• Wanted: compact (chunky) shape (beside > 10 cells)

• C = 4" * (area) / (perimeter)2

circle: C =1square: C = " / 4infinitely narrow shape: C = 0

• Raster land:compare the area (sum of all cells) with the perimeter (sum of “fringe cells”)

• Cheap trick - just use:Compactness = Area / Perimeter

• Small Compactness => skinny

• Large compactness => chunky

• Does is work?

8

• Make sure the site is close (< 50) to any road

• Better done in polygon (vector) world

• Convert to regions to polygons

• (What happens during the raster-to-polygon conversion?)

• Use Select by Spatial attribute:

• Select polygons where:distance to road < 50 m

• Note: We don’t know where (which) road is the closest for a given polygon!

• (How could it be done with rasters only?)

• Pick your “best” site (add human judgement)

0 -100% bare rock

9

Questions?

10