# Special Topics in Advanced GIS slides - Geospatial Innovation

Document Sample

```					Advanced GIS for UCCE -
Analysis
August 22, 2007
We will begin at 1:10 PM.
Department of Environmental Science,
Karin Tuxen-Bettman                      Domain: CAMPUS
GIIF

http://giif.cnr.berkeley.edu
Geospatial Imaging & Informatics Facility
College of Natural Resources @ UC Berkeley
This Afternoon’s Outline
•   Overview of specific GIS analysis
– Spatial statistics
– Landscape ecology
– Hydrologic modeling and watershed delineation
•   Examples of spatial analysis in natural resource science and ecology
•   Overview of land cover datasets
•   Other software for integrated statistical analysis
•   Spatial analysis and statistics tools in ArcGIS 9.2

Computer exercises: Choose from 1 or more applications, including:
• Map & measure polygonal clusters and patterns
• Measure point patterns and distributions
• Hydrologic modeling and watershed delineation using the Model
Builder
• Using Google Earth for 3D visualization
What are Spatial Statistics?
happen to have spatial component…
Spatial statistics take space into account, e.g. distance.

Two types:
Descriptive – characterizes pattern
– How are points distributed?
– What is the pattern?
– Where are the clusters?
Quantitative – quantifies/measures pattern
(e.g. pattern, relationships, trends)
– How clustered/dispersed is the data?
– What are the relationships with other data?
What is Landscape Ecology?
•   Spatial pattern is linked to
ecological process
– i.e. Turner, Forman and Godron, etc.

•   A landscape is made of
– Structure
>   Patch, corridor, mosaic
>   Size, shape, spatial configuration
– Function
>   Population dynamics, nutrient cycling,
competition, succession, physical
processes
– Change
>   Anthropogenic change
>   Natural change
What is Hydrologic Modeling
& Watershed Delineation?

Hydrology concerns the movement of water across a surface,
the flow of water through a drainage system
Methods for performing GIS analysis

3.   Choose a GIS analysis method,
4.   Calculate the statistic(s) and/or metrics,
5.   Interpret the statistics, and
6.   Test significance.
Land Cover Datasets
•   Multi-source Land Cover Dataset (2002, 2006)
– Source: CDF (http://frap.cdf.ca.gov/data/frapgisdata/select.asp)
– Spatial resolution: 100 meter (2002), 30 m (2006)
•   Landfire dataset (2005)
– Source: USGS (http://www.landfire.gov/products_overview.php)
– Spatial resolution: 30 m
•   Coastal-Change Analysis Project (2002) …coastal counties only!
– Source: NOAA (http://csc.noaa.gov/crs/lca/pacificcoast.html)
– Spatial resolution: 30 m
•   National Land Cover Dataset
– Source: USGS (http://edcftp.cr.usgs.gov/pub/data/landcover/states/)
– Spatial resolution: 30 m
•   CalGAP (1986)
– Source: UCSB CalGAP Project     (http://www.biogeog.ucsb.edu/projects/gap/gap_data_state.html)
– Spatial resolution: 4 ha MMU
•   CalVeg77 (1977) (http://frap.cdf.ca.gov/data/frapgisdata/select.asp)
•   Wieslander Vegetation Type Mapping Project (1920s) (http://vtm.berkeley.edu)
Measuring Geographic Distributions
(e.g. How are the points distributed?)

•   Mean
•   Median
•   Central feature
Spatial Statistics
Spatial Pattern Analysis
•   Pattern of point distribution
>   Nearest neighbor index
>   Ripley’s K
>   Theissen polygons, or Voronoi diagrams
>   Semi-variogram

•   Pattern of point and polygon values
– Continuous data: gradients and localized variability
>   Moran’s I
>   Getis-Ord General G
>   Kriging
– Discrete/categorical data
>   Landscape pattern metrics
>   Join count
PATTERN OF POINT DISTRIBUTION:
Neighborhood Operations
•   What is close to me?
•   Methods
– Straight-line distance
(Euclidean distance)
>   Spider diagram
–   Distance of cost over network
–   Cost over a surface
–   Buffers
–   Variable distance buffers
–   Filters
–   Local, Focal and Zonal functions
–   Distance to/from features
–   Theissen polygons, or Voronoi
diagrams
PATTERN OF POINT DISTRIBUTION:
Nearest Neighbor Index
•   Calculates the average distance between points
•   Significance is tested with Z-score
•   Types
– Inter-centroid distance
– Boundary-boundary distance
PATTERN OF POINT DISTRIBUTION:
Ripley’s K Function
•   Counts the # of features within defined distances
•   Measures spatial arrangement                                       h
h
(clustered, uniform, random)
•   Uses multiple simulations to create a
random distribution envelope
•   Detect the scale of those patterns,
e.g. what is the cluster size?                                              Clustering
•   Assumes:                                                                    Random

– Stationary: No trends in the data                                        Upper bound

– Isotropy: No directional detection                                       Lower bound

Lhat(h)-h
(although it is possible to modify the
K function to detect anisotropy.
– Regular study area (rarely encountered)

Distance (m)
PATTERN OF POINT DISTRIBUTION:
Ripley’s K function
Spatial Autocorrelation
•   Spatial autocorrelation measures the level of interdependence
between the variables, the nature and strength of the
interdependence

•   Can be either positive or negative
– Positive spatial autocorrelation has all similar values appearing
together, while negative spatial autocorrelation has dissimilar values
appearing in close association (less common)

•   Measured by:
– Semivariograms
– Moran’s I
– Geary’s C
PATTERN OF POINT DISTRIBUTION:
Semivariograms
semivariance

sill

nugget

h
range

•   Range: the average distance within which the variable remains spatial
autocorrelated  the extent of spatial trends, distance beyond which
sampling is random
•   Sill: the maximum variance of the sample data
•   Nugget: measurement errors or smaller variations within the minimum
sampling distance  the noise in the data
PATTERN OF POINT DISTRIBUTION:
Semivariograms

Random distribution                                                           Trend data                                                         Regular distribution

Pure Nugget Effect                                                             Unbound                                                   Pure nugget effect
(no spatial structure)                                                         semivariograms                                            (has spatial structure)
0.20
0.10                                                                0.50

0.08                                                                                                                                   0.15
0.45

semivariance
Semivariance
Semivariance

0.06
0.10
0.40

0.04

0.05
0.35
0.02

0.00
0.00                                                                0.30
0     10   20        30        40   50   60
0   10   20         30       40   50   60                           0    10   20        30        40   50   60

Distance (m)                                                        Distance (m)
Distance (m)
PATTERN OF POINT DISTRIBUTION:
Semivariograms

100000
80000
60000
gamma
40000
20000
0

objective = 111391211

0   50        100      150
distance
PATTERN OF POINT & POLYGON VALUES:
Moran’s I
•   Shows similarity of neighboring features
•   Provides a single statistics summarizing pattern
•   For continuous data
•   Spatial covariation/total variation
– Ranges from –1 to 1
>   Positive = positive spatial autocorrelation, negative
represents negative autocorrelation. 0 = no spatial
autocorrelation (random).
PATTERN OF POINT & POLYGON VALUES:
Getis-Ord Gi and General G
•   Hot-spot analysis, showing concentration of
high or low values
•   Indicates whether high or low values are
clustered
•   Uses a neighborhood based on a distance you
specify
•   Applies a weight to those within the distance
that have similar values
Other Software for Statistical Analysis
•   Fragstats
– http://www.umass.edu/landeco/research/fragstats/fragstats.html
•   ArcGIS Geostatistical Analyst
– http://www.esri.com/geostatisticalanalyst/
•   GEODA
– Great for categorical (and other!) pattern analysis
– FREE: https://www.geoda.uiuc.edu/
•   VARIOWIN
– Great for semi-variograms
– FREE: http://www-sst.unil.ch/research/variowin/
•   R
– FREE: http://www.r-project.org/
•   S+ spatial statistics module
– NOT FREE: http://www.insightful.com/products/spatial/
•   SAS
– NOT FREE: http://www.sas.com/technologies/analytics/statistics/
PATTERN OF POINT & POLYGON VALUES:
Landscape Pattern Metrics
Landscape Ecology uses “pattern metrics” to quantify structure

•   Size
– Patch size                   

•   Shape
– Elongated, circular, amount of edge                   

•   Spatial configuration
– Measuring patterns in the mosaic (patch metrics)
>   Clustered, dispersed                             
>   Fragmentation, isolation, connectivity
ArcGrid enabled Fragstats
Landscape Metrics:
ONE metric per site (“landscape”)

Whole
landscape
Class Metrics:
ONE metric per class in the map

Each color
represents separate
class
Patch Metrics:
ONE metric per patch (“landscape”)

Each patch metric
calculated for each patch
Problems with Pattern Metrics
•   There has been much scrutiny of these
techniques, and criticism, including…
– Metrics are highly redundant
– Metrics are very sensitive to inputs and to scale
– Conceptual flaws in landscape pattern analysis
>   Unwarranted relationships between pattern and
process
>   Quantifying pattern without considering process
>   Ecological irrelevance of landscape indices

•   Two recent papers discuss these issues and more:
– Wu, J. 2004. Effects of changing scale on
landscape pattern analysis: scaling relations.
Landscape Ecology 19: 125-138.
– Li, H., and J. Wu. 2004. Use and misuse of
landscape metrics. Landscape Ecology 19: 389-
399.
Definitions
•   Drainage system:
– Area upon which water falls, and the network
through which it travels to an outlet
•   Drainage basin:
– Area that drains water to a common outlet
– This area is normally defined as the total area
flowing to a given outlet, or pour point.
– Other common terms for a drainage basin are
watershed, basin, catchment, or contributing
area.
•   Outlet, or pour point:
– Point at which water flows out of an area
– Usually the lowest point along the boundary of
the drainage basin
•   Drainage divide or watershed boundary:
– The boundary between two basins
Definitions
•   Network
•   Outlet
•   Stream channels
•   Junction, or node:
– Intersection of two stream
channels
– Sections of a stream channel
connecting two successive
junctions, or a junction
– Outermost branches of the tree,
(i.e., they have no tributaries).
Hydrologic Analysis
Flow Direction
•   The output of this request is an integer Grid
whose values range from 1 to 255. The
values for each direction from the center are:

•   For example, if the direction of steepest drop
was to the left of the current processing cell,
its flow direction would be coded as 16.
Flow Accumulation
Flow Accumulation creates a grid of accumulated flow to each cell, by
accumulating the weight for all cells that flow into each downslope cell.

Hydrography is usually created with a threshold of accumulated cell values.
Hydrology Tools in ArcToolbox
•   Watersheds & basins
•   Snap Pour Point
•   Stream to Feature:
simplify vs. non-simplify
•   Stream Order
Data for Hydrological GIS
•   Elevation:
– SF Bay Area Regional Database (BARD) 30m and some 10m
DEMs: http://bard.usgs.gov
– SF Bay NGA 2m DEM: see GIIF
– California: 90m DEM: see GIIF
– National Elevation Dataset (NED) 30m DEM: http://ned.usgs.gov
– North America: 1,000m DEM (ESRI): see GIIF
– Global: 1km GTOPO30 (USGS):
http://edcdaac.usgs.gov/gtopo30/gtopo30.html
•   Stream gage data (daily and real-time):
– USGS National Water Information Systems (NWIS)
•   Watersheds, water districts, rivers:
– Calif. Spatial Information Library (CaSIL): http://gis.ca.gov
– U.S. National Hydrography Dataset (NHD): http://nhd.usgs.gov/
Elevation Data

1,000m DEM

SF Bay NGA 2m DEM

2m DSM & DTM

1m LiDAR of Napa County                  90m DEM

```
DOCUMENT INFO
Shared By:
Categories:
Tags:
Stats:
 views: 6 posted: 8/14/2011 language: English pages: 36