Study Guide 2: Geog 621/721 GIS for Environmental Analysis

Modeling

• Modeling : basic concepts
  • Models used to predict a real-world situation based on the state of a number of variables, with many applications: hydrological, geological, soil erosion, ecological, etc.
  • Types of models
  • Deterministic
  • Stochastic
  • Statistical approaches
  • Heuristic approaches
  • Coding models: Map Algebra, ModelBuilder, Python
• Heuristic models: rule-based, subjective
  • Advantage: Allows for working with categorical data by using weights
  • Simple coincidence modeling: binary, additive (zero best)
  • Weighted modeling: Advantages: multiple data types, even nominal (categorical) & disadvantages: subjective
  • Weighted with factors (e.g. environmental, social, tangible, etc., road construction model example)
• Testing models & Accuracy Assessment:
  • Comparison with known values
  • Tweaking models and sensitivity analysis, to see the effect of variables and their weights. Never run just once.
  • Heuristic models should also be assessed: for each run, compare predictions to actual occurrences. If this is impossible, your results are just a guess. The only thing you can do is sensitivity analysis.
• Examples from exercises:  site suitability, etc.

Surface Display & Analysis

• Visualization of maps in perspective (esp. in ArcScene):  draping maps & images
  • Draping images, features, grids. Base heights derived from _____
  • Extruding using data & constants
  • Vertical exaggeration, and XY vs. Z units
  • Navigation & Fly tools
  • Properties of 3D vector datasets: points, lines, polygons. Points and vertices carry z values.
• Raster Surface Derivatives: slope gradient (% and ° ), aspect, shape (curvature), hillshade (for view & solar radiation), other solar radiation tools.
• Plan curvature:  convex (positive curvature, water-spreading) vs. concave (negative curvature, water-gathering).  Purpose:  detecting landslide areas.
• Profile curvature:  convex (positive curvature) vs. concave (negative curvature).
• Volumes from surfaces: to get thickness (isopach map), subtract lower surface from upper.
• Viewsheds:
  • examples of applications
  • meaning of observer & target
  • variables useful for controlling (observer height, target height, azimuth ranges, vertical angle ranges, distance limits)
• Extracting graphics from surfaces: sections, profiles (e.g. longitudinal profile).
• Hydrologic modeling tools: flow direction, flow accumulation (watershed area in cells at each cell), watersheds, stream delineation from threshold flow accumulation.
• Hypsometric analysis: elevation/area gradient/area

Surface Interpolation from Elevation and Sample Data; Using Spatial Statistics

• Elevation vs "Statistical" Surfaces: different level of detail, ability to observe, thus handled differently.
• Vector (TIN: linear) and Raster Interpolation
• Basic process of creating a Terrain or TIN from point and polyline data
  • Terrain: multiple resolution, TIN created on the fly, edit the source data to change, uses LiDAR
  • Mass points: from points or vertices
  • Breaklines: hard and soft, contours, water edge, ridge lines, etc.
  • What do the triangle corners represent? What is stored for the triangles?
• Pycnophylactic interpolation
• Sample Data: Summary Spatial Statistics
  • Pattern and Event Analysis and Exploratory Spatial Data Analysis (ESDA)
    • Spatial Autocorrelation, Tobler's First Law of Geography
    • Nearest Neighbor (know the difference; clustered, random and dispersed distributions)
    • Moran's I: Spatial Autocorrelation measure (difference from Nearest Neighbor?)
  • Measuring Geographic Distributions: central feature, mean center, standard distance, directional distribution, linear directional mean
• Geostatistical Interpolation
  • ESDA tools in Geostatistical Analyst
    • Semivariogram (for seeing the distance (range) over which spatial autocorrelation has an effect)
    • Normal QQ Plot (for seeing if the data are normally distributed)
    • Trend Analysis Plot (to see directional trends, for setting anisotropy values)
    • Crosscovariance (to see the effect of one variable on another, related to distance)
  • Contrasting properties of methods
    • Deterministic vs stochastic
    • Global vs Local
    • Exact vs Smooth
  • Deterministic Methods (Know their properties and idiosyncracies like extrapolation etc.)
    • IDW (also know the equation)
    • RBF (spline)
    • Trend surface (global polynomial)
    • Local polynomial
  • Stochastic Methods: Kriging
    • deterministic trend + spatially autocorrelated random error
    • ordinary Kriging: only the spatially autocorrelated part, but can be used with detrending
    • universal Kriging
    • Maps from stochastic interpolation:  Predictive, Quantile, Standard Error, Probability
    • Advantage of kriging (analysis of confidence, probability)
  • Assessing methods using Cross Validation, RMSE
  • General comments:
    • Always plot your input data points with your surface!
    • Surfaces shouldn’t reveal any surprises not seen in the point data.
    • Do you actually need to interpolate?
    • There is no one best method.
• Spatial Inferential Statistics
  • Ordinary Least Squares
  • Geographically Weighted Regression, and nonstationarity
  • External models like R -- using text files and spreadsheets for data interchange
  • Species Distribution Models:
    • Going from Geographic space to Environmental space and back
    • Fundamental vs Occupied niche in environmental space.
    • Potential vs Actual Distribution in geographic space.
    • Biotic factors (migration, competition, even human impacts) influence actual distribution