4.3 Modeling Geographic Phenomena: Discrete Objects vs. Continuous Fields

Modeling Geographic Phenomena: Discrete Objects vs. Continuous Fields

Quick Summary: Before choosing a spatial data structure, GIS professionals must understand the conceptual nature of the geographic phenomenon being modeled. Geographic phenomena generally follow one of two fundamental conceptual paradigms: the Discrete Object Model (space as an empty container populated by distinct, bounded entities) or the Continuous Field Model (space as a continuous surface where every coordinate location possesses a measurable value). Navigating the trade-offs of representing fields in vector (e.g., isolines, TINs) and discrete entities in raster (e.g., categorical grids, mixed pixels) is a core competency tested on the GISP exam.

---

Conceptual Models of Geographic Space

All spatial analysis begins with human conceptualization of geographic reality. How an analyst conceives of a phenomenon dictates whether vector or raster data structures are appropriate.

                      CONCEPTUAL MODELS OF GEOGRAPHIC SPACE
                      
   +---------------------------------------+---------------------------------------+
   |         Discrete Object Model         |        Continuous Field Model         |
   +---------------------------------------+---------------------------------------+
   | * Space is an empty coordinate frame  | * Space is completely filled          |
   | * Populated by distinct, crisp items  | * Every location (X,Y) has a value    |
   | * Clear geometric boundaries (edges)  | * Variables change smoothly across    |
   | * Countable, individual entities      | * Seamless continuum (no empty space) |
   | * Examples: Parcels, Roads, Buildings | * Examples: Elevation, Temp, Pressure |
   | * Natural Target Format: VECTOR       | * Natural Target Format: RASTER       |
   +---------------------------------------+---------------------------------------+

The Discrete Object Conceptual Model

The Discrete Object Model (also known as the entity-based view) treats geographic reality as an empty space populated by distinct, well-defined entities:

• Crisp Boundaries: Every object has a defined spatial perimeter that cleanly separates the feature from surrounding space (e.g., a property boundary surveyed to the nearest millimeter).
• Identifiable Identity: Each entity can be uniquely identified, counted, tracked, and modified independently of other features (e.g., Fire Hydrant #1042).
• Dimensionality: Entities have intrinsic geometric dimensionality: zero-dimensional points (wellheads), one-dimensional lines (water pipes), or two-dimensional polygons (building footprints).
• Linked Thematic Attributes: A set of descriptive tabular attributes belongs to the entire object as a whole.

The Continuous Field Conceptual Model

The Continuous Field Model treats geographic reality as a seamless mathematical continuum. Space is not empty; rather, every coordinate location $(X, Y)$ within the study domain has a measurable value of the variable:

• No Discrete Boundaries: The variable changes gradually, continuously, or smoothly across space. There are no sudden structural edges or empty voids between features.
• Pervasive Presence: You cannot ask "how many elevations are there in this valley?" because elevation exists everywhere. You can only evaluate the intensity, magnitude, or value of the field at a given position $(X, Y)$.
• Physical and Environmental Variables: Classic continuous field phenomena include topographic elevation, atmospheric temperature, barometric pressure, soil pH, ambient noise levels, and groundwater contaminant plumes.

---

Vector vs. Raster: Architectural Tradeoffs

While GIS practitioners often default to vector for discrete features and raster for continuous fields, either data model can represent either phenomenon—albeit with distinct trade-offs.

---

Representing Discrete Phenomena in Raster: Categorical Grids & The Mixed Pixel Problem

When discrete entities—such as land ownership parcels or soil survey units—are converted into raster format, the continuous grid structure introduces specific conceptual challenges.

                         THE MIXED PIXEL (MIXEL) PROBLEM
                         
            +---------------+---------------+  Cell A: 100% Forest (Pure)
            |               |    Forest     |  Cell B: 65% Forest / 35% Water (Mixed)
            |    Forest     |    (65%)      |  
            |    (100%)     |               |  Assignment Dilemma:
            +---------------+-------+-------+  - Majority Rule: Classified as Forest
            |     Water     | Water |       |  - Loses 35% water coverage entirely
            |     (40%)     | (80%) | Water |  - Sub-pixel mixel uncertainty
            |   Forest(60%) |       |       |
            +---------------+-------+-------+

Categorical (Thematic) Rasters and Value Attribute Tables

In a categorical raster, the cell value is an integer code that functions as an index into a Value Attribute Table (VAT):

• The cell value 1 might represent "Deciduous Forest," 2 represents "Commercial Development," and 3 represents "Agricultural Land."
• Zones: A zone consists of all cells across the grid that share the exact same integer value, regardless of whether they touch.
• Regions: A region consists of a spatially contiguous cluster of connected cells sharing the same value.

[!CAUTION] The Nominal Scale Trap: The numbers in a categorical raster are qualitative identifiers, not quantitative magnitudes. Calculating the "mean" of a raster where $1 = \text{Water}$ and $4 = \text{Urban}$ yields a meaningless decimal (e.g., $2.5$) that has no physical interpretation. Mathematical operations must be restricted to frequency, mode, or cross-tabulation.

The Mixed Pixel (Mixel) Problem

Because real-world boundaries do not conform to rectangular raster grids, cells that fall along the perimeter of two or more discrete units encompass multiple land cover types. These are termed mixed pixels (mixels).

To assign a value to a mixed pixel, raster conversion algorithms employ explicit decision rules:

1. Majority Rule (Dominant Area): The cell is assigned the class that occupies greater than 50% of the pixel's ground footprint. Outlying, narrow features (such as a 2-meter stream passing through a 30-meter cell) are completely obliterated.
2. Cell Center Rule: The class located at the exact mathematical center point of the cell is assigned to the entire pixel, regardless of what occupies the remainder of the cell.
3. Priority / Extreme Weight Rule: Fragile, high-consequence classes (e.g., rare wetland or transportation centerline) are assigned if present anywhere in the cell, regardless of percent coverage.
4. Sub-Pixel Classification (Spectral Unmixing): Advanced remote sensing algorithms model the spectral reflectance of a mixed pixel as a linear combination of pure "endmembers," outputting the fractional percentage of each class present within the cell.

---

Representing Continuous Fields in Vector: Isolines and TINs

Representing continuous surfaces within the vector model requires converting continuous gradients into discrete geometric features.

                    VECTOR REPRESENTATIONS OF CONTINUOUS FIELDS
                    
       Isolines (Contour Lines)             Triangulated Irregular Network (TIN)
       (Planar 1D Lines of Equal Value)     (Continuous Mesh of Planar 2D Triangles)
       
             /------ 120m ------\                         /\------/\
            /                    \                       /  \    /  \
           /   /---- 140m ----\   \                     /    \  /    \
          |   |                |   |                   /------\/------\
          |   |      160m      |   |                   \      /\      /
          |   |     (Peak)     |   |                    \    /  \    /
           \   \--------------/   /                      \  /    \  /
            \--------------------/                        \/------\/

1. Contour Lines (Isolines / Isarithms)

An isoline is a 1-dimensional vector line connecting points of equal numerical value along a continuous surface. Topographic elevation contours, atmospheric isotherms (temperature), and isobars (pressure) are common examples.

• Contour Interval: The vertical or numerical difference between adjacent contour lines. Smaller intervals capture finer detail but crowd the map in steep terrain.
• Index Contours: Heavily weighted contour lines (typically every fourth or fifth contour) labeled with their numerical value to facilitate visual interpretation.
• Analytical Limitations: Contour lines discretize a continuous surface into stepped intervals. They contain zero quantitative data about the terrain between lines; an analyst must interpolate to estimate elevation between contours. Furthermore, contour lines cannot easily model perfectly flat horizontal plains or overhangs without supplemental spot heights.

2. Triangulated Irregular Networks (TIN)

A TIN is a vector-based topological data structure that represents a continuous surface as a seamless sheet of contiguous, non-overlapping planar triangular facets. Each triangle vertex is anchored to an irregularly spaced sample point with an explicit $(X, Y, Z)$ coordinate.

Because TIN triangles are planar, the slope, aspect, and surface elevation at any point inside a triangle can be computed using linear planar interpolation between the three corner nodes.

3. Spot Heights and Point Lattices

A continuous surface can be represented as an array of discrete vector points $(X, Y, Z)$. In a point lattice, points are spaced at regular grid intervals. In a mass point array, points are sampled irregularly, placed at key terrain inflections (peaks, pits, saddle points).

---

Summary of Common Exam Traps

[!CAUTION] Exam Trap 4.7: Conflating Data Models (Vector/Raster) with Conceptual Reality (Discrete/Continuous). Candidates frequently believe that vector "equals" discrete and raster "equals" continuous. Vector models can represent continuous surfaces (using TINs or contour lines), and raster models can represent discrete features (using categorical integer grids). The conceptual model describes geographic reality; the spatial data model describes the digital file architecture.

[!CAUTION] Exam Trap 4.8: Statistical Calculations on Categorical Grids. A favorite GISP exam question presents a raster of soil types or land use codes and asks how to calculate the "average soil type" across a watershed. Mathematical averages (mean) are invalid for nominal categorical data. The only statistically valid summary metric for categorical rasters is the mode (the most frequently occurring class) or a proportional area percentage.

[!CAUTION] Exam Trap 4.9: Assuming Contour Lines Contain Complete Terrain Data. Contour lines do not store a continuous surface; they sample a surface at fixed vertical intervals. In spatial analysis, calculating slope directly from contour vector lines introduces severe bias because no elevation variation is recorded in the horizontal space between the contour lines.