Research Notes

Important findings and summaries within the research papers and concurrent discussions.

Paper: Goodchild - Spatial Information Science

Key Notes:

The paper was written for a conference, the 4th International Symposium on Spatial Data Handling, which took place in 1990. There is some irony in reading this specific conference paper which is over three decades old, due to that fact that the author focuses on the longevity and survival of GIScience as an accepted institutional discipline. They touch on GIS being recongized more as a technology, and presses that GIS itself should be the science, only supported by technology applications rather than GIS as a supporting measure within the sciences. A few pathways Goodchild laid down to accomplish this is by “identifying key scientific questions” in GIScience and expanding on them intellectually. Essentially, they call for three communities specifically, “users, vendors, and researchers” to focus on the foundations of GIScience and then create the tools required to answer critical questions.

Paper: How Long is the Coast of Britain? Statistical Self-Similarity and Fractional Dimensional - Mandelbrot, Good, & Proctor

This paper explores the concept of measuring complex geographical curves, such as coastlines, using fractal geometry.

  • Statistical Self-Similarity:
    • natural shapes, like coastlines, exhibit self-similarity
    • this means that each part of the coastline, when magnified, resembles the whole coastline
  • Fractional Dimension:
    • compared with traditional geometry, which assigns integer dimensions such as
      • point: 0-dimensional
      • line: 1-dimensional
      • polygon: 2-dimensional
      • coastline: 1-dimensional (vector line / grouping of vector lines and indices)
    • with fractal dimensions, the coast of Britain has a dimension of approximately 1.25
  • Measurement Challenges:
    • length of a coastline varies significantly depending on the measurement scale used
    • as the measurement unit becomes smaller, the total length increases, reflecting the intricate details of the coastline
      • larger scale \(\rightarrow\) greater perimeter
  • Practical Applications:
    • aside from geography, the concept of understanding the complexity of natural shapes and phenomena is crucial
    • the paper is foundational in the field of fractal geometry and highlights the limitations of traditional measurement methods for complex natural shapes

Paper: MAUP and Implications for Landscape Ecology - Jelinski & Wu, Nelson & Brewer

This paper examines the the modifiable areal unit problem (MAUP) in the context of landscape ecology.

  • Introduction of MAUP
    • In order to reduce bias and avoid spurious relationships when dealing with aggregated data and multiscaled spatial phenomena, tow problems must be addressed:
      • scale problem: same set of areal data is aggregated into several sets of larger areal units, with each combination leading to different data values and inferences
      • zoning problem: sets recombined into zones that are of the same size but located differently, again resulting in variation in data values and different conclusions
  • Using the imaging data, they addressed scale and zoning effects through several different simulations
  • Spatial Autocorrelation
    • degree of spatial clustering, or degree to which values at one locality are determined in part by the values at neighboring locations
  • Measure of Spatial Autocorrelation
    • Moran’s I
    • Geary’s c
  • Null Hypothesis
    • no spatial autocorrelation for a broad range of aggregation and zoning schems
    • did not use formal testing with p-values for individual hypotheses
    • still concluded that autocorrelation changes with scale, implying that there is evidence of a scale effect related to the MAUP across different aggregations
  • Approaching for Addressing MAUP
    • basic entity approach: advocates identification of individual entities that are not modifiable (base unit)
      • issues:
        • difficulty in identifying the base unit
        • too much detail can lead to overwhelming data collection, analysis, and modelling
    • optimal zoning approach: system that maximizes interzonal variation and minimizes intrazonal variation
      • issues:
        • Fothering and Wong (1991) concluded that an optimal zoning system which minimizes spatial autocorrelation of all possible combinations of variables is not possible
    • sensitivity analysis approach: get a sense of magnitude and scope of the varying statistical characteristics through a series of analyses
      • issues:
        • again, testing all possible combinations would be computationally and timely impractical
    • new / novel methods of analysis:
      • suggested methods:
        • visualization over statsitcal analysis
    • emphasis of spatial analysis on the rates of change: focus on how spatial patterns evolve over time rather than just looking at static snapshots
      • temporal dynamics: how spatial data changes over time periods to understand trends and patterns of change
      • rate of change metrics: metrics to quantify speed and direction of changes in spatial phenomena, which can provide more insights into the processes driving these changes
      • comparative analysis: comparing rates of change across different regions or scales to identify areas of rapid change or stability
  • Summary of Solutions:
    • Multi-Scale Analysis: Conducting analyses at multiple spatial scales to understand how results vary with scale changes. This helps in identifying the scale at which patterns are most pronounced.
    • Zoning Sensitivity Analysis: Testing different zoning schemes to see how changes in the boundaries of spatial units affect the results. This can help in understanding the robustness of the findings.
    • Hierarchical Modeling: Using hierarchical models that account for data at different levels of aggregation. This approach can help in integrating data from various scales and reducing the impact of MAUP.
    • Geostatistical Methods: Applying geostatistical techniques that consider spatial autocorrelation and provide more accurate estimates by accounting for the spatial structure of the data.
    • Simulation and Resampling Techniques: Using simulation and resampling methods to assess the variability of results under different zoning and scaling scenarios. This can provide insights into the stability of the findings.

Although the MAUP is an integral problem within the geospatial sciences, the authors state that MAUP isn’t really a problem, more that it is reflective “of the real systems are hierarchically structured.”

Paper: The Territorial Trap - Agnew

The territorial trap is split into 3 pieces which challenge conventional views in international relations:

  • States as Fixed Units of Sovereign Space:
    • flawed view of states as fixed, unchanging units of territory
    • overlooks historical and geographical variations in exercising power and control over territories
  • Domestic/Foreign Polarity:
    • this binary view fails to account for the complex interactions and influences
  • States as containers of Society:
    • the previous / accepted view: states as containers with societies held in by border
    • author’s challenge: dynamic and fluid nature of social, economic, and political processes that transcend boundaries

While states do have geographical boundaries, the interactions and influences that cross these boundaries make the concept of a state as a rigid, self-contained unit quite limited. Historical relevance, social dynamics, economic exchanges, and political processes often transcend these borders, leading to complex interdependencies and interactions between states.

This perspective encourages a more fluid and interconnected understanding of international relations, recognizing that states are part of a larger, dynamic global system rather than isolated entities.

Paper: Climate As a Risk Factor For Armed Conflict - Mach

This paper seeks to analyze climate change as a factor for armed conflict. Overall, it ranked low on the list of known drivers of organized armed conflict, at least on a direct basis. However, climate change affects resources, and there has been a general trend of climate change increasing, not decreasing, armed conflict. The authors also do state that in the event of climate-related resource scarcity, this can also lead to cooperation between groups.

Takeaways:

  • Climate change not a major predictor for organized armed conflict, or isn’t recognized as such. At the very least, experts disagree, especially when analyzed as a direct contributor of organized armed conflict.
  • Even in the case of resource-management, when climate change threatens or limits resources such as food-yield, cooperation over conflict is expected. Not to say this is true for all climate-related resource scarcity scenarios.
  • Again, experts have some disagreement over the direct contribution of climate change to organized armed conflict.

Discussion: GIS in the Wild

Roughly draws from The Territorial Trap and Mach - Climate As a Risk Factor For Armed Conflict.

In the following notes, state is used as a general term which can be used to identify whole countries or faction-controlled areas.

  • State scale is easier to model, but not always accurate. This can lead to simple stories which can be leaving out complex relationships, and thus be misleading.
  • Armed conflict represented in map, especially constrained to state boundaries, only show part of the story.
  • Political lines / disputed border can change via perspective, which can be illustrated by IP address pings when searching Google Maps.
  • Internal vs. External non-total sovereignty
    • Internal Sovereignty
      • Definition: Internal sovereignty refers to the supreme authority within a state. It is about who holds the ultimate power and how this power is exercised over the state’s territory and population.
      • Focus: It deals with the relationship between the sovereign power and its subjects, ensuring that the state can enforce laws and maintain order within its borders1.
      • Example: The ability of a government to implement policies, collect taxes, and regulate activities within its territory.
    • External Sovereignty
      • Definition: External sovereignty refers to the independence of a state from external control. It is the ability of a state to act autonomously and make decisions without interference from other states or external forces.
      • Focus: It emphasizes the state’s capacity to engage in international relations, enter treaties, and defend its territorial integrity against external threats2.
      • Example: A country’s right to determine its foreign policy and engage in diplomatic relations with other nations.
    • Non-Sovereignty
      • Internal Non-Sovereignty: This could refer to situations where a state lacks full control over its internal affairs, possibly due to internal conflicts, weak governance, or external influence.
      • External Non-Sovereignty: This might describe a state that is heavily influenced or controlled by external powers, limiting its ability to act independently on the international stage.
  • However, the domestic/foreign separation doesn’t necessarily stop at the border.
  • How to escape the territorial trap (theoretically):
    • shatterbelts (crush zones, shatter zones)
    • clash of civilizations (i.e. west vs east, bloody borders)
  • CLIMATE CHANGE AND ITS EFFECT OPERATES THROUGH MECHANISMS

Paper: Scale and Its Histories - Lukinbeal

  • very philosophical paper
  • The author presents scale as an ambiguous definition, however argues that scale should be considered as a mentifact, or abstract cultural trait
    • aids our bodies in daily life
    • basis of exchange, valuation, trade, and transactions that can occur in a moral and ethical way
  • In general, scale has grown to be conceived as
    • ontology: in the technical sciences, a structured framework that defines the relationships between concepts within a specific domain
      • primary: references a hierarchy
      • secondary: means to assess proportion (size)
    • epistemology: in the social sciences, a nature, origin, and limits of human knowledge
      • primary: a system for weighing, a means to estimate quantity including distance, assessing proportion
      • secondary: reference a hierarchy
  • Lam and Quattrochi (1992) give four modalities of scale in the natural sciences and GIScience:
    • cartographic (distance relations)
    • geographic (spatial extent)
    • operational (level at which processes operate)
      • especially significant for how natural and technical scienctists conduct studies and work
    • level of resolution (large scale / small scale, pixel quality)
  • scale as a representational analogy and technique
    • equates things based on similarity while omitting difference for the purpose of clarity
    • provides the organization structure for showing three-dimensional objects as two-dimensional representations
    • we have a reliance on scale to make sense of representations such as near/far, small/large, light/heavy, slow/fast
  • A distinguishing feature of geography is that it is dependent on visual images to illustrate knowledge
  • geography is about the science of studying the proportions of the world, chorography is a humanities-approach to understanding the nuances of smaller parts of the world

Paper: The scale issue in social and natural science - Marceau

  • Recognizes the ambiguity of scale’s definition, but offers a generally accepted version of it:
    • a central concept in the description of the hierarchical organization of our world
  • how processes operate and how they can be linked across scales
  • necessity to predict and control scale and aggregation effects on statistical results and modeling is also recognized as major goals to achieve
  • variability in statistical results and conclusions between scales or aggregation levels, meaning that interpretations aren’t expected to hold across scales, and leads to the concept of preferred scales
  • scales can be defined as a continuum through which entities, patterns, and processes can be observed and linked.
  • the scale problem / scale issue
    1. what is the appropriate spatial scale for the study of a particular geographical phenomenon?
    2. how can we adequately transfer information one spatial scale to another?
  • space
    • absolute: rigid, three-dimensional Euclidean geometry, trigonometric rules can be applied
    • relative: not immune to influence, space exists only with reference to spatial entities and processes (implication of patterns and processes with specific context)
    • example: areas may be close in absolute space but distant in relative space when time, rates, and interactions are considered
    • absolute and relative concepts should be seen as complementary
  • scale refers to the spatial dimensions at which entities, patterns, and processes can be observed and characterized.
  • scaling (generalization?): upscaling consists of taking information at smaller scales to derive processes at larger scales, downscaling consists of decomposing information at one scale into its constituents at smaller scales
  • ecology fallacy: error resulting from making statistical inferences from aggregate relationships to individual relationships
  • solutions for maup:
    • clustering: grouping a set of areal units to partition a study area in such a way that their number and configuration reveal as much information as possible about the phenomenon of interest
      • conventionally represented by a dendogram which shows the stages in grouping and the loss of detail that occurs during the process of aggregation
    • identification of basic geographical entities (basic spatial units)
      • however this approach is not always practical in a universal sense
      • BUT… identification of basic attributes defining the objects under investigation is a fundamental practical exercise that must be done to ensure the validity of any scientific inquiry
    • systematically perform sensitivity analysis
    • using principal axis factor analysis (PCA?), specific MAUP effects are strongly influenced by the presence of spatial autocorrelation in data
  • domain of scale: regions of a scale in which patterns do not change or change monotonically with scale
  • scale threshold: these domains are separated by thresholds, defined by relatively sharp transitions or critical points along the spatial scale continuum
    • identification of relevant domains of scale and scale thresholds when studying a specific phenomenon
    • evaluation of scale effect on the explanatory role of sets of variables
    • determination of scaling laws to adequately link information across scales
  • solutions to the scale problem
    • autocorrelation functions
    • correlograms
    • semivariograms
    • hierarchical path dynamics paradigm (HPDP)
      • theory of patch dynamics
  • final step: create a solid unified theoretical framework from which hypotheses can be derived and tested, and generalizations achieved

Paper: Scale in Geography - Montello

  • Concisely defines scale across three conventions:
    • cartographic scale: depicted feature on a map and its actual size in the world
    • analysis scale: size of the unit at which some problem is analyzed (county or state level)
    • phenomenon scale: size at which human or physical earth structures or processes exist, regardless of method of study or representation
  • the scale of the analysis should be the same scale of the phenomenon
    • inferences across scales can cause a cross-level fallacy

Paper: Modifiable Temporal Unit Problem (MTUP) and its Effect on Space-Time Cluster Detection

  • spatial aggregation requires consideration of problems such as the MAUP and ecological fallacy
  • analogously, temporal aggregation can lead to the Modifiable Temporal Unit Problem (MTUP), which can be examined across three effects:
    • aggregation
    • segmentation
    • boudnary
  • temporal aggregation effect:
    • process of converting observations from a fine interval into a coarse interval
    • can be important for creating meaningful analyses, with conversion of data into measures at regular time intervals (i.e. hourly vs daily vs monthly may make more sense depending on the analysis)
    • various forms of this, but most common is discretization of a fine interval into a course interval with events summed in bins
    • transforming from fine intervals to course intervals is also known as transforming to lower scales
  • temporal segmentation effect:
    • analogous to the zoning effect of MAUP, where spatial statistics vary depending on the adoption of different zoning patterns
    • a series of segmentations can be generated from a single time frame by simply varying the starting point of the temporal intervals
  • temporal boundary effect
    • in spatial analysis, a boundary is usually an arbitrary line drawn around a geographical area indicating its extent
    • extending to temporal analysis, this essentially refers to the starting points and ending points in a time interval
  • analysis via space-time scan statistics (STSS)
    • investigate whether an observed cluster of events has occurred by chance, assuming that the events are distrusted uniformly across the geographical region with no space-time interaction (i.e. a null hypothesis)
    • scans an area using a large set of overlapping windows moving across space and time
      • the window has a shape (i.e. cylinder), a base centroid (x, y), a radius (r), and a time length (t)
      • the size of the shape is increased systematically by space and time, generating a large number of cylinders with all parameters evaluated at each instance
      • statistical analysis is done on with highest likelihood ratios marked as primary candidates for true clusters (can be tested via Monte Carlo replication)

Paper: Linear Trends in Seasonal Vegetation Time Series and the MTUP - Jong

  • commonly, trends are quantified using linear regression methods, while the effect of serial autocorrelation is remediated by temporal aggregation over bins having fixed widths
  • just as different spatial aggregations can cause spurious results due to MAUP, so can different temporal aggregations can too
  • temporal bins (different time interval aggregation windows) can be regarded as temporal units and are modifiable
  • in their research, they found that coarser aggregation levels tend to overestimate change and result in higher variation in predictions with their OLS models