January 3: Introduction to the Course
Research in GIS is applicable to real world problem solving in a multitude of disciplines and industries. GIS allows for greater transparency of assumptions and methods so other GI scientists can see how previous knowledge was attained, and can produce identical results. Landscape ecology, crime analysis, and health geography are three perspectives of GIS-based research on which we will focus, using spatial analytical research methods.
Through class presentations, literature review assignments, and lab reports we will gain the necessary skills needed to perform the fundamental processes which makes GIS such a versatile tool in statistical problem solving: data collection and organization, GIS processing and analyzing, and presenting geographic data to be understood by the desired audience.
January 10: Why is Geography Important?
This week continues exploration of concepts and terms key to GIS and geography, and their role in geospatial analyses.
An issue surrounding geography is selecting the appropriate scale of space, time and organization pertaining to the phenomena under study. Results of spatial studies may not be independent from the spatial units being used. This has ethical and moral, as well as statistical, implications: aggregations of data collected at one scale have the potential to support biased opinions or inaccurate surveys (ex: gerrymandering, MAUP, connections between health, crime and poverty at the neighbourhood level versus census tract level). Scale and extent of study area are primary criteria of spatial analysis (ex: resolution of data), as increasing the extent of study area will reduce resolution, but may reveal patterns hidden at larger scales.
One must consider the geography of the problem as well: a fundamental idea behind GIS and geography in general is the spatial dependence of phenomena, or the effects that geography has alone on the object and the analysis performed on them (ex: spatial autocorrelation, the mere presence of of some quantity in a sampling unit makes its presence in neighbouring sampling units more or less likely).
We also reviewed the Kriging interpolation method, a geostatistical surface analysis which encompasses the structural component (environmental gradient), the spatially correlated component (biological interactions), and the random noise component to discover something about general properties of a surface and estimating properties of missing surface parts.
January 17: Understanding Landscape Metrics and the Link Between Pattern and Process
This weeks lecture contains some of the theory behind landscape ecology and defines some terms we are using in Fragstats for Lab 2. Terms we have been introduced to before (scale, pattern, spatial autocorrelation) are used to help understand landscape ecology (reciprocal interactions between spatial pattern and ecological process, occurring on spatially heterogeneous landscapes). Understanding the process/form relationship allows for descriptive and prescriptive actions with the data.
We explored methodological considerations (spatial autocorrelation, stationarity), processes (abiotic, biotic, anthropogenic, and disturbances), landscape structure and quantification of pattern. For example, topography, human settlement, and the presence of keystone species all create spatial patterns and heterogeneity. These are common considerations in landscape formation, and can be accounted for in GIS analysis.
Understanding what some of the choices we selected in Fragstats to calculate the changes in landscape in Edmonton from 1966-1976 is needed in order to establish a clear connection between the map and statistics behind it. During the Lab, it feels less like blindly pushing buttons and more like an aware and conscious decision.
January 24: Statistics, A Review
The lecture serves as a refresher to statistics to compliment our upcoming lab, which is focused on Geographically Weighted Regression. We revisit statistics terms such as measures of central tendency, measures of dispersion, and the different kinds of data (NOIR, Populations, Samples, sampling frames). Regression analysis, along with terminology and considerations (variables as dependent or independent, relations, number of cases, number of variables, and residuals) was reviewed as well. Gaining a basic understanding of the terms and parameters of linear and multiple regression analysis allows us as GIS users to quickly and easily interpret the results from correlation matrices, regression results, and and grouping analysis.
In Geographically Weighted Regression, coefficients (determined by examining neighbourhoods of points using distance decay functions) are allowed to vary spatially, and a weighting matrix is defined for each point to model spatially varying relationships. We can then ask the question: why? Are there policy implications, personal choice, social or economic factors? Regression and GWR can be used in an array of disciplines.
January 31: This day was dedicated to GEM presentations on Landscape Ecology.