Geographic Information Systems (GIS) Analyses & Methods

All spatial analyses are performed within the context of a Geographic Information System. GIS comprises the software, hardware, and methodologies that allow spatial data to be displayed, modified and analyzed. One of the powerful capabilities of GIS is the ability to combine spatial data with non-spatial data, such as socio-economic and demographic (SED) survey data.

The development, modification and analysis of geospatial data (including remotely sensed imagery) occur within a GIS environment. Although GIS and remote sensing are commonly separated thematically, the processing and analysis of remotely sensed imagery falls within the realm of GIS. Some remote sensing information, therefore, is listed on this page as well as the Remote Sensing Web Pages.

Below is a list of the GIS-based spatial analytical methods and techniques that have been utilized up to this point in the CPC Ecuador Projects. Each method has a brief description, which will be expanded upon in the near future.

  • Nearest Community, Euclidean – Calculation of straight-line distance from each sampled household to the nearest community.
  • Nearest Community, Network – Calculation of the distance along the road network from each sampled household to the nearest community.
  • Distance to Nearest Water Source, Euclidean – Calculation of the straight-line distance from each sampled household to the nearest water source.
  • Land Use & Land Cover (LULC) Classification – While this technically falls under remote sensing, it is a GIS method nonetheless. Classification involves training an algorithm to assign an LULC class to each pixel in an image based on statistics generated from training data, which is itself generated from knowledge of the true LULC on the ground. See the Remote Sensing Web Pages for more detail.
  • Neural Network-Based Cloud Masking – Neural networks, a subset of artificial intelligence, are used to classify and remove clouds and cloud shadows from the remotely sensed imagery.
  • Change Detection – See the discussion of change detections on the Remote Sensing Web Pages.
  • Pattern Metrics – Using the LULC classifications, a variety of numeric indices were calculated that provide insight into the structure of the landscape, including contiguity and patchiness of land cover.
  • Integrated Spatial/Non-spatial Generalized Linear Mixed Modeling – Incorporation of SED data, spatial variables and pattern metrics into a GLMM to analyze SED determinants of LULC change.
  • Analysis of Reliability in Reported Land Use – The reliability of the amount of forest reported by each surveyed farmer was assessed. Correlations were performed between amount of forest on each farm from the LULC classification and the reported amount for the same farm.
  • Feature Extraction – Limited feature extraction of roads from high-resolution IKONOS satellite imagery and medium-resolution Landsat TM satellite imagery.
  • Cellular Automata (CA) Modeling – See Spatial Simulation Modeling for more details.
  • Agent-Based Modeling (ABM) – See Spatial Simulation Modeling for more details.