Geographic_information_system

# Geographic information system A geographic information system (GIS) is an [[Information_system|information system]] organized around one load-bearing fact: every record carries coordinates. It fuses [[Data_storage|data storage]], [[Computational_geometry|computational geometry]], [[Statistics|statistical inference]], [[Digital_image_processing|image processing]], and cartographic display into a single queryable model of territory, so that "where" becomes an operator instead of an afterthought. The lineage is short and concrete — Roger Tomlinson's Canada Geographic Information System began digesting the Canada Land Inventory in 1963, and Howard Fisher's Harvard laboratory released the SYMAP mapping program in 1965 — yet the descendants now route ambulances, site wind farms, track [[Air_pollution|pollution plumes]], adjudicate floodplain insurance, and supply the substrate for [[Agent-based_model|agent-based]] evacuation [[Simulation|simulations]]. For this hub the interest is structural: a GIS is [[Data_engineering|engineered data]] promoted to shared infrastructure, many disciplines computing against one [[Earth]]. ## Raster and vector: two commitments about the world Every GIS stores space in one of two idioms. A raster commits to a grid: territory becomes an array of cells, each holding a value — elevation, rainfall, reflectance, land cover. Rasters are what [[Sensor|sensors]] emit; a satellite scene is a georeferenced image, so the raster toolkit inherits directly from [[Digital_image_processing|digital image processing]] and, increasingly, from [[Computer_vision|computer vision]] classifiers that label crops and rooftops cell by cell. A vector layer commits instead to objects: points, polylines, and polygons in a projected [[Cartesian_coordinate_system|Cartesian plane]], each geometry tied to attribute rows. Topological encodings store shared edges once, making adjacency and containment explicit structure — in effect an [[Adjacency_matrix|adjacency representation]] borrowed from [[Graph_theory|graph theory]] — which is what keeps a parcel fabric or census geography watertight under editing. The trade is permanent: raster algebra parallelizes trivially but [[Data_storage|storage]] grows quadratically with resolution, while vectors are compact and precise but make continuous-field operations awkward. Production systems keep both and convert as needed. ## Flattening a spheroid without lying too much Coordinates mean nothing except relative to a model of the planet. The current reference, WGS 84, treats [[Earth]] as an oblate ellipsoid with equatorial radius 6,378,137 m and flattening ≈ 1/298.257. Gauss's Theorema Egregium (1827) guarantees that no flat map preserves all distances, so every projection is a negotiated distortion — conformal maps keep angles, equal-area maps keep areas, and the Universal Transverse Mercator scheme caps the damage by slicing the globe into sixty 6°-wide zones. Distances that matter must be computed along [[Geodesic|geodesics]] on the ellipsoid, not across the paper. Heights add a second reference: elevations are measured against the geoid, an equipotential surface of [[Gravity|gravity]] that undulates by roughly ±100 m around the ellipsoid. And because the [[Crust_(geology)|crust]] drifts centimeters per year, modern datums carry epochs — position is a function of time. A GIS that skips this bookkeeping produces confident, [[Accuracy_and_precision|precisely wrong]] answers, which is why projection metadata travels with every layer. ## Overlay, buffers, and map algebra The core analytic move is overlay: stack layers on a common geometry and combine them cell-wise or polygon-wise. Ian McHarg did it manually with acetate sheets in 1969; Dana Tomlin formalized it in the 1980s as map algebra, whose local, focal, and zonal operators make a raster stack programmable. Buffering — everything within 300 m of a stream — plus Boolean and weighted overlay yields the suitability models that site landfills and solar farms. Continuous surfaces come from interpolation: inverse-distance weighting at the cheap end, kriging at the principled end. Kriging, descended from D. G. Krige's 1951 mine-valuation work, treats the field as a draw from a spatial [[Probability_distribution|probability distribution]], models its covariance, and returns an [[Estimation_theory|estimation variance]] with every predicted cell. Waldo Tobler's 1970 observation that spatial nearness predicts similarity is the empirical license for all of this, made quantitative in autocorrelation [[Statistics|statistics]] such as Moran's I. Point events become [[Heat_map|density surfaces]] by the same machinery, and [[Spatial_ecology|spatial ecology]] runs its habitat models on exactly these operators. ## The graph underneath: routing and networks Streets, pipes, transmission lines, and transit enter a GIS as [[Directed_graph|directed graphs]]: edges with impedances, turn restrictions, one-way flags. Shortest paths run on Dijkstra's 1959 [[Algorithm|algorithm]] and its A* descendants; drive-time service areas, closest-facility dispatch, and location–allocation models come straight from [[Operations_research|operations research]], with fleet [[Logistics|logistics]] as the industrial customer. The same view supports vulnerability analysis: [[Centrality|centrality]] measures flag the substations and bridges whose loss fragments a system, connecting municipal practice to [[Network_science|network science]] and [[Network_theory|network theory]] proper, and to [[Reliability_engineering|reliability engineering]] when utilities model outage propagation across the [[Electrical_grid|grid]]. That a tax-parcel viewer and a power-restoration optimizer share one substrate is the quiet argument for treating GIS as [[Systems_engineering|systems engineering]] rather than cartography. ## Error is a first-class citizen Every layer carries error: positional (the line sits 5 m off), attribute (the land-cover class is wrong), and conceptual (the wetland boundary is a modeling fiction). Overlay chains compound all three, and because spatial errors correlate, naive worst-case bounds mislead. The honest tool is [[Monte_Carlo_method|Monte Carlo]] perturbation — jitter inputs according to their error [[Probability_distribution|distributions]], re-run the analysis, and report the spread as an explicit [[Uncertainty|uncertainty]] budget. Aggregation adds its own trap: the modifiable areal unit problem means answers change when district boundaries do, the formal cousin of gerrymandering, and the ecological fallacy warns against reading individual behavior off areal [[Statistics|averages]]. Positional standards, ground control, and [[Accuracy_and_precision|accuracy]] metadata are the countermeasures, and [[Estimation_theory|estimation]] discipline is what separates a decision-grade layer from a pretty one. ## From land inventory to ambient infrastructure John Snow's 1854 map of London cholera deaths is the standard ancestor; the machine lineage starts with Tomlinson's CGIS in 1963 and SYMAP in 1965. [[Minnesota]] moved early — its Land Management Information System, begun in the late 1960s, gridded statewide land records into 40-acre cells a generation before desktop tools. Commercial software followed (ARC/INFO, 1982), then three step changes: Landsat imagery flowing since 1972, civilian positioning improving by an order of magnitude when GPS selective availability ended in May 2000, and smartphones fusing [[Radio_navigation|radio navigation]] with [[Inertial_navigation_system|inertial sensing]] into a pocket [[Sensor|sensor]] platform. [[Digitization]] of parcel and survey archives closed the loop. The result is ambient spatial computation: nutrient budgets for the [[Minnesota_River]] and [[Mississippi_River]] basins, [[Urban_metabolism|urban-metabolism]] accounts, [[Weather_forecasting|forecast]] grids, [[Earth_system_science|Earth-system]] model output, and [[Machine_learning|machine-learned]] land cover, all resolvable to shared coordinates — one state vector for the landscape, which is what a [[Systems_theory|systems]] view always wanted geography to be. **On the spine:** [[Information_system]] · [[Computational_geometry]] · [[Graph_theory]] · [[Monte_Carlo_method]] · [[Spatial_ecology]].  *Connected to the Apex Spine:* Geographic information system → [[Hydrology|Hydrology]] — [[WT!Thury_Hydrodynamics_Compendium|Compendium]] section 19, *Hydrology*.  ## Wikipedia : Wikitube **Strict pair:** [Wikipedia](https://en.wikipedia.org/wiki/Geographic_information_system) : [Wikitube](https://en.wikitube.io/wiki/Geographic_information_system) ## Previous hub tags Hubs: `Systems`. Portals: [[PORTAL_Graph_theory]], [[PORTAL_Reliability_engineering]], [[PORTAL_Information_theory]], [[PORTAL_Agent-based_model]], [[PORTAL_Monte_Carlo_method]]. --- *Repopulated 2026-08-12 · redlink fill · 0 deletions.*