GeoSpaX #What GeoSpaX borrows, and how it makes each feature lazy
GeoSpaX borrows proven features from commercial and open-source GIS, remote-sensing and data-science tools. It then automates them: smart defaults, auto-chosen methods, recipes and triggers (VISION.md, AUTOMATION.md).
Rules for borrowing
- From commercial products we borrow ideas and workflows only, through clean-room specs written from public docs. No decompiled code, copied UI assets or doc text (TECH_STACK §1).
- From open-source tools we reuse code where the license allows. GPL/AGPL engines (GRASS, SAGA, SNAP, R, OpenDroneMap) run as separate processes ("providers"), so the GeoSpaX core stays Apache-2.0 OR MIT.
How each feature is built: N native Rust engine · G GDAL/PDAL/PROJ in-process · P provider process · R R step · Py Python step · W WASM in the web edition · AI optional LLM.
Phase (see TECH_STACK §8): P1 Viewer · P2 Autopilot · P3 Field & sensors · P4 Remote-sensing & geostats autopilot · P5 Cartographer & publisher · P6 Platform. ✅ built · 🚧 started.
Plugins (1.8): one tool model that reads the manifests of five ecosystems — GeoSpaX plugin.yaml,
RustGIS .rgplugin (Apache-2.0), GeoLibre plugin.json + WASI tool bundles (MIT), QGIS metadata.txt
(algorithms through qgis_process) and ArcGIS Pro Config.daml + .pyt (ribbon imported, Python
toolbox tools run through arcpy; the .NET assembly cannot). Runtimes: recipe, process, python, r, wasm.
Each plugin reports honestly what it needs when a runtime is missing, and installing one makes it
removable again (uninstall never touches a --plugins dev folder) (PLUGINS.md).
Desktop workbench (1.7): ribbon with 40 commands, dockable panes (Contents/Explorer/Project · Symbology/**
Attributes**/Rasters/Devices · Automate/History/Log), command palette (Ctrl+K), paged attribute table with
zoom-to-row and charts, run history, status bar, resizable and collapsible docks, Identify popups. Structure modelled
on RustGIS (Apache-2.0: "everything is a command", ribbon/pane vocabulary) and GeoLibre (MIT: pane set,
attribute table as a peer of the map) — patterns re-implemented in TypeScript, no source copied
(THIRD_PARTY_NOTICES.md).
Ribbon tools (3.0.6): each Pro tab reviewed against ArcGIS Pro 3.6 and QGIS 4.0. View: Swipe,
Spyglass, Time slider, map grid, go to scale, lock scale, coordinate system search. Edit: split, merge, cut,
reshape, copy and paste, copy along line, trace, GPS point, attribute rules, check topology. Analysis: erase,
zonal statistics with exact cell coverage, network service area, closest facility and shortest path. Share:
a web map in one HTML file and project packages. Tested in CI by scripts/test-ribbon-tools.py and, in a real
browser, scripts/test-web-ribbon.mjs (USER_GUIDE §2b).
Built so far (automation engine v0 + Phase 2 slice 2 + sensor hub v1, October 2026):
- AutoMap rules that explain themselves, automatic page layouts with collision-free point labels, finished PNG/SVG/PDF maps and an HTML report;
- raster maps with relief shading, ramp and class legends;
- recipes with device / folder / schedule / after / stac / webhook / sensor triggers (threshold, geofence);
- SQL, buffer, hexbin and cross-validated IDW;
- spatial statistics (1.6): Gi* hot spots with automatic scale and FDR, global Moran's I, automatic kriging with standard-error maps, native and checked against R spdep/gstat;
- terrain (Copernicus DEM for any area, hillshade, slope, aspect, TRI/TPI, contours);
- remote sensing (STAC search, NDVI and other indices from the newest pass that is clear over the area, true colour);
- raster classification with areas, raster values at points;
- GRASS and SAGA providers;
- R steps (gstat kriging) and Python steps (point-pattern statistics);
- the device and sensor hub: sensor-logger CSV/JSON and FIT readings, live sensors over HTTP and MQTT (OwnTracks phones too), WHO air-quality guideline checks and classes;
- LiDAR (1.4: LAS/LAZ → DTM, DSM, canopy, buildings, tree tops) and 3D scenes (1.5: terrain block diagrams, extruded buildings, trees and point clouds → glTF, an offline viewer and a rendered picture, made by the recipes themselves).
These are marked ✅ / 🚧 in the tables below; details are in AUTOMATION.md, DEVICES.md, RESULTS §11, §12 and §13.
#Desktop GIS
#ArcGIS Pro (Esri, commercial)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Smart mapping style suggestions | AutoMap: profiles every layer and picks symbology, classification, ramp and labels, with a why line and 3 alternatives | N | P2 (✅ v0: field choice, method from skewness, ramp by meaning, labels, map-level rules, why lines; 3 alternatives next) |
| ModelBuilder, iterators, Tasks | Recipe graph view; for_each over files, features, dates or devices; guided recipes for first-timers |
N | P2 (🚧 recipes + runner + 13 gallery recipes ✅; graph view, for_each next) |
| Geoprocessing history | Every action is logged as a command; "Save as recipe" turns the session into a workflow | N | P2 (🚧 "Save this map as a recipe" ✅; session log next) |
| Geostatistical Analyst (kriging wizard, EBK, cross-validation) | Auto-kriging: three variogram models fitted, ordinary kriging cross-validated against IDW (lower error wins), prediction + standard-error maps, variogram chart, plain-language report | N (interpolate.auto); R (gstat, automap) / Py (PyKrige, GSTools) for more |
P4 (✅ v1.6 native interpolate.auto, equal to gstat krige/krige.cv; r-kriging recipe ✅; EBK, anisotropy, trends next) |
| Image Analyst, deep-learning classify pixels/objects | Pretrained land-cover and object models run as recipe steps; training samples suggested | Py (TorchGeo, SAM) / N (ONNX) | P4 |
| Layouts, map series, dynamic text | Auto layout (title, legend, scale bar, north arrow, credits) and auto atlas per feature or region | N (Typst/SVG) | P5 (🚧 auto layout → PNG/SVG/PDF ✅ in export.map, rasters with relief shading ✅; atlas next) |
| Local scenes: elevation surfaces, vertical exaggeration, extrusion, LAS dataset layers, Export to glTF / 3D Object | 3D scene in every LiDAR and terrain recipe: block diagram draped with map colours, relief shading, imagery or contours; buildings extruded to their heights, trees, point cloud; exaggeration chosen from the relief; glTF + offline viewer + picture in the report (AUTOMATION §6d) | N (own mesher, glTF writer, CPU renderer, WebGL viewer) | ✅ v1.5; LOD2 roofs, I3S/3D Tiles next |
| Network Analyst (routes, service areas) | "What can I reach in 15 min?" as one step (isochrones, OD matrices) | P (Valhalla/OSRM) / Py (osmnx) | P4 |
| Contents pane (drawing order, group layers, expandable legends), contextual tabs | Pro mode: a layer tree with Map node, group layers, twisties, drag into groups and a group switch; a contextual Layer tab for the selected layer. Lazy mode turns the same layers into cards that explain themselves | N | ✅ next |
| Swipe, Spyglass, time slider | Compare and Time on the View tab: swipe or a lens shows the map without the selected layer; the time slider plays any date or year field through a definition query, so tools and exports see what the map shows | N | ✅ 3.0.6 |
| Split, Merge, Cut, Reshape, Trace, attribute rules, topology checks | The same edits on the Edit tab, attributes kept and every edit undoable; attribute rules for value lists, defaults and required fields; Check topology returns overlaps, gaps and invalid shapes as a layer | G (GEOS) | ✅ 3.0.6 |
| Zonal Statistics as Table | Zonal statistics with exact cell coverage: each cell weighted by the share inside the polygon, as exactextract, so thin and small polygons get correct values | N | ✅ 3.0.6 |
| Geoprocessing pane (search tools, favourites, history) | Toolbox: recipes, building blocks and every plugin's tools in one ranked search; each building block lists the ArcGIS Pro / QGIS tools it stands in for, so their names find it; recently used on top | N | ✅ next |
| Pro assistant (help, SQL, ArcPy code) [1] | Ask pane: questions about the data answered through GeoSpaX's own tools (queries, statistics, selections, recipes without code), every tool call listed; local (Ollama, OpenAI-compatible) or Claude with your key; the same tools as an MCP server. Natural language → recipe: P2 | AI | ✅ Ask + MCP next; recipe writer P2 |
#QGIS 4 (GPL)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Processing toolbox with GDAL/GRASS/SAGA providers | Provider federation: one step vocabulary, and the best installed engine is picked automatically (and recorded) | N + P | P2 (🚧 GDAL in-process ✅, grass and saga steps ✅; automatic engine choice next) |
| Processing toolbox search across providers | The Toolbox searches GeoSpaX, GeoLibre (1,062 tools), QGIS (qgis_process), ArcGIS Pro (.pyt) and RustGIS tools at once, with a badge per source |
N | ✅ next |
| Graphical modeler, batch processing | Recipes, batch over folders and devices, triggers | N | P2 (🚧 recipes + device/folder/schedule/after/stac/webhook/sensor triggers ✅) |
| Expressions, data-defined overrides | DuckDB/SQL expressions in recipes and styles | N | P2 |
| Print layouts, atlas, layout charts and legend sync modes (new in 4.0) [2] | Auto layouts with charts; legend synced to visible layers by default | N | P5 (🚧 legend from styles with class counts ✅; charts next) |
| Labeling engine (collision, priorities) | Auto-labels: label field guessed (name/title/label), collision-free, priority by size or rank | N | P5 (🚧 points ✅: field guessed, 4-position collision avoidance; lines/polygons next) |
| GPS panel, QField companion | Device hub: live GNSS, phone streaming, OwnTracks over MQTT, auto-import, geofence triggers (DEVICES.md) | N | ✅ v1 (Linux mounts, NMEA, gpsd, phone page, MQTT); Windows/macOS MTP, BLE in P3 |
| Point-cloud algorithms (Height Above Ground, SMRF ground, virtual point clouds; new in 4.0) [2] | LiDAR autopilot: ground → DTM/DSM/CHM → building/tree extraction, started by dropping the files (AUTOMATION §6c) | N (las/laz-rs, own SMRF) |
✅ v1.4 (LAS/LAZ, vendor classes or SMRF, buildings, tree tops); colourised point cloud in 3D ✅ v1.5; E57/PLY next |
| 3D map view, Qgis2threejs export | The 3D scene opens from the run card (🧊 3D) and travels as one scene.html that works offline in any browser; scene.glb for other 3D tools |
N | ✅ v1.5; 3D inside the main map next |
| Geometry checker, fix geometries | Auto-clean on import (validity, duplicates, encoding, empty geometries), with a report | G/N | P2 |
| Georeferencer | Assisted georeferencing (control points suggested from a reference layer) | N | P5 |
#MapInfo Pro (Precisely, commercial)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Thematic map wizards, heat maps, temporal mapping [3] | Covered by AutoMap; time fields detected, so an animation is offered automatically | N | P2 |
| Spatial services: Search Nearby, Overlap, Summarize (v23.2) [3] | One-click "what's near / what overlaps / summarize by area" steps | N | P2 (🚧 buffer + hexbin summarise ✅) |
| SQL Select, saved queries, MapBasic | Plain-language filter → SQL; recipes + R/Python instead of a proprietary language | N / AI | P2 (🚧 vector.sql ✅, R/Python steps ✅) |
| Redistricting / territory balancing | "Balance territories by X" recipe (regionalization with constraints) | Py (PySAL max-p) | P4 |
| Drone data add-in [3] | Drone autopilot (device hub + OpenDroneMap) | P | P3 |
| 3DMap window (Vertical Mapper surfaces in 3D) | Any elevation raster → scene.3d (one step after dem.fetch, interpolate.idw, a GRASS/SAGA surface) |
N | ✅ v1.5 |
#GeoLibre (MIT) and RustGIS (Apache-2.0)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| 1,000+ WebAssembly geoprocessing tools (GeoLibre, geolibre-wasm / whitebox_next_gen) | Bundled: all 1,062 tools as the geolibre-tools plugin, run locally under Node's WASI; layers resolve to their files and results join the map (PLUGINS §1a) |
P (WASI) | ✅ next |
| DuckDB Spatial SQL workspace (GeoLibre) | SQL steps inside recipes, in the native engine (larger-than-memory) | N | P2 (🚧 SQLite/SpatiaLite ✅; DuckDB next) |
| Batch runner, model/pipeline chaining (GeoLibre) [4] | Recipes with triggers, caching and provenance | N | P2 (🚧 triggers, after chains, provenance ✅; cache next) |
| Auditable, undoable assistant operations (GeoLibre) [4] | The AI only emits recipe steps, so everything stays auditable, undoable and re-runnable | AI | P2 |
| SamGeo / SAM segmentation (GeoLibre) [4] | Segment-anything step in the RS autopilot | Py | P4 |
| Story maps, offline region download, Android (GeoLibre) [4] | Story generated from a recipe run (one step per slide); offline packs built automatically before field trips | N | P3/P5 |
| Command registry, undo, transaction log, CLI, sandboxed WASM plugins (RustGIS) [5] | Command log → recipes; geospax run CLI; WASM recipe steps and plugins |
N / W | P2/P6 (🚧 geospax run/check/watch ✅) |
#Global Mapper (Blue Marble, commercial; known for ease of use)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| "Open anything" format breadth, instant display | GDAL/PDAL breadth with auto-open, auto-CRS, auto-overviews | G | P1 ✅ / P2 |
| Terrain tools: volumes, contours, profiles | "Terrain report" recipe from any DEM | N / P | P4 (✅ terrain-report: Copernicus DEM, relief, slope classes with ha, contours, PDF; volumes and profiles next) |
#Open-source analysis engines (called as providers)
#GRASS 8.5 (GPL)
New in 8.5: a grass.tools Python API that calls tools as functions with NumPy I/O, wider parallelization, and JSON outputs [6]. We use it as a provider.
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
r.watershed, r.stream.*, r.fill.dir |
"Watersheds & streams" from a DEM (fetched automatically for the area if missing) | P | P4 |
r.sun, r.viewshed, r.cost, r.walk |
Solar potential, visibility and cost-distance as one-click recipes | P | P4 |
i.segment, i.maxlik, i.cluster |
Segmentation and classification steps in the RS autopilot | P | P4 |
Temporal framework (t.*) |
Time-series recipes (monthly NDVI trends, anomalies) | P | P4 |
r.geomorphon and any other module |
Landforms in plain words (flat, peak, ridge, …) with GRASS's colours; any GRASS script as a recipe step | P (grass --tmp-project --exec) |
P2 (✅ grass step, landforms-grass) |
NVIZ / m.nviz.image (3D surfaces with draped rasters and vectors) |
scene.3d drapes any raster (drape:) and paints vector layers on (overlay:), with no camera to set: the view and exaggeration are chosen and explained |
N | ✅ v1.5 |
#SAGA (GPL) · Whitebox (MIT open core) · Orfeo ToolBox (Apache-2.0)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| SAGA terrain analysis (TWI, LS-factor, channel network, TPI, landforms), kriging | Part of the terrain report; landform classes labelled in plain words | P (saga_cmd) |
P4 (✅ saga step, wetness-saga) |
| GDAL DEM tools (hillshade, slope, aspect, TRI, TPI, roughness, contours) | Built-in terrain steps with automatic z-factor, contour interval and smoothing | N (GDAL in-process) | P2 (✅ terrain.derive, terrain.contours) |
| Whitebox hydrology, LiDAR and terrain tools (500+) | Default native provider for terrain/hydrology steps | N/P | P2 |
| OTB streaming apps (pan-sharpening, LSMS segmentation, SVM/RF classification, SAR) | Large-image steps with automatic tiling and streaming | P (otbcli) |
P4 |
#Remote sensing and photogrammetry
#ERDAS IMAGINE / Octave Imagine (Hexagon, commercial)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Spatial Modeler, "no-code workflows with AI and automation" [7] | Recipes (graph view + YAML) | N | P2 (🚧 YAML ✅; graph view next) |
| Indices, band math, Recode, Subset | rs.index (NDVI, NDWI, NBR, NDMI, SAVI, EVI, cloud-masked), raster.classify with areas, rs.fetch for the area only |
N | P2 (✅) |
| Supervised/unsupervised classification (ISODATA, max likelihood), accuracy assessment | Auto-classification: training samples suggested, accuracy report written automatically | P (OTB/GRASS) / Py | P4 |
| Orthorectification, mosaicking, pan-sharpening, ATCOR atmospheric correction [7] | Auto-ortho, mosaic, pan-sharpen; surface reflectance via open processors (Sen2Cor/6S-style) | P | P4 |
| Change detection (DeltaCue, zonal change) | "What changed?": two dates → change map + area table + report | N / P | P4 |
| Photogrammetry (semi-global matching dense clouds) [7] | Drone autopilot via OpenDroneMap (AGPL, separate process) | P | P3/P4 |
| VirtualGIS / 3D viewing of imagery on terrain | Imagery fetched in the run (rs.fetch) is draped on the terrain automatically |
N | ✅ v1.5 |
#ENVI (NV5 Geospatial) · eCognition (Trimble) · TerrSet (Clark Labs), all commercial
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| ENVI spectral tools: SAM, MNF, PPI, spectral libraries, band math | Index library picked automatically from sensor metadata (NDVI, NDWI, NBR, …); SAM against open spectral libraries | N / Py | P4 |
| ENVI Modeler | Recipes | N | P2 |
| eCognition object-based image analysis (multiresolution segmentation + rule sets) | OBIA recipes (segmentation → features → classifier), with rules suggested from samples | P / Py | P4 |
| TerrSet Land Change Modeler, Earth Trends Modeler | Land-change projection and trend/seasonality recipes | Py / R | P4 |
#ESA SNAP (GPL) · Google Earth Engine (cloud, proprietary)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
SNAP Sentinel-1/2/3 toolboxes, Graph Builder, gpt batch |
SAR autopilot: orbit → calibration → speckle → terrain correction → flood map | P (gpt) |
P4 |
| Earth Engine catalog, reducers, cloud-free composites, time series | STAC-based "virtual catalog": search, cloud mask, median composite and zonal stats for your area, locally or through openEO backends | N / Py | P4 (🚧 STAC search, per-area cloud check, masked indices ✅ (rs.*, latest-ndvi, stac trigger); composites, zonal stats next) |
| Copernicus DEM / SRTM access | The DEM for any area, fetched as COG windows and reprojected to UTM | N | P2 (✅ dem.fetch) |
| Earth Engine code editor | Recipes + notebooks; an optional connector uses your own Earth Engine/openEO credentials | Py | P4 |
#LiDAR and drones (PDAL, LAStools, OpenDroneMap, Pix4D/Metashape)
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
PDAL pipelines, LAStools-style ground/height/classify (lasground, lasheight, las2dem), lidR rasterize_terrain/locate_trees, ArcGIS LAS Dataset to Raster + Classify LAS Ground + Extract Buildings |
lidar-to-terrain: one drop → DTM (vendor classes or SMRF), DSM, canopy height, building outlines with roof heights, tree tops with heights, maps + report; 2 cm terrain error and 40/40 trees on the test scene |
N | ✅ v1.4 |
| Pix4D / Metashape / ODM processing | Plug in the SD card: orthomosaic, DSM, and NDVI if multispectral, then map + report | P (ODM) | P3/P4 |
| CloudCompare / Potree point-cloud viewing | The run's point cloud in the 3D scene (about 250,000 points in the file's own colours, else by class or elevation), next to the modelled terrain, buildings and trees | N | ✅ v1.5; full-density streaming (COPC/Potree-style) next |
#Spatial statistics, geostatistics, data science
#GeoDa / PySAL · R · Python
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| GeoDa ESDA: Moran's I, LISA clusters, spatial regression, regionalization (SKATER, max-p) | "Find hotspots / clusters": weights chosen automatically, significance explained, map + table | N (stats.hotspots); Py (PySAL esda/spreg) / R (spdep) for more |
P4 (✅ v1.6 global Moran's I + Gi* hot spots, equal to spdep; LISA clusters, regression, regionalization next) |
| ArcGIS Optimized Hot Spot Analysis + Incremental Spatial Autocorrelation | points-to-hotspots recipe: incidents counted in auto-sized hexagons, distance band from the first clustering peak, Gi* with FDR, hot/cold map at 90/95/99 % and the reasons in the report |
N | P4 (✅ v1.6: 1,120 incidents in 0.04 s, 1 M points in 2.2 s) |
| ArcGIS Spatial Statistics: Mean Center, Standard Distance, Directional Distribution, Average Nearest Neighbor | python-spatial-stats recipe: centre, spread, deviational ellipse and the Clark–Evans index (clustered / random / dispersed) in plain words, on the map |
Py | P2 (✅ gallery recipe, plain Python) |
| ArcGIS Generate Tessellation + Summarize Within; QGIS Count points in polygon | points-to-hexbins recipe: auto-sized hexagons with counts (or sum/mean), classes chosen from the distribution |
N | P2 (✅ aggregate.hexbin, 1M points in 0.5 s) |
R sf, terra, stars |
R steps receive layers as sf/terra objects (GeoParquet/COG exchange) |
R | P2 (✅ v1: sf in/out via GeoJSON/GPKG; GeoParquet/COG next) |
R gstat + automap::autoKrige |
The auto-kriging engine, fittingly an "automap" for lazy users | R | P4 |
R spatstat, mgcv, INLA |
Point-pattern analysis and spatial models as advanced recipe steps | R | P4 |
R tmap, mapsf, ggplot2 |
"Show me the R code" for any AutoMap; publication maps | R | P2/P5 |
Python geopandas, shapely, rasterio, xarray, dask |
Python steps with Arrow/GeoParquet exchange; notebooks beside the map | Py | P2 (✅ v1: GeoDataFrame or GeoJSON in, layers + numbers out) |
Python scikit-learn, PyKrige, GSTools, mgwr |
ML regression/classification, kriging, geographically weighted regression | Py | P4 |
Python pystac-client, stackstac/odc-stac, TorchGeo, segment-geospatial |
RS autopilot building blocks | Py | P4 |
| webR, Pyodide | Light R/Python steps in the browser edition, with no server-side install | W | P4 |
#Automation and lazy web mapping
#FME (Safe Software) · CARTO Workflows · Felt / kepler.gl / Datawrapper
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| FME transformers, automations (triggers, schedules, webhooks), validation | Recipe triggers, connectors, data-quality checks | N | P2 (✅ device / folder / schedule / after / stac / webhook / sensor triggers) |
| CARTO no-code workflows in data warehouses | Recipe steps against DuckDB, PostGIS and cloud warehouses | N | P6 |
| Felt: instant good-looking defaults, share links, natural language → map + style [8] | AutoMap + one-click publish to a static web map (PMTiles + MapLibre, no server needed) | N | P2/P5 (🚧 AutoMap + finished PNG/SVG map + report ✅; web-map publish next) |
| kepler.gl / Datawrapper: beautiful defaults with no setup | Colour-blind-safe palettes and typographic defaults out of the box | N | P2 |
#Field data capture
#Esri Field Maps / Survey123 · QField / Mergin Maps · ODK / KoboToolbox
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Form-based capture (XLSForm) | Import XLSForm/ODK/Kobo submissions; a recipe builds the map + summary | N | P3 |
| Offline field maps, sync | Offline packs made automatically; device-hub sync when the phone is plugged in or on Wi-Fi | N | P3 |
| Live location sharing | Phone browser streaming, OwnTracks over MQTT and GNSS feeds into the live layer; a phone entering an area starts a recipe | N | ✅ (phone page, NMEA, gpsd, MQTT; geofence trigger) |
#Real-time feeds and sensors
#ArcGIS Velocity / GeoEvent Server · Garmin Connect / Strava · OpenAQ / PurpleAir / Sensor.Community · R openair · OGC SensorThings
| Borrow | GeoSpaX lazy version | How | Phase |
|---|---|---|---|
| Velocity / GeoEvent: real-time feeds (MQTT, HTTP), geofencing, incident detection, alerts [9] | sensor trigger: threshold (above/below), geofence (inside: <polygon layer>), cooldown per device, on HTTP pushes, MQTT messages and GNSS fixes; the alert is a finished map + report (air-quality-alert) |
N | ✅ v1 |
| MQTT ingestion (IoT platforms, Home Assistant, Node-RED) [10] | Built-in MQTT 3.1.1 subscriber (--mqtt mqtt://host/topic/#): generic sensor JSON, flat numeric payloads, device id from payload or topic |
N | ✅ (mqtts:// and LoRaWAN payload mappings in P3) |
| OwnTracks phone tracking [11] | OwnTracks location messages become live phones (no GeoSpaX app on the phone) | N | ✅ |
| Garmin Connect / Strava activity import (FIT) [12] | Watches, bike computers and fitness sensors plug in like any device: track + per-second heart rate, cadence, power, temperature as a readings layer | N (fitparser) |
✅ |
| OpenAQ / PurpleAir / Sensor.Community air-quality maps | Logger CSV/JSON with coordinates → readings → PM classes at the WHO 2021 guideline and interim targets [13], share of readings above, hotspot map, report | N | ✅ |
R openair (pollution summaries, polar plots) |
sensor.summary (count, mean, p95, max, WHO exceedance); openair plots through the R step |
N / R | ✅ summary · polar plots P4 |
| OGC SensorThings API clients (FROST, OSGeo istSOS) | Poll Things/Datastreams/Observations into the live store, same triggers | N | P3 (next) |
| Phone sensor apps (SensorLog, phyphox exports) | Their CSV exports (locationLatitude(WGS84) …) are sensor logs without configuration |
N | ✅ (CSV) |
#Sources
- Esri, ArcGIS Pro assistant 3.6 beta (Jan 2026): https://www.esri.com/arcgis-blog/products/arcgis-pro/announcements/join-the-arcgis-pro-assistant-3-6-beta-help-shape-the-future-of-desktop-gis
- QGIS 4.0 changes: https://www.reddit.com/r/QGIS/comments/1rodk65/qgis_40_visual_changelog_official_video/ · https://giscarta.com/blog/creating-a-demographic-map-copy · https://usemaps.com/en/qgis-4-0-whats-changed-webinar-summary/
- MapInfo Pro v23.2 announcement (Nov 2025): https://groups.google.com/g/mapinfo-l/c/x4nndYq8KxU · feature list: https://www.pelicangeographics.com/pbmapinfo
- GeoLibre v1.3.0 and docs: https://github.com/opengeos/GeoLibre/discussions/385 · https://github.com/opengeos/geolibre/blob/main/docs/index.md
- RustGIS README: https://github.com/opengeos/RustGIS
- GRASS 8.5.0 release (May 2026): https://github.com/OSGeo/grass/releases/tag/8.5.0
- Hexagon, Octave Imagine (formerly ERDAS IMAGINE): https://hexagon.com/products/erdas-imagine
- Felt blog (Felt AI, MCP server): https://felt.com/blog
- Esri, ArcGIS Velocity overview: https://www.esri.com/en-us/arcgis/products/arcgis-velocity/overview
- OASIS, MQTT Version 3.1.1 (standard): https://docs.oasis-open.org/mqtt/mqtt/v3.1.1/mqtt-v3.1.1.html
- OwnTracks JSON format: https://owntracks.org/booklet/tech/json/
- Garmin, FIT protocol: https://developer.garmin.com/fit/protocol/
- WHO global air quality guidelines (2021): https://www.who.int/publications/i/item/9789240034228