How kbs2007-22/isro works
1. This repository appears to be an architecture/visualization repo, not a full application codebase in the provided tree. The README contains a high-level system diagram for a geospatial air-quality analytics pipeline.
2. The project is split into two independent analytics flows that share a common theme: remote-sensing and ground-observation fusion for atmospheric monitoring.
3. The first flow is an AQI mapping pipeline. It ingests two data sources:
- Satellite and weather grid data
- CPCB ground-station measurements
These are combined as model inputs rather than treated as separate outputs.
4. The core computation in that AQI flow is a CNN-LSTM model. Architecturally, this is the main prediction layer that fuses spatial patterns from gridded inputs with temporal dynamics, implying both spatial and time-series processing.
5. The output of that first flow is “Daily Pan-India AQI Maps.” So the model boundary ends at geospatial map production, not at station-level analytics. The downstream consumer is a map layer or visualization/reporting layer.
6. The second flow is a hotspot attribution pipeline focused on formaldehyde (HCHO). It starts from:
- TROPOMI HCHO observations
- Active fire detections
These are separate upstream datasets merged for attribution analysis.
7. A wind vector overlay sits in the middle of the HCHO flow. That indicates a physical-context layer used to interpret plume transport and likely infer probable source direction or dispersion path.
8. The output of the second flow is “Source Attribution & Hotspot Maps.” This is a different class of product from the AQI map: instead of forecasting or estimating air quality, it explains likely emission sources and spatial hotspots.
9. The main architectural boundary is between data acquisition and geospatial analytics. Upstream systems provide satellite, weather, station, and fire observations; downstream systems consume derived raster/map products and attribution maps.
10. The project likely depends on geospatial processing throughout: gridded raster alignment, spatial overlay, and map generation. Those concerns are central because both flows operate on spatial data rather than tabular business records.
11. The two flows are not shown as directly coupled, so they should be diagrammed as parallel pipelines with shared environmental-domain context, not as a single end-to-end chain.
12. If you are drawing an architecture diagram, the key nodes are:
- Satellite/weather grids
- CPCB ground stations
- CNN-LSTM inference
- Daily AQI map output
- TROPOMI HCHO
- Active fire data
- Wind-vector context
- Source attribution / hotspot map output
13. In repository terms, the visible content suggests the README is the authoritative architecture description, and the “graph visualiser” directory is likely for rendering or presenting that diagram rather than implementing the data pipeline itself.
Main entry points: README.md README · graph visualiser Graph visualiser
Components 13
Repo artifacts
Repository README and the diagram renderer/viewer.
- README · Authoritative architecture description.
README.md - Graph visualiser · Renders or presents the architecture diagram.
graph visualiser
AQI flow
Daily air-quality mapping pipeline.
- Satellite grids · Satellite and weather gridded observations for AQI modeling.
- CPCB stations · Ground-station measurements used to anchor AQI estimation.
- Input fusion · Combines spatial and station signals before inference.
- CNN-LSTM · Fuses spatial patterns and temporal dynamics for AQI prediction.
- AQI maps · Daily pan-India AQI map products.
HCHO flow
Hotspot attribution pipeline for formaldehyde.
- TROPOMI HCHO · Formaldehyde observations used for hotspot attribution.
- Active fires · Fire detections used as a source-candidate signal.
- Wind context · Vector context for plume transport and source direction.
- Attribution · Infers likely sources and hotspot structure from combined signals.
- Hotspot maps · Source attribution and hotspot map products.
Other components
- Spatial align · Aligns gridded and point observations into a common spatial frame.
Connections 13
- README → Graph visualiser: describes · The README feeds the diagram visualiser.
- README → Satellite grids: documents · Architecture spec includes the AQI data inputs.
- README → TROPOMI HCHO: documents · Architecture spec includes the HCHO data inputs.
- Satellite grids → Spatial align: align · Gridded AQI inputs are normalized spatially.
- CPCB stations → Spatial align: align · Station observations are mapped into the shared frame.
- Spatial align → Input fusion: feed · Aligned inputs are fused for modeling.
- Input fusion → CNN-LSTM: input · Prepared features enter the CNN-LSTM model.
- CNN-LSTM → AQI maps: produce · Model inference generates daily AQI maps.
- TROPOMI HCHO → Spatial align: align · HCHO observations are spatially normalized.
- Active fires → Spatial align: align · Fire detections are brought into the common spatial frame.
- Wind context → Attribution: context · Wind vectors inform plume interpretation.
- Spatial align → Attribution: feed · Aligned remote-sensing and fire signals support attribution.
- Attribution → Hotspot maps: produce · Analysis generates source attribution and hotspot maps.