Bharat in Betastartup field guide
← Back to the atlas
No. 1180Problem statementEnergy and environment

Downscaling of Satellite based air quality map using AI/ML

Source

Source summary

Develop an AI/ML model using existing Python libraries to derive fine-resolution NO2 air-quality maps from coarse satellite observations. Handle large datasets and cloud-related gaps, choose and fit a suitable model, and validate predictions with unseen independent data. Intended users are researchers and government bodies assessing local air quality.

No. 1299Problem statementEnergy and environment

OceanEmbed - Satellite Embedding-Based Deep Learning Framework for Reconstruction of Subsurface Ocean Temperature from Surface Satellite Observations.

OceanEmbed should reconstruct subsurface temperature from surface satellite observations over the North Indian Ocean, from 5–30°N and 45–105°E. Daily inputs should be harmonized to 0.25° grids and include temperature, salinity, height/sea-level anomaly, currents and winds. Learned satellite embeddings should estimate profiles at 15 standard depths: 0, 5, 10, 20, 30, 50, 75, 100, 125, 150, 200, 300, 500, 700 and 1,000 m. Independent ARGO validation and a Bay of Bengal or Arabian Sea demonstration are required; the referenced dataset table is missing.

No. 1300Problem statementEnergy and environment

Develop a web-based interactive 3D visualization platform that integrates numerical ocean model outputs and in-situ observations.

INCOIS requests browser-based 3D visualization that combines ocean-model fields with Argo, glider, CTD and biogeochemical observations. It should ingest NetCDF and delimited text, show volumes, depth slices, profiles and time animation, and expose variable, colour, opacity and vertical-exaggeration controls. Modular data additions, lightweight APIs and open geospatial/CF standards should support institutional deployment and public science communication without desktop software dependencies.

No. 1390Problem statementEnergy and environment

AI-Based Detection and Classification of Industrial Fires and Persistent Thermal Sources Using NASA FIRMS, OSM & Satellite Data

Develop AI geospatial classification separating industrial fires and persistent thermal sources from natural/forest fires and other burning. Combine NASA FIRMS anomalies, OSM/land-cover, industrial-site data and satellite imagery, distinguishing contexts such as flares, mines, refineries and power plants. Store results in GIS and visualise map overlays. Thermal anomalies are inputs to classify, not automatic proof of an accident.