Advanced Remote Sensing and GNSS for Resource Management

Posted on May 29, 2026 in Geography

Detector Arrays and Sensor Calibration

1. Pushbroom Architecture: Pushbroom scanners use CCD or CMOS arrays consisting of thousands of detectors aligned perpendicular to the flight direction. Each detector continuously images a specific ground pixel line as the satellite moves forward.

2. Detector Material Types: Silicon CCD is used for visible/NIR (0.4–1.1 µm), InGaAs for SWIR (1–1.7 µm), and Mercury Cadmium Telluride (MCT) for thermal infrared. Each material has a different quantum efficiency suitable for specific wavelength ranges.

3. Detector Non-Uniformity: Each detector has slightly different gain and offset values, with variations up to 5–10%. This creates vertical striping artifacts in images. Relative calibration is applied to equalize all detectors across the array.

4. Noise Sources: Major noise types include shot noise (random photon arrival), dark current noise (thermal electrons), readout noise (electronics), and quantization noise (ADC). Cooling the detector reduces dark current significantly.

5. Radiometric Calibration: This converts raw Digital Number (DN) to physical radiance units. The formula is: L = Gain × DN + Bias. Gain and bias are determined in the laboratory before the launch and monitored during the mission using onboard calibration sources.

6. Dark Current Calibration: The camera shutter is closed periodically during the orbit. The detector output with no light input is measured and subtracted from the image data to remove the thermal noise contribution from the detector electronics.

7. Vicarious Calibration: After launch, the sensor images ground targets of known reflectance, such as desert playas and calibrated tarps. Results are compared with the actual reflectance to validate and update calibration coefficients.

8. Temporal Stability: Sensors degrade over time in space. Regular calibration generates time-dependent gain coefficients. Without this, trends in NDVI values could be artifacts of calibration rather than actual environmental changes on the ground.

Hyperspectral Data for Material Discrimination

1. Narrow Band Coverage: Hyperspectral sensors have 200+ contiguous bands with 5–10 nm bandwidth, compared to multispectral sensors with only 3–10 bands of 100–200 nm width. This fine spectral sampling captures diagnostic absorption features missed by multispectral sensors.

2. Spectral Signature Detection: Every material absorbs light at specific wavelengths, creating a unique spectral signature. Hyperspectral narrow bands detect these exact absorption positions, while wide multispectral bands average across features and lose diagnostic information.

3. Spectral Library Matching: Hyperspectral pixel spectra can be matched against USGS/JPL reference libraries containing thousands of material spectra using the Spectral Angle Mapper (SAM) algorithm. Multispectral data has insufficient bands for unique spectral matching.

4. Sub-pixel Detection: Linear spectral unmixing detects materials covering less than 10% of a single pixel. Multispectral data requires a material to cover 30–50% of a pixel for reliable detection, making it unsuitable for trace material identification.

5. Similar Color, Different Composition: Two rocks may appear identical in RGB imagery but have completely different mineralogy. For example, calcite and dolomite are both carbonates but have different SWIR absorption positions. Hyperspectral distinguishes them; multispectral cannot.

6. Quantitative Concentration Mapping: Absorption band depth correlates with material abundance (e.g., chlorophyll, water content, lignin). Hyperspectral imaging provides quantitative concentration maps, while multispectral gives only qualitative presence or absence classification.

7. Geological Case Study: AVIRIS hyperspectral data identified 20+ distinct minerals at Cuprite, Nevada, including alunite, kaolinite, and buddingtonite. The same area with Landsat MSS (4 bands) could only separate altered rock from unaltered rock, giving just 2 classes instead of 20+.

8. Economic Benefit: Hyperspectral mineral exploration reduces field sampling by 70% and discovery costs by 50%.

Detecting Industrial Heat Pollution

  1. Problem Statement: Industries discharge hot effluent into water bodies, raising temperatures, reducing dissolved oxygen, and destroying aquatic ecosystems; manual sampling misses most violations.
  2. Satellite Data: Landsat 8/9 TIRS (100 m, 16-day revisit) for routine monitoring; ECOSTRESS (70 m) for high frequency; Sentinel-3 (1 km, daily) for large-scale coverage; all are freely available.
  3. Radiometric Calibration: Raw TIRS digital numbers are converted to at-sensor radiance using gain and bias from metadata, then converted to brightness temperature using the inverse Planck’s Law.
  4. Atmospheric Correction: Water vapor causes a 2–5°C error in thermal data; a MODTRAN-based correction is applied using NCEP atmospheric profiles to remove this error and retrieve accurate surface temperatures.
  5. Emissivity Correction: Water emissivity is stable at 0.99; the NDVI threshold method identifies water pixels and assigns emissivity; the true Land Surface Temperature (LST) is then calculated.
  6. Thermal Anomaly Detection: Background temperature is measured upstream; the anomaly threshold is set as the mean plus three standard deviations. Any pixel exceeding the threshold is flagged as thermally polluted.
  7. Thermal Plume Mapping: Flagged pixels are grouped to form a plume polygon; the area, maximum temperature, and gradient from the discharge point are calculated and compared with permitted limits.
  8. Legal Evidence and Output: Satellite LST is validated with a ground thermal camera and serves as legal evidence in court; automated alerts are sent to the pollution control board when an anomaly is detected.

Monitoring Lake Eutrophication

  1. Project Objective: Assess water quality (chlorophyll-a, turbidity) and detect land-use changes around a local lake over five years using free Sentinel-2 and Landsat-8 satellite data.
  2. Study Area and Data: Any city lake showing algae blooms (e.g., Bellandur Lake, Bengaluru); Sentinel-2 (10 m, 5-day) for recent data; Landsat-8 (30 m) for a 10-year historical time series.
  3. Field Data Collection: GPS water sampling points are recorded; chlorophyll-a, turbidity, and Secchi depth are measured. The lake boundary GPS track and photographs of algae blooms are collected for ground truth.
  4. Water Quality Indices:
    • NDCI = (B5–B4)/(B5+B4): A higher value indicates more chlorophyll or algae.
    • Turbidity = B4/B3: A higher ratio indicates more suspended sediment.
    • NDWI = (B3–B8)/(B3+B8): This delineates the lake extent.
  5. Land Cover Mapping: A Random Forest supervised classification of a 500 m lake buffer zone into water, vegetation, built-up, and bare soil is performed; the 2020 map is compared with the 2025 map to quantify encroachment.
  6. Temporal Analysis: A monthly NDCI time-series is plotted to detect seasonal patterns and long-term trends; results are correlated with IMD rainfall and temperature records.
  7. Expected Outputs: Outputs include a chlorophyll-a concentration map (low/moderate/high/very high), a lake extent change map, and the identification of sewage inflow points as consistent high-chlorophyll patches near inlets.
  8. Deliverables and Tools: Deliverables include a project report, poster, and interactive web map on QGIS Cloud. Software used: QGIS with the SCP plugin and SNAP for Sentinel-2. Timeline: 6 weeks. A recommendation letter is sent to the municipality.

Satellite Imagery for Decision-Making

  1. Radiometric Correction: This removes sensor noise and converts DN to physical reflectance values; without this, two images of the same area on different dates cannot be compared for change detection.
  2. Geometric Correction: This removes distortions from Earth’s rotation, sensor geometry, and terrain relief; without this, distances, areas, and locations measured from imagery will be significantly incorrect.
  3. Road Alignment Planning: Orthorectified imagery provides accurate horizontal distances, slopes, and curve radii; engineers can plan road alignments without extensive field visits, saving time and costs.
  4. Pipeline Corridor Selection: Corrected imagery overlaid with slope, soil, and land-use maps helps select optimal routes, avoiding steep terrain, forests, settlements, and water bodies.
  5. Urban Drainage Design: An accurate DEM from corrected imagery enables correct watershed delineation and flood modeling; a wrong DEM gives incorrect flow directions, causing improperly designed drainage.
  6. Structural Deformation Monitoring: Corrected SAR imagery processed by InSAR detects millimeter-level ground movement over dams and bridges; uncorrected imagery shows false movement due to geometric errors.
  7. Land Use Zoning: Atmospherically corrected imagery accurately classifies urban, agricultural, forest, and wetland areas; uncorrected hazy imagery misclassifies urban areas as bare soil, leading to incorrect zoning decisions.
  8. Flood Plain Mapping: Corrected imagery provides an accurate DEM and land cover for hydraulic modeling; incorrect data leads to false flood risk maps, potentially causing construction in flood-prone areas.

Mineral Discrimination: Hyperspectral vs Multispectral

  1. Narrow Band Resolution: Mineral absorption features are 10–50 nm wide; hyperspectral bands of 5–10 nm resolve them clearly, whereas multispectral bands of 100–200 nm average across features, losing diagnostic shapes.
  2. Continuum Removal: Hyperspectral imaging allows the removal of broad background albedo to isolate specific mineral absorption features (e.g., the alunite doublet at 2.165 and 2.208 µm), which is impossible with multispectral data.
  3. Spectral Library Matching: USGS/JPL libraries are matched using the SAM algorithm (an angle < 0.1 radian indicates an excellent match); multispectral data has too few bands for reliable unique mineral identification.
  4. Clay Mineral Discrimination: Kaolinite (2.20 µm sharp doublet) vs. Montmorillonite (2.21 µm broad single): Hyperspectral separates them clearly, while a Landsat single SWIR band cannot detect either.
  5. Iron Oxide Discrimination: Hematite vs. Goethite show different absorption at 0.9 µm and 1.4 µm; hyperspectral 200 VNIR bands capture subtle shifts that multispectral blue/green/red/NIR bands cannot differentiate.
  6. Carbonate Discrimination: Calcite absorbs at 2.33 µm vs. Dolomite at 2.31 µm, only 20 nm apart; hyperspectral with 5 nm bandwidth resolves this, whereas a multispectral SWIR band (2.08–2.35 µm) cannot separate them.
  7. Quantitative Abundance Mapping: Hyperspectral linear unmixing calculates the percentage of each mineral per pixel, providing a quantitative abundance map; multispectral data gives only a binary present or absent result.
  8. Case Study: AVIRIS mapped 20+ minerals at Cuprite, Nevada, while Landsat MSS gave only two classes; hyperspectral imaging reduces field sampling by 70% and exploration costs by 50%.

Urban Heat Island Mapping Workflow

  1. Data Acquisition: Landsat 8/9 Band 10 TIRS (thermal), Band 4 (red), and Band 5 (NIR) are downloaded; a summer daytime image (10:30 AM local) is selected to capture maximum surface heating.
  2. Radiometric Calibration: Band 10 DN is converted to TOA radiance using the formula: Lλ = ML × Qcal + AL (where ML is the radiance multiplier and AL is the radiance add band from the metadata).
  3. Atmospheric Correction: MODTRAN/FLAASH is applied using NCEP atmospheric profiles; skipping this step causes a 2–5°C temperature error in the final Land Surface Temperature map.
  4. Brightness Temperature: Radiance is converted to brightness temperature using the inverse Planck’s Law: BT = K2 / ln(K1/Lλ + 1), using K1 and K2 calibration constants from the metadata.
  5. Emissivity Estimation: NDVI is calculated from Band 4 and Band 5; emissivity is assigned as: bare soil (NDVI < 0.2) ε=0.96, full vegetation (NDVI > 0.5) ε=0.98, mixed pixels are interpolated, and water ε=0.99.
  6. Land Surface Temperature: The true LST is calculated by correcting BT for emissivity: LST = BT / [1 + (λ × BT / ρ) × ln ε]; the output is then converted from Kelvin to Celsius.
  7. UHI Intensity Mapping: The rural reference LST is calculated from forests and farms outside the city; UHI intensity = urban LST – rural LST. It is classified as Low (0–2°C), Moderate (2–4°C), High (4–6°C), or Severe (>6°C).
  8. Outputs and Recommendations: Outputs include a color-coded UHI map, hot spot identification, and cool island mapping. Recommendations include green roofs, urban forests, reflective pavement, and water features in severe zones.

Hyperspectral Mineral Exploration Systems

  1. System Components: An airborne sensor (AVIRIS-NG/HySpex) or spaceborne sensor (PRISMA/EMIT) with GPS/IMU for georeferencing is used, along with a processing workstation with ENVI software or the Python SpectroPy library.
  2. Target Minerals: Target hydrothermal alteration minerals include Alunite, Kaolinite, Illite, Muscovite, Chlorite, Hematite, and Goethite; their presence indicates potential gold, copper, or zinc ore deposits below.
  3. Data Acquisition: The spectral range is 0.4–2.5 µm (VNIR-SWIR) for alteration minerals; the ground sampling distance is 5–10 m (airborne) with a swath width of 1–5 km. Solar illumination correction is required.
  4. Radiometric Calibration and Atmospheric Correction: DN is converted to surface reflectance using FLAASH or the empirical line method with ground calibration targets of known reflectance.
  5. Bad Band Removal: Water absorption bands at 1.4 µm, 1.9 µm, and 2.5 µm are removed; low SNR bands (<100:1) are also removed. This reduces 200+ bands to 150–180 usable bands for analysis.
  6. MNF Transformation: The Minimum Noise Fraction (MNF) transformation separates signal from noise and reduces dimensionality; the first 20–30 MNF components are retained, containing more than 95% of the useful signal.
  7. Endmember Extraction and Spectral Matching: The Pixel Purity Index (PPI) finds pure mineral pixels; spectra are matched to the USGS/JPL library using SAM (angle < 0.1 rad = excellent match). MTMF is used for low-abundance minerals.
  8. Output and Validation: Outputs include mineral distribution maps, alteration zone maps (phyllic/argillic/propylitic), and drill priority maps (high/medium/low). Field GPS validation is performed with XRD/XRF lab analysis; this method increases exploration success 3–5 times.

National-Level Operational Remote Sensing

  1. Crop Yield Estimation: The FASAL system uses SAR and optical data to provide district-level crop yield estimates 4–6 weeks before harvest for import/export and PDS decisions.
  2. Drought and Flood Assessment: NDVI anomaly maps identify drought districts; SAR flood maps are generated within 24 hours for NDRF deployment and crop insurance settlement.
  3. Forest Cover Monitoring: ISRO’s biennial assessment using LISS-III provides state-wise forest classification, mangrove mapping, and fire alerts via MODIS/VIIRS.
  4. Air Quality Monitoring: AOD from MODIS and Sentinel-5P maps PM2.5, NO₂, and SO₂; these are used by the National Clean Air Programme to rank cities and enforce GRAP measures.
  5. Reservoir and Snow Cover: India-WRIS tracks reservoir storage and Himalayan snow cover to support irrigation planning and CWC flood forecasting.
  6. Cyclone Tracking: INSAT-3D (every 15 min) provides cyclone position, intensity, and storm surge height for IMD early warnings and evacuation planning.
  7. Urban Sprawl Monitoring: NUIS uses LISS-IV (5.8 m) to map built-up areas, supporting master plan preparation, property tax assessment, and the AMRUT mission.
  8. Coastal Monitoring: CRZ mapping detects construction violations within 500 m of the high tide line; Cartosat-3 is used for border surveillance along the LoC and LAC.

Geospatial Tech in Natural Resource Management

  1. Groundwater Mapping: An overlay of geology, slope, lineament, and rainfall using AHP analysis identifies groundwater potential zones; Maharashtra’s check dams raised groundwater levels by 2–3 m.
  2. Watershed Prioritization: Morphometric analysis ranks sub-watersheds for soil conservation; treatment in the Narmada basin reduced soil loss by 40%.
  3. Reservoir Sedimentation: NDWI water spread vs. DEM estimates capacity loss; the Tawa Reservoir desiltation restored 15% of its storage capacity.
  4. Deforestation Monitoring: Annual Landsat/Sentinel-2 change detection monitors forest loss; Amazon’s PRODES detected a 13,000 km² loss in 2020–21, triggering enforcement.
  5. Forest Fire Risk Mapping: An overlay of fuel load, slope, aspect, and historical fires creates a risk map; Uttarakhand’s fire crews are pre-positioned based on these risk zones.
  6. Biodiversity Corridor Planning: Least-cost path analysis identifies wildlife corridors; underpasses in the Kanha-Pench corridor reduced human-elephant conflict by 60%.
  7. Soil Erosion Mapping: The RUSLE model using satellite inputs maps erosion risk; afforestation in Chambal’s ravines reduced erosion by 60% over 10 years.
  8. Solar Site Selection: An overlay of solar radiation, slope, transmission line proximity, and wasteland identifies sites; the Bhadla Solar Park (2.25 GW) was selected using this method.

Remote Sensing for Environmental Decisions

  1. Rapid Disaster Response: A SAR flood map is generated within 24 hours, whereas a ground survey for the same area takes 7–10 days, losing critical rescue time.
  2. Large Area Coverage: One Landsat scene (185×185 km) covers an entire district in minutes; an equivalent ground survey costs ₹50 lakh and takes two weeks.
  3. Frequent Revisit: Sentinel-2 monitors crops every five days, enabling timely fertilizer and pest control; monthly ground sampling misses critical stress periods.
  4. Objective Consistent Data: Satellites apply consistent definitions across time and space, unlike ground reports that vary due to observer bias.
  5. Beyond Human Vision: NDVI detects vegetation stress 2–3 weeks before visible browning, enabling early pest and disease control.
  6. Historical Archive: Landsat data since 1972 shows glacier retreat (10–20 m/year), coastal erosion, and desertification trends that are impossible to assess from ground data alone.
  7. Legal Evidence: Satellite change detection was used in the Aravalli illegal mining case, showing 200 ha of extra excavation; the Supreme Court imposed a ₹1000 crore penalty.
  8. Cost Effectiveness: ISRO’s Bhuvan costs ₹200 crore/year, but benefits exceed ₹8000 crore from insurance savings, flood reduction, and mining penalties, giving a BCR > 40:1.

Crop Monitoring Workflow and Technologies

  1. Data Acquisition: Sentinel-2 (10 m, 5-day) for NDVI/LAI, Sentinel-1 SAR for soil moisture, and Landsat-8 for thermal stress are used; all cloud-free seasonal images are acquired.
  2. Atmospheric Correction: Sen2Cor converts TOA to BOA reflectance, removing haze, water vapor, and aerosol effects for consistent NDVI values across all dates.
  3. Cloud Masking: The Sentinel-2 SCL layer masks clouds and shadows; cloudy pixels are replaced by the nearest cloud-free date, giving a composite every 10–15 days.
  4. NDVI Time-Series: NDVI is plotted against time to extract the emergence date (NDVI > 0.2), peak canopy, senescence date, and growing season duration.
  5. Derived Indices: LAI measures light interception; NDWI detects crop water stress; and TCI from Landsat thermal data identifies heat stress zones.
  6. Crop Type Mapping: Random Forest classification is performed using 10–15 multi-date images; classes include paddy, wheat, cotton, sugarcane, and fallow, with an accuracy > 90%.
  7. Stress Detection: Stress detection includes drought (NDWI < 0.1), pests (NDVI drop > 0.2 in 5 days), and floods (SAR backscatter drop > 3 dB); automatic SMS alerts are sent to farmers.
  8. Yield Estimation and Output: A yield model uses the NDVI peak, LAI, and rainfall data (accuracy ±10–15%); outputs include weekly crop health maps, stress alerts, yield forecasts, and a district dashboard.

GIS-Based Flood Risk Mapping Systems

  1. Data Requirements: Data requirements include a DEM (SRTM 30 m), SAR flood maps (Sentinel-1), IMD rainfall, CWC discharge, Sentinel-2 land use, census population, and critical infrastructure locations.
  2. Watershed Delineation: The DEM is processed in ArcHydro or QGIS to extract the stream network, delineate sub-basins, and assign curve numbers from land use and soil types.
  3. Rainfall-Runoff Modeling: HEC-HMS with the SCS-CN method uses an IMD IDF design storm; the output is the peak discharge (m³/s) at basin and sub-basin outlets.
  4. Hydraulic Modeling: A HEC-RAS 2D simulation with Manning’s n from land use is performed; flood depth and inundation extent are generated for 10, 25, 50, and 100-year return periods.
  5. Flood Hazard Zonation: Depth is classified as Low (0.1–0.5 m), Moderate (0.5–1.5 m), High (1.5–3 m), or Very High (> 3 m); this is combined with flow velocity for accurate hazard zones.
  6. Risk Assessment: Risk = Hazard × Vulnerability × Adaptive Capacity; this is overlaid with population density, poverty, and age groups to classify very high, high, moderate, and low-risk zones.
  7. Flood Risk Atlas: Output maps include flood hazard (100-year), population at risk, infrastructure at risk, evacuation routes, and no-build zones for land-use planning.
  8. Early Warning System: Real-time rainfall triggers an automatic model run; the flood forecast has a 6–24 hour lead time. SMS alerts are sent to the collector and SDRF; the Assam system reduced mortality by 80%.

Impact of Outdated DEMs on Flood Hazard Maps

  1. Elevation Inaccuracy: The SRTM 2000 DEM has a 10–20 m vertical error; new construction changes elevation by 2–5 m, causing a ±2 m flood depth error and incorrect hazard zone classification.
  2. Missing Man-made Features: New levees, embankments, and check dams are absent in the old DEM; the model computes incorrect flow paths, creating false inundation areas.
  3. River Channel Change: Meander migration and bank erosion over 10–20 years change river geometry; the old channel is shown, causing false positives and negatives in flood mapping.
  4. Reservoir Sedimentation: An old DEM overestimates reservoir storage; a sediment-filled reservoir releases more water, causing downstream flood peaks to be underestimated by the model.
  5. Runoff Coefficient Change: Urbanization raises impervious surfaces from 10% to 60%; the CN increases from 60 to 85. Peak discharge is underestimated by 30–50% when using old land-use data.
  6. Manning’s Roughness Error: Old land-use data shows forest (n=0.12), but the reality is agriculture (n=0.06); incorrect roughness overestimates the inundation extent, causing unnecessary land abandonment.
  7. Loss of Life: The 2015 Chennai floods: An outdated DEM showed low risk in a reclaimed floodplain; no evacuation was ordered, resulting in 500+ deaths. 80% of the drowned areas were previously mapped as safe.
  8. Policy Failure: The 2008 Bihar Kosi floods used 15-year-old maps, giving a false sense of security; after the disaster, the NDMA mandated annual hazard map updates as a compulsory requirement.

Cyclone Monitoring and Recovery Technologies

  1. Cyclone Track Monitoring: INSAT-3D imagery every 15 minutes with the Dvorak technique provides the cyclone position, wind speed, and category (1–5); warnings are issued at 72, 48, and 24 hours before landfall.
  2. Storm Surge Prediction: Satellite altimeters measure sea surface height; hydrodynamic models (ADCIRC) predict surge height (2–10 m) and coastal inundation for evacuation decisions.
  3. Rainfall and Wind Mapping: GPM measures rainfall rates through clouds; SAR captures a 1 km wind field. This guides port closures, flight cancellations, and power grid shutdowns.
  4. Vulnerable Population Mapping: A storm surge map overlaid with a population density grid estimates evacuation numbers (e.g., 500,000 for Cyclone Fani (2019) in Odisha).
  5. Shelter and Evacuation Routes: GIS assigns each village to the nearest cyclone shelter; least-cost path analysis defines evacuation routes, avoiding flooded and damaged roads.
  6. Immediate Damage Assessment: Cartosat-3 (0.25 m) or SAR change detection classifies buildings as collapsed, partially damaged, or intact; this guides rescue team prioritization.
  7. Agriculture Damage Assessment: The NDVI difference (pre- minus post-event) maps crop loss (paddy, coconut, cashew) for food aid distribution and compensation calculations.
  8. Long-term Recovery: Time-series NDVI monitors rebuilding and regrowth; for Cyclone Fani (2019), compensation of ₹1500 crore was disbursed in three months, and rebuilding took 18 months instead of 36.

Forest Fire Detection and Assessment Models

  1. Satellite Fire Detection: MODIS (1 km, 4–6 passes daily) and VIIRS (375 m) detect fires using thermal thresholding (BT > 360K); the output includes fire coordinates and Fire Radiative Power (FRP) in MW.
  2. Real-time Alert System: An automatic SMS is sent to the nearest forest guard within 30 minutes of detection, along with an email to the DFO and a WhatsApp message to village volunteers with the exact location.
  3. Fire Perimeter Confirmation: The Sentinel-2 SWIR band (1.6 µm) confirms the burn scar and fire perimeter; PlanetScope (3 m, daily) is used in emergencies for high-resolution mapping.
  4. Fuel Model Mapping: A forest-type GIS layer assigns fuel loads: teak/sal (40–60 t/ha), bamboo (60–80 t/ha), or pine (extreme); a Rothermel 13-type fuel model is assigned for simulation.
  5. Fire Spread Simulation: The FARSITE model uses the ignition point, fuel model, wind, temperature, humidity, and DEM slope/aspect; it predicts the fire perimeter at 6, 12, 24, and 48 hours.
  6. Real-time Dashboard: The current perimeter (red), 6-hour predicted (orange), 12-hour (yellow), and 24-hour (pink) perimeters are shown on a dashboard; villages at risk are highlighted for evacuation decisions.
  7. Burn Severity Mapping: dNBR = NBR_pre – NBR_post using Sentinel-2 SWIR bands; it is classified as unburned, low, moderate, or high severity, indicating everything from surface fires to complete canopy loss.
  8. Post-fire Damage Output: Carbon emissions (biomass lost × 0.5) are reported for the UNFCCC; outputs include timber loss, wildlife corridor impact, and soil erosion risk. The 2019 Bandipur fire was controlled in three days versus seven days previously.

NavIC in Civilian and Strategic Sectors

  1. Military Navigation: NavIC is used in warships, fighter aircraft, missiles, and troop movements in Ladakh and Arunachal without depending on foreign GPS.
  2. Missile Guidance: Agni, BrahMos, and Prithvi missiles use NavIC for mid-course guidance, even if GPS is denied during a conflict.
  3. Border Surveillance: NavIC is integrated with drones and ground sensors along the LoC and LAC for real-time positioning without foreign system dependency.
  4. Secure Military Timing: Encrypted NavIC signals are used for military radar synchronization, communication networks, and command-and-control systems.
  5. Strategic Autonomy: During the 1999 Kargil War, the USA denied GPS access to India; NavIC ensures India never faces such denial in future conflicts.
  6. Disaster Alerts: NavIC broadcasts cyclone, flood, and tsunami warnings directly to receivers within India without requiring internet connectivity.
  7. Fishermen Safety: NavIC provides IMBL alerts, preventing fishermen from accidentally crossing into Pakistani or Sri Lankan sea borders.
  8. Vehicle Tracking: NavIC chips are mandated in all commercial vehicles under the AIS-140 standard for emergency notification and fleet tracking.

GNSS-Based Public Bus Fleet Management

  1. On-Board Unit: Each bus is fitted with a NavIC + GPS receiver, GSM/4G modem, ESP32 microcontroller, and backup battery as an on-board unit (OBU).
  2. Data Transmission: The bus sends its ID, position, speed, heading, and passenger count to a cloud server every three seconds via the MQTT protocol.
  3. Cloud Backend: AWS or Azure stores data in a time-series database and runs arrival predictions, congestion detection, and route deviation alerts.
  4. User Interfaces: The driver receives the route on a tablet, office staff see a live map on a dashboard, and passengers see real-time bus locations and crowd levels on a mobile app.
  5. Data Flow: Data flow: GNSS → OBU → 4G → Cloud → Processing → App.
  6. Geofencing: Virtual boundaries (geofencing) at stops and depots automatically detect missed stops, route violations, and early departures.
  7. Passenger Benefit: Average waiting times are reduced from 15 to 5 minutes, and crowd levels are shown in the app to help avoid overcrowded buses.
  8. City Benefit: The city benefits from 8–12% fuel savings, a 20–30% increase in ridership, reduced traffic congestion, and better route-planning data.

Understanding GNSS Error Sources

  1. Ionospheric Delay: Free electrons at 100–1000 km altitude slow the GPS signal, causing a 5–50 m range error; this is worst in the afternoon and during high solar activity.
  2. Tropospheric Delay: Water vapor, pressure, and temperature in the lower atmosphere bend and slow the signal, causing a 2–20 m range error; the wet component is the hardest to model.
  3. Satellite Clock Error: Atomic clocks on satellites drift a few nanoseconds per hour; a 1-nanosecond drift equals a 30 cm position error. The residual error is 1–3 m after correction.
  4. Receiver Clock Error: A cheap quartz clock in a receiver drifts ~1 µs/sec; the same bias affects all measurements. A fourth satellite is used to solve this unknown clock error.
  5. Orbital/Ephemeris Error: The predicted satellite position differs from the true orbit due to Moon/Sun gravity and solar radiation pressure, causing a 1–5 m position error.
  6. Ionosphere Mitigation: Dual-frequency receivers compare L1 and L2 signals to cancel ionospheric delay, improving accuracy from 5 m to less than 1 m.
  7. Satellite Error Mitigation: DGPS and RTK use corrections from a nearby reference station to effectively remove satellite clock and orbit errors.
  8. Overall Impact: Combined uncorrected errors result in a 10–50 m position error; with DGPS/RTK corrections, accuracy improves to the centimeter level.

GNSS-Based Vehicle Tracking Systems

  1. Three-tier Architecture: Architecture: Edge layer (vehicle device) → Network layer (4G cellular) → Cloud layer (server, database, dashboard).
  2. Hardware: Hardware includes a GNSS receiver (u-blox NEO-M9N), ESP32 microcontroller, SIM7600 GSM, accelerometer, and a backup lithium battery on the vehicle.
  3. Data Acquisition: The GNSS locks every five seconds, recording latitude, longitude, altitude, speed, heading, HDOP, and satellite count.
  4. Data Transmission: Data in JSON format (e.g., {vehicle_id, lat, lon, speed, timestamp, ignition}) is sent via HTTP POST or MQTT to the cloud server.
  5. Cloud Processing: The server validates the data, stores 90 days of raw data, updates the live tracking cache, and triggers geofence violation checks.
  6. Power Management: The device uses a deep sleep mode (< 10 mA) between updates and wakes automatically on movement via an accelerometer interrupt.
  7. Real-time Outputs: Real-time outputs include a live map (updating every 3 seconds), route history, geofence alerts, and overspeed notifications sent via SMS or email.
  8. Analytics Outputs: Analytical outputs include daily distance, idle time, fuel estimates, driver behavior scores (e.g., harsh braking or acceleration), and maintenance reminders based on kilometers traveled.

Integrated GNSS and GIS Land Survey Solutions

  1. Hardware: Hardware includes an RTK-GNSS rover (1 cm accuracy), a base station or NTRIP correction via 4G, and a data collector tablet with QField software.
  2. Software: Software used: QGIS or ArcGIS Pro for desktop mapping, RTKLIB for post-processing, and ArcGIS Online for cloud sync and team sharing.
  3. Pre-Survey Planning: Existing cadastral maps and satellite imagery are loaded onto the tablet; data collection forms with validation rules are designed before the fieldwork.
  4. Data Collection: The surveyor occupies each boundary corner for 30 seconds (static), achieving 1 cm accuracy; attributes like boundary type and condition are recorded.
  5. Quality Control: Real-time alerts are generated for low satellite counts, a PDOP greater than 2, or a loop closure error greater than 5 cm, requiring repeat measurements.
  6. Post-Processing: Raw GNSS data is corrected using a CORS network or precise ephemeris, improving accuracy to 5–10 mm for control points.
  7. GIS Integration: Corrected points are imported into the GIS, snapped into closed polygons, and the area is calculated and overlaid with soil maps and flood zone layers.
  8. Final Outputs: Final outputs include a digital cadastral map (DWG/SHP), a contour map (0.5 m interval), a 3D terrain model, a legal area report, and as-built drawings.

GNSS and INS Integration Models

  1. GNSS Weakness: GPS signals can be blocked in tunnels, urban canyons, basements, or underwater, and are susceptible to jamming and multipath errors.
  2. INS Weakness: Accelerometers and gyroscopes provide continuous positioning, but drift accumulates at 1–10 meters per minute depending on sensor quality.
  3. Kalman Filter: The Kalman filter is the core integration engine that fuses GNSS positions with INS data; GNSS corrects INS drift, and INS bridges GNSS signal outages.
  4. System Components: System components include a MEMS IMU (accelerometer and gyroscope triad), a multi-constellation dual-frequency GNSS receiver, Kalman filter firmware, a magnetometer, and a barometer.
  5. Data Flow: Data flow: IMU measures at 100–1000 Hz → computes position/velocity/attitude → GNSS updates at 1–10 Hz → Kalman filter corrects IMU errors in a feedback loop.
  6. State Vector: The state vector consists of 15 states: 3 position, 3 velocity, 3 attitude, 3 accelerometer biases, and 3 gyroscope biases; these are propagated by the IMU and updated by the GNSS.
  7. GNSS Available Mode: In GNSS Available Mode, the Kalman filter continuously corrects INS errors, giving an overall accuracy of 0.5–2 meters depending on GNSS signal quality.
  8. GNSS Denied Mode: In GNSS Denied Mode, the INS continues alone; the error grows by 0.1–0.5% of the distance traveled (e.g., a 100 m tunnel results in only a 0.1–0.5 m error). A magnetometer prevents heading drift.

GPS Positioning Techniques and Error Sources

Techniques:

  1. Standard Point Positioning (SPP): This uses pseudorange only from broadcast ephemeris; it has an accuracy of 5–10 m and works with basic receivers like mobile phones and car GPS.
  2. Differential GPS (DGPS): A base station broadcasts pseudorange corrections via radio or 4G to the rover; it has an accuracy of 0.5–2 m and removes atmospheric, orbit, and clock errors.
  3. Real-Time Kinematic (RTK): This uses carrier phase measurements with base station corrections; it has an accuracy of 1–3 cm and requires high-end receivers and a radio communication link.
  4. Precise Point Positioning (PPP): This uses precise satellite ephemeris and clocks from the internet; it has an accuracy of 10–30 cm and requires no base station, though it has a 15–30 minute convergence time.
  5. Network RTK: Multiple CORS stations model spatially correlated errors over a large area; it has an accuracy of 1–5 cm and works up to 50 km from the nearest reference station.

Error Sources:

  1. Atmospheric Errors: These include ionospheric delay (5–50 m) and tropospheric delay (2–20 m); they are mitigated by dual-frequency receivers, SBAS, and atmospheric correction models.
  2. Satellite and Receiver Errors: These include satellite clock errors (1–3 m), orbit errors (1–5 m), receiver noise (0.1–1 m), and multipath interference from buildings (1–10 m).
  3. Geometric and Relativistic Errors: A high PDOP from clustered satellites reduces accuracy; the relativistic effect of +38 µs/day from speed and gravity is already corrected in the satellite signal.