KALOPATHOR®
S1 C-BAND · 360° SWIVEL

Monsoonal Early Warning System · Microwave radar detection at 30m resolution — sense, forecast, route, and alert before the waters rise.

LIVE MISSION OPERATIONAL CONSOLE

Spaceborne Radar Command

Access the full-resolution spatial intelligence interface, interactive MapLibre GL radar overlays, flood risk telemetry, and automated CAP 1.2 bilingual alerting.

EVENT TELEMETRY · 12 AUGUST 2024

National Inundation Footprint

FENI TRIPWIRE IoU 0.5338 · OPS MODEL d3v4.2
0polygons
Detected Flood Extents
Raw Sigmoid τ=0.5 Threshold
0km²
Submerged Terrain
Sentinel-1 Dual-Pol (VV/VH)
0.0Million
Exposed Population
HRSL 30m Settlement Overlay
0stations
FFWC River Telemetry
Real-Time Stage & Danger Levels
SPATIAL INTELLIGENCE & TELEMETRY

Operations Console Capabilities

Engineered for disaster management operators, hydrologists, and emergency dispatchers requiring sub-kilometer precision and honest uncertainty disclosure.

SWIPE COMPARISON

Split-Curtain Optical vs Radar Wipe

Compare NASA VIIRS 375m TrueColor pre-flood optical imagery directly against Sentinel-1 30m microwave SAR flood detection. Drag the curtain divider anywhere on the canvas to distinguish turbid rivers from flash-inundated agricultural fields.

RIVER TELEMETRY

115 FFWC Telemetry Hydrographs

Hover over any gauging station on the Jamuna, Meghna, Padma, or Surma rivers to inspect 24-hour rate of rise (▲ Rising / ▼ Falling), peak crest records, and Danger Level margins.

ACTIONABLE IMPACT

Multi-Hazard Peril Scorecards

Every detected flood polygon features a unified severity badge (Critical Lvl 5 to Moderate Lvl 2), exposed union demographics, severed transport infrastructure, and verified shelter routing.

COORDINATION

URL Deep-Linking & Real-Time Sync

Synchronize viewports, layer configurations, and active polygons directly into shareable URLs to coordinate emergency response personnel with zero ambiguity.

MISSION WORKFLOW

How to Use the Operations Console

5-STEP RAPID SITUATIONAL ASSESSMENT
STEP 01

Inspect Inundation Extents

Pan and zoom across Bangladesh on the MapLibre GL radar map. Click any red flood polygon or choose from the top affected regions list in the rail to open the Action Card showing exposed population and model provenance.
STEP 02

Engage Split-Curtain Swipe

Click the Swipe Split button in the top navigation bar. Drag the interactive divider bar across the map to compare the NASA VIIRS TrueColor optical ground truth on the left against the Sentinel-1 SAR microwave radar layer on the right.
STEP 03

Interrogate River Gauges

Select Gauges in the left workflow rail. Click any of the 115 national telemetry stations (red = danger, amber = warning, teal = normal) to inspect responsive SVG hydrographs with 24-hour rate of rise and flood crest telemetry.
STEP 04

Plan Evacuation & Shelters

When viewing an active flood incident, the console calculates Dijkstra depth-cost evacuation routes to designated elevated concrete shelters, warning operators if primary arterial roads are severed by standing water.
STEP 05

Review Alerts & Share View

Review bilingual OASIS CAP 1.2 digital emergency alerts gated by multi-sensor verification. Click Share in the top bar to copy a precision deep link with exact coordinates and layers for inter-agency coordination.
Verification Engine

Six Pinned Verification Stages

Every operational stage is deterministic, versioned, and bound by the Honesty Doctrine.

STAGE 01

SAR capture

Sentinel-1 IW GRD · VV+VH · 30 m

Sentinel-1 IW GRD · Dual-Pol (VV/VH) · 30m resolution

STAGE 02

Detection

d3v4.2 U-Net · τ=0.5 · 1,199 polygons

d3v4.2 U-Net · 6.3M parameters · τ=0.5 sigmoid threshold

STAGE 03

Forecast

LightGBM + uncertainty bands

LightGBM F5 Spatio-Temporal Model + GloFAS runoff integration

STAGE 04

Exposure

HRSL population overlay

Meta / Columbia CIESIN High Resolution Settlement Layer (30m)

STAGE 05

Routes & shelters

EVE passability → evacuation

EVE Dijkstra Network · Road breaches & concrete shelter graph

STAGE 06

Alert (CAP 1.2)

Bilingual draft → human approval

OASIS CAP 1.2 Digital Emergency Payload · Bilingual EN/BN

System Catalog

Complete 10-Layer Geospatial Directory

All physical sensors, optical basemaps, radar algorithms, and humanitarian exposure models in the console.

LAYER 01
Optical Reflectance
375m · Daily Refresh (~13:30 LT)
Sensor: Suomi NPP / VIIRS (Visible Spectrum)

NASA VIIRS TrueColor (375m)

NASA GIBS VIIRS SNPP Corrected Reflectance (EPSG:3857)

“The sky as the eye would see it from 824 kilometers above. Daily sun-synchronous passes frame the macro geography of the Ganges-Brahmaputra-Meghna delta.”

Physical Basis

Natural visual-spectrum composite capturing natural surface coloration, cloud dynamics, and river sedimentation plumes in the Bay of Bengal.

Technical Spec

Native EPSG:3857 WebMercator Level 9 tile matrix (256×256px), dynamic date query.

LAYER 02
Radar & Flood
30m · 6–12 Day Orbit Revisit
Sensor: Sentinel-1A/1B C-Band Synthetic Aperture Radar

Detected Flooding (Sentinel-1 SAR)

Sentinel-1 C-Band SAR (IW Mode, VV + VH, 30m)

“When torrential rain obscures the delta for weeks, radar sees in the dark, mapping flooded paddies and submerged homesteads at 30-meter precision.”

Physical Basis

Active microwave pulses penetrate monsoonal cloud decks. Smooth standing water acts as a specular reflector, bouncing radar energy away from the antenna (dark signature).

Technical Spec

d3v4.2 U-Net (EfficientNet-B0, 6ch, τ=0.5). 1,199 polygons across 21,954 km² with interactive Split-Curtain swipe against NASA VIIRS optical baseline.

LAYER 03
Radar & Flood
20m · Near Real-Time Overpass
Sensor: Sentinel-1 Multi-Algorithm Ensemble

Copernicus GFM Live Flood

Copernicus Emergency Management Service GFM WMS-T

“An independent European orbital witness providing multi-model cross-validation against national detection inferences.”

Physical Basis

Harmonized ensemble of three independent SAR flood algorithms (TU Wien, DLR, LIST) executing on all global Sentinel-1 downlinks.

Technical Spec

OGC WMS-T streaming layer with ISO 8601 temporal querying.

LAYER 04
Optical Reflectance
250m · 2-Day Rolling Window
Sensor: Terra + Aqua MODIS (NDWI / MNDWI Indices)

NASA MCDWD 2-Day Flood Composite

NASA LANCE Near Real-Time MCDWD (MODIS Composite)

“A rapid macro-catchment sensor. Grey denotes dry ground; bright red indicates flood water; cyan denotes sensor shadow.”

Physical Basis

Multi-spectral water indices across optical and thermal infrared bands identifying surface water anomalies during cloud breaks.

Technical Spec

NASA GIBS TileMatrixSet with daily compositing.

LAYER 05
Hydrology & Terrain
Vector Transects · Meters/Year Migration
Sensor: Multi-Temporal Landsat & Sentinel-2 Archives

River Erosion & Bankline Migration

Centroid Transect Vector Migration (2016–2021)

“Bangladesh is living land that breathes and carves. The Jamuna and Padma shift entire kilometers in a single flood season.”

Physical Basis

Morphological contour tracking across dynamic braided river systems (Jamuna, Padma, Lower Meghna) measuring structural bank collapse.

Technical Spec

Shoreline vector polygons, erosion hazard zones, and longitudinal rate transects.

LAYER 06
Hydrology & Terrain
Point Telemetry · 3-Hour Update Cycle
Sensor: 115 Automated & Observer River Gauges

Water Level Telemetry Gauges

Flood Forecasting and Warning Centre (FFWC) Network

“The ground-truth pulse of the rivers anchoring satellite inferences with observed stage hydrographs.”

Physical Basis

Physical water level gauges measuring river stage height relative to Public Works Datum (PWD) and historical Danger Levels.

Technical Spec

Interactive SVG hydrographs with 24h rising/falling rate, danger thresholds, and real-time pointer crosshair tracking across 115 national stations.

LAYER 07
Exposure & Routing
30m Spatial Grid · National Extent
Sensor: Building Footprint CV + National Census Microdata

Population Exposure (HRSL 30m)

Meta / Columbia CIESIN High Resolution Settlement Layer

“Flood polygons are geographical data; people are the human reality. Computes affected population down to individual unions.”

Physical Basis

Spatial intersection of SAR inundation polygons with high-resolution population density rasters aggregated to Union Parishad administrative units.

Technical Spec

20.5M total national exposure estimate (Feni basin: 184,615 directly exposed).

LAYER 08
Exposure & Routing
T+1, T+3, T+5, T+7 Lead Days
Sensor: GloFAS Runoff + Open-Meteo + Antecedent S1 Saturation

Forecast Risk Horizon (T+1 to T+7)

LightGBM F5 Two-Branch Spatio-Temporal Model

“Water that has fallen in the transboundary hills today arrives in the haor basin in 48 hours. Forecasts the wave before it crests.”

Physical Basis

Hydrological wave propagation estimating downstream inundation expansion from upstream Meghalaya/Assam hill runoff.

Technical Spec

Probability rasters stamped with explicit honest uncertainty bounds.

LAYER 09
Exposure & Routing
Segment-Level Road Passability
Sensor: OpenStreetMap Road Topology + LGED Shelter Inventory

Evacuation Routing & Shelters (EVE)

EVE (Evacuation Vector Engine) Depth-Cost Dijkstra Graph

“Knowing where the water is is only half the battle; knowing where to run is life-saving.”

Physical Basis

Depth-cost graph traversal routing stranded communities along elevated dry ridges to designated concrete flood and cyclone shelters.

Technical Spec

Honest fallback: If roads are severed, triggers shelter-in-place boat rescue guidance.

LAYER 10
Exposure & Routing
Digital Warning Payload (SMS/Broadcast/Web)
Sensor: Automated 4-Gate Alert Verification Engine

CAP 1.2 Bilingual Alerting

OASIS Common Alerting Protocol v1.2 (Bilingual EN/BN)

“Converts complex radar and demographic intelligence into unambiguous, actionable public warnings.”

Physical Basis

Structured digital emergency payload requiring 4 verification gates (schema validity, exposure threshold, multi-sensor corroboration, human sign-off).

Technical Spec

Bilingual Bengali and English advisories with time-bounded go-before instructions.

Cloud Infrastructure

Backend Status, Storage & Upgrades

Zero static secrets, permanent Google Cloud ADC OAuth token refresh at $0.00/mo, and automated gating transparency.

INFRASTRUCTURE 01

Permanent ADC OAuth Storage ($0.00/mo)

GCP Project: project-300d4e0e-5c73-49bf-b8a

Google Cloud organization policy enforces disableServiceAccountKeyCreation. The pipeline authenticates through a verified ADC token refresh loop:

ADC Identity:monarqlabs@gmail.com
Token Cycle:Auto-refreshing short-lived bearer
GCS Bucket:gs://monarqlabs-gemini-workspace/kalopathor/
Compute Cost:$0.00 / month
HONESTY GATING 02

Why Live Sentinel-1 Loop Was Paused

Operational Safety & Calibration Gate

1. Broken Conformal Calibration (A1 Gate):

One-sided lower quantile regression on binary masks produced a tautological coverage (0/10 publishable deciles). Under the Honesty Doctrine, automated advisories with uncalibrated percentages are blocked.

2. Channel 5 Dry-Reference Soil Moisture Shift:

Automated dry-season mosaics raised Negative Control False Positive Rate from 0.335 to 0.611. Live triggers were safely paused until the index upgrade.

Immediate Upgrades

Three Ready-to-Deploy Zero-Compute Upgrades

UPGRADE A · ZERO GPU COMPUTE

CONSEMA Morphological Margin Bands

Replaces the broken parametric conformal band with deterministic 2D Euclidean morphological distance transforms (scipy.ndimage.distance_transform_edt). Computes core inundation via erosion and uncertainty margins bounded by HAND gradients in <400ms on CPU.

UPGRADE B · ZERO DEPENDENCY

Channel 5 Swap to SDWI (Dual-Pol Index)

Eliminates the external Google Earth Engine 80GB dry-reference requirement using SDWI = ln(10 · VV · VH) - 8. Cancels seasonal soil dielectric moisture shifts and isolates specular open water without false positive spikes.

UPGRADE C · ~280ms/TILE LATENCY

Sub-Second CPU ONNX Export

Converts the 6.3M parameter PyTorch checkpoint (d3v4.2_best.pt) into an optimized ONNX runtime graph. Cuts memory footprint from 4.2GB (PyTorch+CUDA) to <350MB on standard lightweight CPU workers.

Nothing is presented as live that isn't.

Honest by design

In life-safety systems, transparency is non-negotiable.

SEEDED / DEMO

Replay data is flagged amber everywhere it appears — never a silent green.

LIVE PILOT · FENI

One district fully real beats twenty half-real. Depth over breadth, on purpose.

EVIDENCE TRAIL

Every advisory links back to its SAR pass, model version and provenance.

Operational Commitments
No Ghost Data

Replay data labeled SEEDED DATA; never disguised as live telemetry.

No False Guarantees

Timestamps carry "estimate · pending recalibration".

Independent Auditing

Every advisory links directly to raw satellite pass timestamps & model version.

Depth Over Breadth

Feni benchmark validated before expanding to national breadth.