CycloNet — Intelligent Tropical Cyclone Tracking & Satellite Intensity Estimation Platform

Next-Generation Meteorological & Infrastructure Vulnerability Intelligence Platform combining Deep Learning Computer Vision, GEE-Compatible Satellite Inundation Pipelines, Parametric Hydrodynamic Storm Surge Modeling, and Gemini 3.7 Flash Multimodal AI Reasoning.

---

CycloNet is an enterprise-grade meteorological intelligence, hydrodynamic storm surge forecasting, and critical infrastructure vulnerability platform. Engineered for meteorological departments, state disaster management authorities (SDMAs), and humanitarian response agencies across the Bay of Bengal and coastal APAC, CycloNet shifts disaster response from post-landfall recovery to proactive pre-landfall anticipatory action.

The platform unifies Deep Convolutional Neural Networks (ResNet-50), GEE-Compatible Sentinel-1 SAR & SRTM 30m DEM flood inundation modeling, Parametric Hydrodynamic Storm Surge Simulation (Jelesnianski Formulation), DEM-informed catchment rainfall damage pathway simulation, 5-factor critical infrastructure vulnerability scoring (IVF), parametric insurance liquidity disbursement triggers, and Gemini 3.7 Flash Multimodal AI for automated pre-landfall executive briefings in 7 regional languages.

---

📋 Table of Contents

• ✨ Key Platform Capabilities
• 🌊 Track 5: Cyclone Impact & Infrastructure Vulnerability Forecaster
• 🤖 Gemini 3.7 Flash Multimodal AI & Dual-Inference
• 🏗️ System Architecture & Pipelines
• 🧠 Deep Learning & Intensity Classification Architecture
• 🌊 Hydrodynamic Storm Surge & Catchment Modeling
• 🔌 Comprehensive REST API Reference
• 🛠️ Tech Stack Matrix
• 🔑 Environment Configuration
• 🚀 Quickstart & Installation
• 🧪 Running the Automated Test Suite
• 📝 License & Attribution

---

✨ Key Platform Capabilities

Feature	Description	Architectural Core
🛰️ Dual-Inference AI Classifier	Pairs fine-tuned ResNet-50 (PyTorch) with Gemini 3.7 Flash Multimodal Vision for dual-inference satellite verification.	PyTorch 2.2 + Gemini 3.7 Flash Vision API
🛡️ Out-of-Distribution (OOD) Guard	Detects and rejects non-meteorological images (landscapes, wallpapers) with high-confidence NOT A CYCLONE verdicts.	ResNet-50 2048-dim Latent Space Separation
🌊 Parametric Storm Surge Simulator	Physics-informed hydrodynamic surge simulation incorporating inverted barometer rise, shelf wind stress, and astronomical spring tide superposition.	Jelesnianski Parametric Formulation + Bathymetric Profiler
🛰️ GEE-Compatible SAR Inundation Pipeline	Inundation simulation utilizing Sentinel-1 SAR dual-pol (VV/VH), Sentinel-2 MNDWI, and SRTM 30m DEM elevation deficit schemas.	GEE-Compatible Geospatial Engine + GeoJSON Vector Output
⚡ 5-Factor Infrastructure Vulnerability (IVF)	Multi-factor mathematical scoring for 400kV / 220kV power grids, national/state evacuation highways, and cyclone shelters.	Continuous IVF Scoring Engine
💰 Parametric Insurance Liquidity Triggers	Automated smart policy payout rules triggered automatically based on sustained wind and surge thresholds within 7 days.	Parametric Threshold Evaluator
📢 Automated Multilingual Briefings	Synthesizes complex storm kinematics into executive risk memos in 7 languages (English, Hindi, Bengali, Odia, Telugu, Tamil, Gujarati).	Gemini 3.7 Flash Prompt Engine
🚨 NDMA-Compatible Multi-Channel Dispatches	Generates OASIS CAP-CP v1.2 XML feeds (Test status), SMS alerts, coastal siren triggers, VHF marine broadcasts, and SACHET notifications in sandbox mode.	CAP-CP v1.2 Sandbox Engine
🗺️ Geospatial Trajectory Tracking	Interactive Leaflet map displaying active storm centers, historical path waypoints, quadrant wind radii, and predictive cones of uncertainty.	Leaflet.js + GeoJSON GIS Layers

---

🌊 Track 5: Cyclone Impact & Infrastructure Vulnerability Forecaster

The Problem

Extreme weather events in the Bay of Bengal and coastal APAC require rapid anticipatory action. Shifting disaster response from post-landfall recovery to pre-landfall evacuation planning, infrastructure hardening, and parametric insurance liquidity saves lives and livelihoods.

The Solution

CycloNet implements an end-to-end predictive vulnerability pipeline:

1. GEE-Compatible Inundation Modeling: Employs Sentinel-1 SAR backscatter parameters combined with SRTM 30m DEM slope calculation to estimate inland flood penetration distance (with GEE cloud service connector and simulation fallback).
2. Compound Catchment Runoff Pathways: Evaluates precipitable moisture and basin topography to pinpoint flash flood choke points across river deltas (Mahanadi Delta, Hooghly Estuary, Godavari Basin).
3. Parametric Storm Surge Simulator: Computes oceanic water pileup using the Jelesnianski hydrodynamic formulation, factoring in central barometric pressure, astronomical tide stage, and forward speed.
4. Critical Asset Vulnerability Index (IVF): Evaluates power substations, transmission lines, highway corridors, and medical shelters using a 5-factor weighted index.
5. Parametric Insurance Liquidity Triggers: Establishes pre-agreed financial payout criteria ($V_{\text{max}} \ge 65\text{ kt}$, Surge $\ge 2.5\text{ m}$) for immediate disaster relief disbursement.

---

🤖 Gemini 3.7 Flash Multimodal AI & Dual-Inference

CycloNet leverages Gemini 3.7 Flash across three mission-critical capabilities:

1. Multimodal Satellite Frame Inspection:

   • Evaluates Dvorak T-numbers ($T3.0 - T7.0+$).
   • Measures eye structure morphology (pinhole, ragged, elliptical) and diameter ($\text{km}$).
   • Computes cloud-top brightness temperatures (deep convective cores down to $-80^\circ\text{C}$).
   • Detects Eyewall Replacement Cycles (ERC) and spiral band feeder asymmetry.

2. Automated Multilingual Pre-Landfall Executive Briefings:

   • Synthesizes storm trajectory, surge heights, power grid status, and evacuation priorities into structured municipal action memos.
   • Available in English, Hindi (हिन्दी), Bengali (বাংলা), Odia (ଓଡ଼ିଆ), Telugu (తెలుగు), Tamil (தமிழ்), and Gujarati (ગુજરાતી).

3. Domain-Specific Meteorological Copilot:

   • Context-aware chatbot with persistent access to real-time UI telemetry, SLOSH physics parameters, and GEE flood risk indices.

---

🏗️ System Architecture & Pipelines

High-Level Architecture

flowchart TB
    classDef clientStyle fill:#4F46E5,stroke:#3730A3,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef apiStyle fill:#0EA5E9,stroke:#0284C7,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef mlStyle fill:#E11D48,stroke:#BE123C,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef mapStyle fill:#0D9488,stroke:#0F766E,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef dbStyle fill:#1E40AF,stroke:#1D4ED8,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef extStyle fill:#7C3AED,stroke:#6D28D9,stroke-width:2px,color:#FFFFFF,font-weight:bold
    classDef geeStyle fill:#15803D,stroke:#166534,stroke-width:2px,color:#FFFFFF,font-weight:bold

    subgraph ClientLayer["🖥️ CLIENT PRESENTATION LAYER (Next.js 16 + React + Tailwind)"]
        Dashboard["Dashboard & Active Storm Monitor"]:::clientStyle
        ClassifierUI["AI Satellite Frame Classifier (Dual-Inference)"]:::clientStyle
        ForecastUI["Trajectory & Cone-of-Uncertainty Engine"]:::clientStyle
        InfraUI["Hydrodynamic Surge & GEE Flood Risk Center"]:::clientStyle
        ReportsUI["Multi-Channel Broadcast & CAP-CP v1.2 Advisories"]:::clientStyle
        ChatbotUI["Gemini 3.7 Flash Copilot (Voice-Enabled)"]:::clientStyle
    end

    Dashboard & ClassifierUI & ForecastUI & InfraUI & ReportsUI & ChatbotUI -->|REST API / JSON / Multipart Form| Gateway

    subgraph BackendLayer["⚡ FASTAPI SERVICE LAYER (Python 3.11+)"]
        Gateway["FastAPI Gateway & CORS Middleware"]:::apiStyle

        subgraph CoreServices["Backend Microservices"]
            MLService["🧠 PyTorch ResNet-50 Inference Engine"]:::mlStyle
            GeminiService["🤖 Gemini 3.7 Flash Multimodal AI Engine"]:::extStyle
            SurgeService["🌊 Parametric Hydrodynamic Surge Simulator"]:::mapStyle
            GEEService["🛰️ GEE-Compatible SAR Inundation Service"]:::geeStyle
            GISService["🗺️ GIS & Trajectory Projection Engine"]:::mapStyle
            ReportService["📑 CAP-CP v1.2 Sandbox Advisory Dispatcher"]:::dbStyle
        end

        Gateway --> MLService & GeminiService & SurgeService & GEEService & GISService & ReportService
    end

    subgraph ExternalData["🌐 EXTERNAL EARTH OBSERVATION & SATELLITE FEEDS"]
        GEE["GEE / Satellite Inundation Pipelines (Sentinel-1 SAR, Sentinel-2 MNDWI, SRTM 30m DEM)"]:::geeStyle
        IMD_NOAA["IMD / NOAA GFS Real-Time Gridded Meteorological Feeds"]:::mapStyle
        GEEService --> GEE
        GISService --> IMD_NOAA
    end

    subgraph StorageLayer["🗄️ PERSISTENCE & DATA STORAGE"]
        DB[("🗄️ SQLite Database / SQLAlchemy ORM")]:::dbStyle
        ReportService & MLService --> DB
    end

Loading

---

🌊 Hydrodynamic Storm Surge & Catchment Modeling

The surge engine calculates coastal sea surface elevation using the physics-informed Parametric Hydrodynamic Formulation (Jelesnianski Methodology):

$$S_{\text{peak}} = \alpha \cdot \Delta P + \beta \cdot V_{\text{max}}^2 \cdot \cos(\theta_{\text{coast}}) + H_{\text{tide}}$$

Where:

• $\Delta P = P_{\text{ambient}} - P_{\text{central}}$: Inverted Barometer Effect ($\sim 1\text{ cm}$ sea surface rise per $1\text{ hPa}$ atmospheric pressure drop).
• $V_{\text{max}}$: Maximum sustained surface wind speed ($\text{m/s}$ or $\text{knots}$).
• $\theta_{\text{coast}}$: Angle of storm approach relative to the coastal shelf normal vector ($75^\circ - 85^\circ$ creates peak shoaling).
• $\alpha, \beta$: Bathymetric friction and shelf shoaling coefficients ($1.45\times$ for shallow Bay of Bengal shelf vs $1.10\times$ for Arabian Sea).
• $H_{\text{tide}}$: Astronomical tidal height at landfall (Spring Tide vs Neap Tide superposition).
• Inland Flood Penetration Distance: Modeled via SRTM 30m DEM terrain slope ($\sim 0.75\text{ m/km}$).

---

🏗️ Multi-Factor Infrastructure Vulnerability Index (IVF)

CycloNet replaces basic map overlays with a continuous, 5-factor mathematical Infrastructure Vulnerability Index (IVF):

$$\text{IVF Score} = 0.30 \times \text{FloodDepth} + 0.25 \times \text{WindGust} + 0.15 \times \text{ElevationDeficit} + 0.10 \times \text{CoastalProximity} + 0.20 \times \text{GridCriticality}$$

IVF Score	Risk Classification	Action Directive
75 – 100	🔴 CRITICAL	Pre-emptively de-energize 400/220kV busbars at T-3h; switch medical shelters to isolated DG microgrids; close inundated road corridors.
50 – 74	🟠 HIGH	Erect mobile switchyard flood barricades; pre-stage NDRF water-rescue teams; restrict roads to emergency vehicles only.
25 – 49	🟡 MODERATE	Inspect water pump drainage sumps; verify 6-hour DG battery charge.
0 – 24	🟢 LOW	Routine baseline monitoring with telemetry standby.

---

🔌 Comprehensive REST API Reference

Interactive Swagger documentation is available at http://localhost:8000/docs.

1. Automated Early Warning Advisory Dispatch (/api/alerts)

Method	Endpoint	Description
POST	/api/alerts/dispatch	Automated Multi-Channel Dispatch (Sandbox): Generates OASIS CAP-CP v1.2 XML (Test status), trilingual Cell Broadcast SMS, NDMA SACHET JSON feeds, acoustic siren tower triggers, and Coast Guard VHF Ch 16 distress broadcasts.
POST	/api/alerts/broadcast-test	Simulates multi-channel emergency broadcast dispatch with delivery metrics.

2. Google Earth Engine & Inundation (/api/gee)

Method	Endpoint	Description
GET	/api/gee/layers	Lists available GEE-compatible satellite layers (Sentinel-1 SAR, Sentinel-2 MNDWI, SRTM 30m DEM).
GET	/api/gee/flood-inundation	Generates GEE-compatible Sentinel-1 SAR flood extent GeoJSON polygons for active cyclone coordinates.
GET	/api/gee/rainfall-pathways	Returns DEM-informed catchment precipitation accumulation, river discharge (cumecs), and drainage choke points (Diamond Harbour, Kakdwip, Sagar Island, Dhamra).

3. Hydrodynamic Storm Surge (/api/surge)

Method	Endpoint	Description
GET	/api/surge/calculate	Computes parametric peak surge height ($S_{\text{peak}}$), inverted barometer rise, and inland penetration.
GET	/api/surge/profile	Computes coastal sector-by-sector surge profile and critical asset risk ratings.

4. Gemini 3.7 Flash AI Engine (/api/ai & /api/chat)

Method	Endpoint	Description
POST	/api/ai/analyze-multimodal	Submits satellite image to Gemini 3.7 Flash for Dvorak T-number & eye inspection.
POST	/api/ai/pre-landfall-briefing	Generates executive pre-landfall disaster management risk memo in 7 languages.
POST	/api/chat	AI meteorological copilot query endpoint with live UI telemetry context.

5. Active Cyclone Monitoring & Trajectory (/api/active-systems & /api/history)

Method	Endpoint	Description
GET	/api/active-systems	Returns active tropical cyclones with current telemetry, pressure, and forecast trajectory.
GET	/api/history/search	Searches historical cyclone best-track archives (IBTrACS/IMD).

6. AI Satellite Classification (/api/classify)

Method	Endpoint	Description
POST	/api/classify	PyTorch ResNet-50 inference endpoint returning continuous softmax probabilities.

---

🛠️ Tech Stack Matrix

Component	Technology	Purpose
Backend Gateway	FastAPI 0.110+ (Python 3.11+)	High-throughput asynchronous REST API gateway
Multimodal Generative AI	Google Gemini 3.7 Flash	Satellite computer vision, pre-landfall briefings & chat
Earth Observation	Google Earth Engine (GEE)	Sentinel-1 SAR, Sentinel-2 MNDWI, SRTM 30m DEM
Deep Learning	PyTorch 2.2 + Torchvision	ResNet-50 classification backbone and MLP head
Hydrodynamic Modeling	SLOSH Formulation (NumPy/SciPy)	Parametric storm surge & tidal superposition modeling
Frontend Framework	Next.js 16 (App Router + Turbopack)	Modern React 19 web application
Geospatial GIS Mapping	Leaflet & React-Leaflet	Multi-layer GIS rendering for tracks, cones & flood polygons
Styling & UI Tokens	Tailwind CSS + Framer Motion	Glassmorphic, responsive dark/light theme
Relational Database	SQLite & SQLAlchemy 2.0	Persistent storm telemetry and audit logs

---

🔑 Environment Configuration

Backend (backend/.env)

APP_NAME=CycloNet
PORT=8000
DEBUG=True
DATABASE_URL=sqlite:///./cyclone_tracker.db

# Deep Learning Model
MODEL_CHECKPOINT_PATH=app/models/cyclone_classifier.pth
DEVICE=cpu                              # Options: "cpu" or "cuda"

# Gemini 3.7 Flash Multimodal AI
GEMINI_API_KEY=your_gemini_api_key_here

# CORS Configuration
CORS_ORIGINS=http://localhost:3000,http://127.0.0.1:3000

Frontend (frontend/.env.local)

NEXT_PUBLIC_API_URL=http://localhost:8000/api

---

🚀 Quickstart & Installation

Option 1: One-Click Windows Launch (Recommended)

start_all.bat

Launches both servers:

• FastAPI Backend: http://localhost:8000 (Docs: http://localhost:8000/docs)
• Next.js Frontend: http://localhost:3000

---

Option 2: Manual Developer Setup

1. Backend Setup (FastAPI)

cd backend
python -m venv venv

# Windows
.\venv\Scripts\Activate.ps1
# Linux / macOS
source venv/bin/activate

pip install -r requirements.txt
uvicorn app.main:app --reload --port 8000

2. Frontend Setup (Next.js)

cd frontend
npm install
npm run dev

---

🧪 Running the Automated Test Suite

Run Backend PyTest Suite (22 Unit & Integration Tests)

cd backend
python -m pytest tests/ -v

Run Frontend Production Build & TypeScript Check

cd frontend
npm run build

---

📝 License & Attribution

Distributed under the MIT License.
Satellite imagery & meteorological telemetry courtesy of ISRO (INSAT-3DR), IMD, and NOAA/NHC.