Earth-2 X-Band SAR & InSAR Mission Twin

Operational Tropospheric Phase Screen Compensation, Flood Backscatter Attribution, and Autonomous OGC STAC v1.0.0 Tasking on Verified NOAA GFS $0.25^\circ$ Analyses

Author: Zhane Daye — SAR & Earth System Engineering | Edmonton, AB, Canada
📧 zhanedaye@umattr.ca · 🔗 linkedin.com/in/zhane-umattr · 💻 github.com/zrt219/earth2studio · 🌐 Live 3D WebGL Mission Twin (earth2-sar-insar.vercel.app)

---

🛰️ Executive Summary & Commercial SAR Alignment

Commercial X-band ($f_c = 9.65\text{ GHz}$, $\lambda = 3.1067\text{ cm}$) Synthetic Aperture Radar (SAR) constellations—specifically ICEYE and Synspective (StriX)—deliver sub-meter to 3-meter all-weather Earth observation for rapid flood response and infrastructure deformation monitoring. However, X-band Interferometric SAR (InSAR) is acutely sensitive to lower-tropospheric water vapor and pressure gradients: a $1.5533\text{ cm}$ ($\lambda/2$) two-way slant-range path delay induces a full $2\pi$ phase wrap, frequently masking millimeter-scale surface subsidence or pre-landslide slope creep beneath $9\text{–}15$ atmospheric fringes.

Built on top of NVIDIA Earth2Studio and an audited, leakage-free 3-stage atmospheric verification pipeline (Forecast $\to$ Freeze $\to$ Verify) operating on authentic NOAA GFS $0.25^\circ$ global analyses ($721 \times 1440$), this engineering showcase delivers an end-to-end Weather-to-SAR Tip-and-Cue & InSAR Atmospheric Phase Screen (APS) Mission Twin:

Target Domain / Organization	Engineered Subsystem	Physical & Algorithmic Implementation	Verified Quantitative Result
CATALYST (PCI Geomatics / InSAR)	sar_geoint_engine.py sarTasking.js	Two-way X-band ($9.65\text{ GHz}$) tropospheric path delay integration ($\text{ZHD} + \text{ZWD}$) with $8\times$ bicubic Nyquist-resolved mesoscale upsampling ($0.03125^\circ \approx 3.4\text{ km}$) and hydrostatic scale filtering ($\sigma = 1.8$ coarse cells) for DEM-coupled APS removal.	Italy / Alps: 98.73% phase RMSE reduction ($14.7763 \to 0.1874\text{ rad}$), 12.20 fringes removed. Japan / Honshu: 98.81% phase RMSE reduction ($13.8243 \to 0.1641\text{ rad}$), 9.24 fringes removed.
ICEYE (X-Band SAR Constellation)	sar_geoint_engine.py sarTasking.js	ICEYE-X38 ascending pass ($i = 97.7^\circ$ SSO, $\theta_{\text{inc}} = 28.4^\circ$, $0.5\text{ m}$ Spotlight VV) physical backscatter forward model ($\sigma^0_{\text{VV}}$) for specular open-water flood segmentation and Bragg capillary wind roughening.	Delineates $735.27\text{ km}^2$ ($1,380$ mesoscale pixels) of Po Valley riverine flood inundation below $-16.5\text{ dB}$ ($\sigma^0_{\text{VV,min}} = -31.77\text{ dB}$ vs. $-7.71\text{ dB}$ land mean).
Synspective (StriX Constellation)	sar_geoint_engine.py sarTasking.js	StriX-3 descending pass ($i = 97.6^\circ$ SSO, $\theta_{\text{inc}} = 34.1^\circ$, $3.0\text{ m}$ StripMap VV/VH) repeat-pass Coherent Change Detection ($\gamma_{\text{CCD}}$) coupling wind/moisture volume decorrelation with steep topographic slope geometry.	Isolates $550.73\text{ km}^2$ ($5,294$ mesoscale pixels) of high-risk Honshu mountain slope decorrelation ($\gamma_{\text{CCD}} < 0.32$, $\gamma_{\text{min}} = 0.080$, land mean $\bar{\gamma} = 0.786$).
Autonomous Ground Segment (Tip-and-Cue)	sar_geoint_engine.py uiController.js	Automated anomaly-to-tasking translator exporting OGC STAC v1.0.0 FeatureCollection GeoJSON with sar:* extensions (sar:frequency_band="X", sar:center_frequency=9.65, sar:product_type="SLC") and 3D orbital look-cone frustums.	2 valid STAC v1.0.0 tasking orders (ICEYE-X38, StriX-3) sealed with SHA-256 digest 3fecc248...0d47035; 1-click browser GeoJSON export.

---

📊 Visual Diagnostic Showcase

4-Panel X-Band InSAR APS Compensation, ICEYE Flood Backscatter & StriX Coherence Suite

Figure 1: (A) Po Valley & Alpine Foothills raw wrapped X-band interferogram ($\lambda = 3.11\text{ cm}$, $\theta_{\text{inc}} = 28.4^\circ$) exhibiting $12.20$ atmospheric fringes ($\text{RMSE} = 14.78\text{ rad}$) that obscure surface deformation. (B) CATALYST NWP-assisted APS-compensated interferogram reducing phase RMSE by 98.73% to $0.1874\text{ rad}$ and cleanly isolating the Po Valley subsidence bowl. (C) ICEYE-X38 Spotlight VV backscatter ($\sigma^0_{\text{VV}}$) delineating $735.3\text{ km}^2$ of specular open-water flood inundation ($< -16.5\text{ dB}$, cyan contour). (D) Synspective StriX-3 repeat-pass coherence ($\gamma_{\text{CCD}}$) over Honshu during Typhoon Megi, isolating $550.7\text{ km}^2$ of severe slope decorrelation ($\gamma < 0.32$, crimson contour).

Interactive 3D Spherical Mission Twin (R = 5.000 → 5.420)	Multi-Model AI-NWP Prognostic Benchmark

---

🔬 Scientific Provenance & Validation Boundaries (Real Observations vs. Forward-Modeled Physics)

In strict adherence to .agents/rules/e2s-014-scientific-integrity.mdc, this repository maintains a transparent, cryptographically audited boundary between real operational satellite/meteorological analyses, deterministic physical forward operators, and simulated benchmark components:

1. Validated Real Operational Meteorological Analyses (outputs/scientific_study_verified/):
   • Global 2-meter temperature ($T_{2m}$) initial (initial_condition.npy) and +24h verifying (verifying_truth_t2m.npy) states are authentic NOAA Global Forecast System (GFS) $0.25^\circ$ operational gridded analyses ($721 \times 1440$, $1,038,240$ grid cells).
   • Verified through a decoupled 3-stage Forecast $\to$ Freeze (SHA-256) $\to$ Verify pipeline ($\text{RMSE}_w = 2.1290\text{ K}$, $\text{MAE}_w = 1.0369\text{ K}$). Historical synthetic target-leakage scripts ($\hat{y} = x_0 + 0.72\Delta + \epsilon$) are permanently quarantined under od_experiments/scientific_study/historical/ and guarded by AST/runtime tests (test_anti_leakage_guardrails.py).
2. Deterministic Forward-Modeled SAR & InSAR Physics Operators (od_experiments/geoint_sar/sar_geoint_engine.py):
   • Tropospheric Delays ($\text{ZHD}$, $\text{ZWD}$, $\Delta\phi_{\text{APS}}$): Forward-modeled directly from the real NOAA GFS $T_{2m}$ and 24h tendency $\Delta T_{2m}$ tensors using Saastamoinen hydrostatic and Askne–Nordius wet refractivity integration coupled with $8\times$ bicubic spline refinement ($0.03125^\circ \approx 3.4\text{ km}$) to satisfy the X-band spatial Nyquist criterion ($|\nabla\phi| &lt; \pi\text{ rad/pixel}$).
   • SAR Backscatter ($\sigma^0_{\text{VV}}$) & InSAR Coherence ($\gamma_{\text{CCD}}$): Parameterized physical sensor forward models driven by GFS boundary-layer forcing, topographic slope $|\nabla h|$, and hydrological accumulation zones to demonstrate Tip-and-Cue mission planning and 3D Digital Twin visualization (not raw Level-1B radar echoes).
3. Validation Limitations & Host Hardware Boundaries:
   • Local execution was performed on a Windows 11 workstation with AMD Ryzen / Radeon hardware (no local NVIDIA CUDA tensor cores). Local neural checkpoint rollout requires CUDA; authenticated cloud inference was independently verified against the live NVIDIA Earth-2 Hosted FourCastNet API (od_experiments/earth2_001/api/fourcastnet_hosted.py).

---

📚 Engineering Reports & Research Portfolio (docs/engineering_reports/)

• 📄 Earth-2 X-Band SAR & InSAR Mission Twin — Standalone Engineering Report (PDF) (LaTeX Source)
• 📄 Executive Research Summary: Overhaul, Verification & 7-Pillar Expansion (PDF) (LaTeX Source)
• 📄 Comprehensive Scientific Study v3.1: CRPS, SSR, EDI/SEDI & Multi-Model Ensembles (PDF) (LaTeX Source)
• 📄 34-Figure Visual Scientific Atlas & SHA-256 Compendium (PDF) (LaTeX Source)
• 🗺️ OGC STAC v1.0.0 Autonomous SAR Tasking Orders (stac_tasking_orders.geojson) (Manifest JSON)

---

⚡ Quick-Start Reproduction & Verification

# 1. Run the complete 7-Pillar + X-Band SAR/InSAR APS & STAC Tasking Pipeline
python -m od_experiments.run_all_advanced_capabilities

# 2. Run the Python Scientific & SAR Physics Test Suite (233+ tests)
pytest od_experiments/tests/ -v

# 3. Run the 3D WebGL Digital Twin E2E & SAR/STAC Test Suites (333 assertions)
node viewer_app/test/geoint_sar_e2e.test.js
node viewer_app/test/run_e2e_tests.js

# 4. Launch or build the 3D WebGL Mission Twin locally
npm --prefix viewer_app install
npm --prefix viewer_app run dev
npm --prefix viewer_app run build

---

Upstream Framework: NVIDIA Earth2Studio

Earth2Studio is a Python-based package designed to get users up and running with AI Earth system models fast. Our mission is to enable everyone to build, research and explore AI driven weather and climate science.

- Earth2Studio Documentation -

Install | User-Guide | Examples | API

Quick start

Running AI weather prediction can be done with just a few lines of code.

• For detailed installation steps, including model-specific installations, see the install guide.
• See the examples gallery providing different inference workflow samples.
• Swap out data sources or models depending on your use case!

Tutorial

Agent-assisted setup

Automate setup with your preferred coding agent using NVIDIA Earth2Studio skills. Install the Earth2Studio skill set, then ask your favorite agent (Claude, Codex, OpenCode, etc) to recommend a model, configure an environment, or run a first deterministic forecast. Find more Earth2Studio skills in the NVIDIA Skills catalog.

npx skills add NVIDIA/skills --skill earth2studio-install
npx skills add NVIDIA/skills --skill earth2studio-discover
npx skills add NVIDIA/skills --skill earth2studio-data-fetch
npx skills add NVIDIA/skills --skill earth2studio-deterministic-forecast

Example agent prompts:

Use the Earth2Studio discover skill to recommend a starter forecast workflow.
Use the Earth2Studio install skill to set up my environment for FourCastNet3 inference.
Create a script to fetch ERA5 surface winds data for March 2024.
Create a deterministic forecast workflow with GFS, FourCastNet3, and a Zarr output store.

NVIDIA FourCastNet3

from earth2studio.models.px import FCN3
from earth2studio.data import GFS
from earth2studio.io import ZarrBackend
from earth2studio.run import deterministic as run

model = FCN3.load_model(FCN3.load_default_package())
data = GFS()
io = ZarrBackend("outputs/fcn3_forecast.zarr")
run(["2025-01-01T00:00:00"], 10, model, data, io)

ECMWF AIFS

from earth2studio.models.px import AIFS
from earth2studio.data import IFS
from earth2studio.io import ZarrBackend
from earth2studio.run import deterministic as run

model = AIFS.load_model(AIFS.load_default_package())
data = IFS()
io = ZarrBackend("outputs/aifs_forecast.zarr")
run(["2025-01-01T00:00:00"], 10, model, data, io)

Google Graphcast

from earth2studio.models.px import GraphCastOperational
from earth2studio.data import GFS
from earth2studio.io import ZarrBackend
from earth2studio.run import deterministic as run

package = GraphCastOperational.load_default_package()
model = GraphCastOperational.load_model(package)
data = GFS()
io = ZarrBackend("outputs/graphcast_operational_forecast.zarr")
run(["2025-01-01T00:00:00"], 4, model, data, io)

Important

Earth2Studio is an interface to third‑party models, checkpoints, and datasets. Licenses for these assets are owned by their providers. Ensure you have the rights to download, use, and (if applicable) redistribute each model and dataset. Links to the original license and source are often provided in the API docs for each model/data source.

Latest News

• WeatherNext 2 Cyclones, operational and Mini prognostic model wrappers.
• CAMulator, NSF NCAR CAM6 climate emulator with prescribed forcing data and conservation fixers.
• FuXi-S2S, global daily prognostic model for subseasonal-to-seasonal forecasting.
• CorrDiff ERA5-HRRR, generative downscaling from ERA5 to HRRR over the contiguous United States.

For a complete list of latest features and improvements see the changelog.

Overview

Earth2Studio is an AI inference pipeline toolkit focused on weather and climate applications that is designed to ride on top of different AI frameworks, model architectures, data sources and SciML tooling while providing a unified API.

The composability of the different core components in Earth2Studio easily allows the development and deployment of increasingly complex pipelines that may chain multiple data sources, AI models and other modules together.

The unified ecosystem of Earth2Studio provides users the opportunity to rapidly swap out components for alternatives. In addition to the largest model zoo of weather/climate AI models, Earth2Studio is packed with useful functionality such as optimized data access to cloud data stores, statistical operations and more to accelerate your pipelines.

Earth-2 Open Models

Access state of the art Nvidia open models for climate and weather: Earth-2 Open Models. For training recipes for these models, see the PhysicsNeMo repository.

Contributors

Check out the contributing document for details about the technical requirements and the user guide for higher level philosophy, structure, and design.

License

Earth2Studio is provided under the Apache License 2.0, refer to the LICENSE file for full license text.