Open scientific modeling workflow

AutoCF

An ecosystem for Streamlining Compound Flood Simulation

AutoCF provides a streamlined, user-friendly, automated framework for compound flood simulation, supporting model setup, forcing preparation, simulation, validation, and impact analysis while letting users retain full control over the hydrodynamic solver.

Download AutoCF How to cite
Live audience
Where visitors come fromLoading country-level visits…
Unique visitors by countryNo raw IP addresses stored
Latest stable release
AutoCF v1.0.0Latest
—Total downloads
—Countries
1Publications
—Unique visitors
Not just a tool, but an entire compound flood simulation ecosystem.
Get AutoCF

WindowsLinux | Docker

v1.0.0Download

Windows desktop package with a managed Linux runtime through WSL2 or Docker.

macOSDocker

v1.0.0Download

Apple Silicon desktop package using Docker Desktop for its managed runtime.

HPCGPU | CPU

v1.0.0Download

Portable Apptainer release for shared Linux clusters. Nothing is installed on the host.

Global reach

AutoCF around the world

Workflow

What AutoCF handles

Forcing, streamed Wind, pressure, rain, discharge and water level pulled and clipped to your domain
Study area One GeoJSON polygon and the grid, bathymetry and land cover are built around it
Boundaries Tide, surge and river conditions assembled and written onto the model edges
Orchestration Build, simulate and post-process chained end to end, picking up where a job stopped
Flexibility Solver, resolution and every workflow step stay yours, set in one YAML file
Impact attribution Which driver caused the flooding, and what it exposed: people, buildings, roads
The AutoCF agent

Meet Riva

Riva: the bank of a river, the edge where the water meets the land.

I live in the dock at the bottom right of AutoCF, beside Jobs and log, resting and blinking while you work, waving when you greet me, thoughtful while I read. Click me and I open up.

I exist for the questions that interrupt a study: what a solver parameter really does, why a run came out the way it did, what the evaluation is actually telling you. So I look things up before I speak, searching the solver manual, the HydroMT API docs you have installed, and the AutoCF workflow guidance. For evaluation and attribution I take exact values straight from your model's JSON and CSV. I never ask a language model to do your arithmetic.

The numbers are yours; the explaining is mine. All of it stays on your computer, and I explain rather than act: I will not edit your YAML or launch your simulations.

How to cite

Use the tabs below to cite AutoCF and the supporting modeling tools used in your work.

If you use AutoCF in any part of your work—including model preparation, simulation, evaluation, or impact attribution—please cite:

Radfar, S., Maghsoodifar, F., Lin, N., & Moftakhari, H. (2026). AutoCF: An Automated LLM-Assisted Ecosystem for Compound Flood Simulation, Evaluation, and Impact Attribution. arXiv:2609.35753. https://doi.org/10.48550/arXiv.2609.35753
Download BibTeX Read the preprint
Selected case studies

Model Results

Maximum-depth flood maps and compound-flood attribution results produced through the AutoCF workflow. Open any figure to view it at full resolution.

Modeled maximum water depth for the Hurricane Harvey case study
Flood mapHurricane Harvey · 2017Maximum modeled water depth across the study domain, with observation locations.
Modeled maximum water depth for the Hurricane Helene case study
Flood mapHurricane Helene · 2024Maximum modeled water depth across the study domain, with observation locations.
Hurricane Sandy flood impact metrics by depth threshold
Impact analysisHurricane Sandy · Impact summaryFlooded area, exposed buildings, and exposed population across modeled depth thresholds.
Shapley attribution of flooded area to coastal, precipitation, and river drivers
Driver attributionHurricane Sandy · Flood driversShapley attribution of flooded area to coastal, precipitation, and river forcing.
Hurricane Harvey flooded area by depth threshold for combined and individual flood drivers
Impact analysisHurricane Harvey · Exposure-depth curveFlooded area by depth threshold for combined and individual coastal, precipitation, and river forcing.
Hurricane Harvey spatial Shapley drivers, forcing contributions, and compound amplification
Driver attributionHurricane Harvey · Spatial driversDominant Shapley driver, forcing contributions, and compound amplification across the study area.
The team

Team

Portrait of Dr. Soheil Radfar
Name
Dr. Soheil Radfar
Role
Lead Software Developer; Hydroinformatics Software Engineer
Affiliation
Associate Research Scholar, Princeton University
Website
sradfar.github.io
Portrait of Faezeh Maghsoodifar
Name
Faezeh Maghsoodifar
Role
Scientific Software Architect; Model Integration & User Experience Engineer
Affiliation
Ph.D. Candidate, The University of Alabama
Website
faezehmaghsoodifar.com

Contact the AutoCF team

We welcome inquiries, suggestions, comments, and collaboration opportunities.

autocf.project@gmail.com