SWATGenXSWATGenX
ResearchDocs
Open Explorer Work with us

Watershed Modeling Software for U.S. SWAT+ Models

Screen any U.S. watershed in your browser, then download a ready-to-run SWAT+ model of it.

SWATGenX is watershed modeling software for U.S. SWAT+ (SWAT Plus). Pick a USGS station, HUC12 outlet, or HUC8 basin and it builds the package for you — terrain, climate, soils, land cover, routing and SWAT+ project files from NHDPlus HR and national datasets. Calibrate in the cloud, with validation on years the optimizer never saw.

Open the Watershed Explorer Download example modelsGenerate a SWAT+ model
SWATGenX Watershed Explorer: USGS streamflow stations across the contiguous United States, coloured by current flood status
25,000+
USGS stations
live streamflow context
30M+
NHDPlus HR features (USGS)
high-resolution routing
SWAT+
Editor-ready packages
downloadable ZIP
50
Example models
sign in to download and inspect
NHDPlus HRPRISM climategSSURGO soilsNLCD land coverNSRDB solarSWAT+ Editor-readyParallel SWAT+ enginePython · REST · MCP

What SWATGenX does

Step 1 · Screen

Screen watersheds

Find the right basin before building.

Browse USGS stations, HUC8 and HUC12 boundaries, stream networks, watershed indicators, and live streamflow context in the Watershed Explorer.

Step 2 · Build

Generate SWAT+ models

Build an Editor-ready ZIP for any U.S. basin.

Create downloadable SWAT+ project packages from NHDPlus HR routing, national climate, soils, land use, and terrain—delivered for QSWAT+ and SWAT+ Editor workflows.

Step 3 · Report

Report and calibrate

Document the basin and prepare calibration.

Produce PDF watershed briefings, and cloud calibration and validation outputs when the account runs them.

SWATGenX is a web-based watershed modeling software platform for the continental United States. It turns national geospatial datasets and NHDPlus HR hydrography into calibrated SWAT+ models, and, in a groundwater study, a steady-state MODFLOW 6 groundwater model beside the SWAT+ model (in Michigan Lower Peninsula, Indiana, Minnesota, Ohio, Wisconsin) — an end-to-end research-and-development environment for engineers and scientists who need trustworthy, reproducible watershed models across many U.S. basins without building the data and calibration pipeline themselves. Map-based screening and PDF reporting are the on-ramp; the deliverable is a defensible model.

Workflow

Select a basin → screen in Explorer → generate a report (optional) → build a SWAT+ package → open in Editor for calibration and scenarios.

Select
USGS gage, catalog HUC12 outlet, whole HUC8 basin, or — by drilling into a HUC12 and hand-picking catchments — a 30 m catchment-set site model, all in the Watershed Explorer.
Screen
Map layers, stream networks, live flow context, and watershed indicators before you commit.
Report
Generate a PDF watershed briefing for stations and HUC8 (when your access level allows).
Build
Receive a downloadable SWAT+ ZIP wired for QSWAT+ / SWAT+ Editor with NHDPlus HR delineation — every build runs a 10-year diagnostic and a SWAT+ Check review.
Calibrate & validate
Cloud calibration: particle-swarm optimization (MMPSO by default) against USGS gauges, scored on daily and monthly NSE over its calibration years and again over validation years the calibration did not use.
Example modelsHow it works

Calibration & validation, built in

A model you cannot evaluate is a liability. Every SWAT+ model SWATGenX builds runs a 10-year diagnostic simulation and an automated SWAT+ Check water-balance review, so you see how the uncalibrated model behaves at every gauge before spending anything on calibration. From there, cloud calibration runs particle-swarm optimization (MMPSO by default) against USGS observations on elastic compute — scored on daily and monthly NSE — and then re-runs the calibrated parameters over a validation window of years the calibration did not use.

This is not a private benchmark: our calibration results are published on the calibration examples page, 43 basins calibrated coast to coast, 25 of them validated on years their calibration did not use, with the calibration and validation scores side by side.

Cloud calibration runs your whole swarm at once on one AWS box, one core per model, with MMPSO (multi-memory particle swarm) as the default optimizer. Our own calibrations add a diagnosis-guided search, whose models each run on several cores with SWATGenX's parallel build of the SWAT+ engine; with routing serial it gives the same results on several threads as on one in our tests.

Who SWATGenX is built for

SWATGenX is not a replacement for every hydrologic or hydraulic model. It is built for teams that need calibrated SWAT+ models at scale, and, in a groundwater study, a steady-state MODFLOW 6 groundwater model beside the SWAT+ model (in Michigan Lower Peninsula, Indiana, Minnesota, Ohio, Wisconsin)—on national datasets and NHDPlus HR routing, assembled through QSWAT+ and SWAT+ Editor and calibrated in the cloud—without rebuilding the data, delineation, and calibration pipeline for each study area.

Next steps

Build a model of your watershed
Watershed modeling platform
How It Works

New to the category? Read the Watershed Software Guide. Outcomes: use cases. Screening scope: watershed screening tools.

Technical foundation

The hosted pipeline prepares NHDPlus HR channel connectivity, PRISM and NSRDB climate drivers, gSSURGO and NLCD land inputs, and a SWAT+ Editor–aligned project tree—documented end to end in the methodology hub.

Modeling engine: SWAT vs SWAT+

SWAT (the Soil & Water Assessment Tool) runs as command-line executables with text-file input and output.

SWAT+ is a completely restructured redesign of SWAT in which the code organization becomes object-based and the input system becomes relational-based. The core process algorithms remain similar, but SWAT+ was built to address limitations that accumulated over decades of additions and modifications. It is not a patch; it is a major rewrite.

What changes in practice:

  • Relational toolchain — SWAT+ Editor stores all model inputs in a SQLite database with foreign-key relationships, so users do not have to manually edit raw text files. The executable still reads ASCII input files at runtime, but the modern toolchain (QSWAT+ + SWAT+ Editor) generates and manages them from the database.
  • Flexible spatial connections — SWAT+ allows explicit, user-defined routing between watershed objects (HRUs, landscape units, channels, reservoirs), rather than the fixed subbasin → channel hierarchy of SWAT classic.
  • Active extension trajectory — Recent peer-reviewed work demonstrates surface–subsurface coupling via gwflow and SWAT+MODFLOW-6, which matters for any analysis that includes groundwater interaction.

SWATGenX employs SWAT+ (assembled via QSWAT+ and SWAT+ Editor) as its modeling framework, and runs simulations on its open, accelerated build of the SWAT+ engine — multi-core capable (see the parallel engine for what its identity tests cover).

Hydrography backbone: NHDPlus V2 vs NHDPlus HR

The hydrography framework a platform "locks in" determines the size and count of routed stream segments, and therefore the representational detail of channels, catchments, and routing structure. This is often a stronger practical differentiator than the modeling engine alone.

AttributeNHDPlus V2 (medium resolution)NHDPlus HR (high resolution)
Scale1:100,000 (medium-resolution NHD)mostly 1:24,000 or finer (high-resolution NHD)
Elevation source30 m NED (2012 snapshots)10 m USGS 3DEP
Features (USGS OFR 2019-1096)about 2.6 millionmore than 30 million (over ten times as many)

NHDPlus V2 is mapped at 1:100,000. NHDPlus HR increases the flowline count by roughly an order of magnitude, enabling modeling at local scale while nesting into national context.

Finer hydrography means more data, more computation and more parameters to identify. SWATGenX automates the preprocessing, calibration, and model-assembly pipeline so that the cost of working with high-resolution hydrography is absorbed by the platform, not by the user.

For the very largest NHDPlus HR basins, teams may still prefer staged calibration passes or more constrained delineation strategies (including TauDEM-oriented alternatives where appropriate) to keep runs operationally manageable.

Below the land surface: groundwater & PFAS

SWATGenX also builds below the land surface. In a groundwater study, a steady-state MODFLOW 6 groundwater model is built with the SWAT+ model, in Michigan Lower Peninsula, Indiana, Minnesota, Ohio, Wisconsin: its recharge comes from the SWAT+ model's annual percolation, its aquifer geometry and hydraulic properties from well lithology logs at build time, and its heads are calibrated automatically to measured well heads.

The data under those models is a capability in its own right. The national groundwater inventory — 9.5 million well records in 48 states, with 33.4 million lithology depth intervals — is harvested exclusively from state agency records, with no synthetic gap-filling: what you query is what an agency logged. That standard is what makes the derived aquifer models testable.

Our research adds PFAS fate-and-transport modeling: MODFLOW 6 solute transport driven by SWAT+ recharge, demonstrated at contaminated-site scale, with per-soil PFAS initialization drawn from a national soil PFAS inventory and observations from the 98,733+-record monitoring inventory.

National data inputs

All SWATGenX analyses draw from publicly available, nationally consistent datasets — no proprietary data is required. Documented vintages and sources ensure reproducibility.

  • NHDPlus HR — high-resolution stream network, catchments, and routing (mostly 1:24,000 or finer, 10 m 3DEP)
  • USGS NWIS — real-time and historical streamflow for 25,000+ stations
  • PRISM — gridded daily precipitation and temperature (4 km; our builds use 2000–2024)
  • NSRDB — solar radiation, humidity, wind (~4 km; gridMET wind when NSRDB's is implausibly low)
  • gSSURGO — gridded soil properties (250 m)
  • NLCD / USDA CDL — land cover and crop classification
  • USGS 3DEP — elevation, used through the NHDPlus HR elevation raster at 30 m

One platform, three doors: web, Python, AI agents

Everything on this page is reachable without a browser. Guests — no account — can read the example-model catalog through the public API, from Python or any HTTP client; the AI-assistant connector needs a SWATGenX account. The national groundwater inventory (9.5 million well records in 48 states, with 33.4 million lithology depth intervals) and the PFAS monitoring inventory(98,733+ sampling records) are for verified accounts.

Web application

The Watershed Explorer, model dossiers, calibration wizard, and downloads — the full visual workflow described above.

Python

pip install swatgenx

import swatgenx as sg
sg.catalog(state="FL")

Read the example-model catalog as a guest; build and download models with an API key.

AI agents (MCP)

A public MCP server at swatgenx.com/mcp gives Claude, ChatGPT, and other AI assistants ten tools to search models, read calibration results, query groundwater and PFAS data, and start builds — with the access ladder explained in every response.

Data backbone (citations & resolutions):

  • 25,000+ USGS stream gauges; current readings, refreshed once a day, for the 7,000+ that report them (USGS marks recent data provisional)
  • NHDPlus HR — built mostly from 1:24,000-or-finer NHD, 10 m 3DEP, and WBD; the USGS NHDPlus HR User’s Guide (OFR 2019-1096) counts more than 30 million features against about 2.6 million in NHDPlus V2 (USGS NHDPlus HR)
  • PRISM gridded climate (~4 km cells; PRISM defines a “day” as 24h ending 12:00 GMT); NLCD land cover 30 m; gSSURGO soils (gridded SSURGO per NRCS); NSRDB solar/meteorology nominally ~4 km (≈0.038°), 30-minute series — SWATGenX may resample/regrid for operations; see methodology for lineage

Full methodology & provenance table →

Appropriate use

Flood screening compares each gauge's latest reading with flood flows from a log-Pearson Type III fit to its USGS annual peaks. Short gage records, regulation, land-use change, and non-stationarity increase uncertainty — treat classes as triage, not design certification.

Rainfall context compares recent 24h totals to PRISM-derived historical percentiles of rainy days at the station’s grid cell — not the same as NOAA Atlas design storms or MRMS-based hydrologic design. See Data & methodology for baseline years and caveats.

  • Not for regulatory floodplain mapping without a licensed engineering study
  • Not a replacement for official watches, warnings, or evacuation orders
  • Event-focused engineering hydrology (e.g. USACE HEC-HMS) remains the standard for many design contexts.

Use of the platform is subject to the Terms of Use and Privacy Policy.

Get started

Open the Watershed Explorer, pick a USGS station, catalog HUC12, or HUC8, and move from map context to automated briefings or a SWAT+ model build. Most builds finish within about an hour once they start; some take several hours. The same national stack runs end to end.

Model Creation API examples (Jupyter + Python): GitHub.