iterative-ensemble-smoother

What it is and what it does

iterative_ensemble_smoother is a Python library for data assimilation and history matching using ensemble-based methods. It focuses on inverse problems where you have many parameters (potentially millions) but relatively few ensemble realizations (hundreds). The library implements ESMDA (Ensemble Smoother with Multiple Data Assimilation), a non-iterative algorithm that performs multiple sequential data assimilation steps to update an ensemble of model realizations toward observations.

The package depends on numpy, scipy, scikit-learn, joblib, and networkx to handle the numerical computations and ensemble operations. It provides both correlation-based (AdaptiveESMDA) and distance-based (DistanceESMDA) localization methods to improve results in high-dimensional problems. The library is actively maintained and supports modern Python versions (3.12–3.14).

Use it for:

• Calibrate reservoir simulation models using production data in petroleum engineering workflows
• Update weather or climate model ensembles with observational data for forecasting
• Inverse problems in geophysics where you estimate subsurface properties from seismic or well data
• History matching in hydrogeology to constrain groundwater flow parameters
• Parameter estimation in computational models with expensive forward simulations

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

Implements ensemble-based data assimilation and history matching algorithms (ESMDA with correlation and distance-based localization) for inverse problems with many parameters and few realizations.

Yes, if you work on inverse problems or data assimilation in scientific computing. The package is actively maintained, has low install friction, and fills a specific niche in ensemble-based parameter estimation. The GPL-3.0 copyleft license is suitable for research and open-source projects but rules out proprietary closed-source use. The 'Development Status :: 1 - Planning' classifier warrants checking whether all advertised localization methods are production-ready.

Install

iterative-ensemble-smoother on PyPI

pip install iterative-ensemble-smoother

uv add iterative-ensemble-smoother

poetry add iterative-ensemble-smoother

Installing iterative-ensemble-smoother

Before you install

Low friction install with standard scientific Python dependencies (numpy, scipy, scikit-learn, joblib, networkx). Active maintenance with a release 9 days ago; repository shows recent commits and is not archived.

License in practice

GPL-3.0 copyleft license requires that derivative works and distributions remain under GPL-3.0 or compatible terms; suitable for research and open-source projects but incompatible with proprietary closed-source use.

Quickstart

pip install iterative_ensemble_smoother

from iterative_ensemble_smoother import ESMDA

# Initialize and run ESMDA on your ensemble and observations
esmda = ESMDA()
esmda.update(ensemble, observations, std)

Requires Python 3.12 or later (supports 3.12, 3.13, 3.14); numpy, scipy, scikit-learn, joblib, and networkx must be available.

Verify before relying

• Whether AdaptiveESMDA and DistanceESMDA are fully implemented or still in development given 'Development Status :: 1 - Planning' classifier
• Performance characteristics and scalability limits for the 'millions of parameters' use case mentioned in description
• Availability and completeness of examples referenced in the usage section

Package facts

Evidence: iterative_ensemble_smoother-1.3.0-py3-none-any.whl

Tags

More Scientific/Engineering packages