PESTpy

A pure-Python parameter estimation and uncertainty analysis package for environmental and geophysical models. pestpy bridges the gap between scipy.optimize.least_squares and PEST++ by providing two core algorithms:

• GLM (Gauss-Levenberg-Marquardt) – A deterministic optimizer with an explicit multi-lambda testing loop and Marquardt scaling. Suitable for non-linear least-squares calibration problems.
• iES (Iterative Ensemble Smoother) – A stochastic ensemble-based smoother based on the mathematics from Fienen et al. (2025). It simultaneously upgrades an ensemble of parameter realizations while adding observation noise to prevent ensemble collapse.

Installation

pip install pestpysuite

Or directly from source:

pip install .

Quick Start

import numpy as np
from pestpy import GLMOptimizer, IESOptimizer

# --- GLM example ---
def model(params):
    return params[0] * np.array([1.0, 2.0, 3.0]) + params[1]

obs = np.array([2.1, 4.0, 6.1])
p0 = np.array([1.0, 0.0])

glm = GLMOptimizer(model, obs)
result = glm.fit(p0)
print("GLM solution:", result)

# --- iES example ---
n_params, n_obs, n_real = 2, 3, 50
ensemble = np.random.randn(n_params, n_real)

ies = IESOptimizer(model, obs)
ensemble_updated = ies.fit(ensemble)
print("iES posterior mean:", ensemble_updated.mean(axis=1))