pyinsyde 0.1.0

A faithful, dependency-light Python port of the INSYDE synthetic flood depth-damage model (Dottori et al. 2016).

pyinsyde — a Python port of the INSYDE flood-damage model

pyinsyde is a faithful, dependency-light Python port of INSYDE, the synthetic (component-based) flood depth-damage model for residential buildings of Dottori et al. (2016, NHESS 16, 2577–2591). It reproduces the original R model (ComputeDamage) component by component — in both deterministic and Monte-Carlo modes — and ships the model's original reference (EUR) cost data.

The only runtime dependencies are numpy and scipy.

Port of the reference R implementation by Dottori, Figueiredo, Martina, Molinari and Scorzini (GPL-3.0). See PORTING.md for fidelity notes and the known R quirks that are reproduced on purpose.

---

What INSYDE does

INSYDE estimates flood damage to a residential building by summing physical damage to individual components (structure, finishes, windows/doors, systems, clean-up/dehumidification, …) as a function of the flood hazard (water depth, velocity, duration, sediment load, water quality) and the building's exposure (geometry, materials, finishing level). Damage is returned in absolute terms and as a relative damage ratio (0→1) against the building's replacement value, and can be produced as an expected value or as a Monte-Carlo distribution.

---

Install

pip install -e .            # or: uv pip install -e .

For development (tests + linter):

uv venv --python 3.12
uv pip install -e ".[dev]"
uv run pytest -q

Quick start

import numpy as np
from pyinsyde import (ExposureVariables, HazardVariables,
                      compute_damage, load_unit_prices, replacement_value_eur)

exp = ExposureVariables()                 # INSYDE reference building
up = load_unit_prices()                   # original EUR unit prices
rv = replacement_value_eur(exp.BS, exp.BT)

hz = HazardVariables(he=np.arange(0.0, 3.0 + 1e-9, 0.5))   # depth vector
res = compute_damage(exp, hz, up, rv, uncert=0)            # deterministic
print(res.rel_damage)     # relative damage ratio at each depth

A complete runnable script — the Python equivalent of the R example — is in examples/run_insyde.py:

uv run python examples/run_insyde.py

---

Regional localization (bring your own prices)

INSYDE's damage logic is region-agnostic; only the cost inputs are local. The model ships with its original Italian (EUR) reference data, but you can run it for any country without modifying the package:

from pyinsyde import compute_damage, load_unit_prices, ExposureVariables, HazardVariables

up = load_unit_prices(path="prices/my_region.txt")   # your unit-price table
rv = 2000.0                                           # your replacement value (per m²)
res = compute_damage(ExposureVariables(), HazardVariables(), up, rv, uncert=0)

Your price table uses the same name value #unit format as pyinsyde/data/unit_prices.txt. Supply your region's replacement value directly to compute_damage.

Only commit openly-licensed data. Keep commercial cost tables out of the repository and load them at runtime via path=. Contributions of open regional datasets are very welcome — see the roadmap (multi-region support).

---

Repository layout

pyinsyde/
  variables.py   ExposureVariables / HazardVariables  (mirror INSYDE inputs)
  model.py       compute_damage — faithful port of insyde_function.R
  prices.py      unit-price + replacement-value loaders (original EUR data)
  data/          unit_prices.txt, replacement_values.txt  (INSYDE EUR reference)
examples/
  run_insyde.py  the R INSYDE example, in Python (deterministic + Monte-Carlo)
tests/           model validation tests
r_reference/     the original INSYDE R sources, for line-by-line comparison
PORTING.md       fidelity notes and reproduced R quirks
DATA.md          provenance of the bundled reference data

---

License / attribution

Ported from INSYDE (Dottori, Figueiredo, Martina, Molinari, Scorzini), released under the GNU GPL-3.0. This port is likewise GPL-3.0 — see LICENSE. The bundled unit_prices.txt / replacement_values.txt are the model's original reference data; see DATA.md.