KML Earthworks: Turn a Google Earth sketch into cut/fill volumes in seconds

Show the Community!
1

Hey community! :construction_worker:

I’m a civil engineer and construction cost coordinator from Brazil.
I built this app to solve a real problem from my daily work —
and I want to share it here.

kml_estrada_perfil
kml_estrada_perfil1513×635 52.2 KB

The Problem

Early-stage access roads and haul roads are often sketched quickly
in Google Earth. Turning those alignments into earthworks estimates
usually means:

  • Manual coordinate extraction from KML
  • External elevation queries (one by one)
  • Repetitive spreadsheet formulas for cut/fill volumes
  • No mass balance diagram

This process takes hours. And it’s error-prone.

The App

KML Earthworks automates the entire pipeline:

  1. Upload .kml files (exported from Google Earth)
  2. App parses coordinates, builds chainage every 20m (Haversine)
  3. Queries terrain elevation via Open-Meteo API (OpenTopoData fallback)
  4. Fits a constrained grade line (SciPy optimization, max slope +
    max height limits)
  5. Computes cut/fill using prismatoid formula
    (5–15% more accurate than average-end-area)
  6. Applies shrink/swell factor, calculates borrow/waste per segment
  7. Generates interactive Plotly charts: plan view, terrain profile,
    cut/fill bars, mass diagram, 3D view
  8. Exports detailed Excel report

Zero GIS software required. Runs entirely in the browser.

:link: Live app: https://kml-earthworks.streamlit.app
:laptop: GitHub: GitHub - CAIOZANETTI/kml-earthworks: Automate terrain cut/fill estimation from KML access alignments, with interactive Plotly visualization and a Streamlit app. Includes optional elevation enrichment · GitHub

Tech Stack

  • UI: Streamlit
  • Charts: Plotly
  • Data: pandas + NumPy + SciPy
  • Elevation: Open-Meteo API + OpenTopoData fallback
  • KML parsing: Python stdlib (no GDAL needed)
  • Exports: openpyxl
  • Type safety: TypedDict for all data structures
  • Tests: 58 pytest tests covering all core modules (100% passing)

What I learned building this

The hardest part wasn’t the code — it was encoding engineering
judgment into logic.

For example: when should the grade line “cut” vs “fill”?
That depends on soil type, haul distance, shrink/swell factor,
and cost tradeoffs. A spreadsheet formula can’t hold that reasoning.
Python can.

This is my thesis: domain experts should build their own tools.
Streamlit made it possible for an engineer (not a software developer)
to go from problem to deployed app.

Happy to answer questions about the engineering logic, the
architecture decisions, or how I used AI assistance (Claude)
to accelerate development while keeping full technical
responsibility for every formula.

— Caio Zanetti | Civil Engineer & Builder

1 Like