Getting Started#

This tutorial introduces the usage of MOCA for different user backgrounds. The three sections below are expandable and independent of each other. Choose the section that best matches your experience.

I have never worked with uncertainty or Monte Carlo before

Life Cycle Assessment is affected by many sources of uncertainty. This package helps to quantify and understand the uncertainty that is contained in the numerical data that make up your LCA model.

This data includes:

your foreground process data
the background database data you use, such as ecoinvent
characterisation factors used in the LCIA

The easiest way to get started on uncertainty analysis with Brightway2 is to use the Activity Browser. This tool provides a user-friendly interface to explore and define uncertainty in your LCA model, without needing to write code.

Using the Activity Browser, you can:

assign uncertainty distributions to exchanges
define parameters such as the mean, standard deviation or bounds
use pedigree matrices to derive uncertainty parameters

If you are completely new to uncertainty analysis, chances are high you do not have a lot of uncertainty data on hand. In this case, the Activity Browser’s pedigree matrix functionality can be a great starting point to define uncertainty based on your judgement of your model quality. You can rank your different exchanges between 1 and 5 on several categories and then get an estimation of the uncertainty.

Once you have defined uncertainty in the Activity Browser, you can directly use this information to run large Monte Carlo simulations using moca_uncertainty_lca. This allows you to transfer the uncertainty information that you have defined for your input parameters into uncertainty information for your LCA results. You can then analyse the resulting distribution of impact scores to see histograms, create boxplots and more.

To give you a quick overview of how Monte Carlo simulation works, here is a simplified workflow:

Randomly sample all uncertain inputs according to their distributions
Recalculate the LCA result
Repeat this process many times
Analyze the resulting distribution of impact scores

If you do not want to build up an own lca project, you can use a rudimental dummy project, which can be used to get familiar with the functionality of this package. To create the project go to the FAQ.

Now everything is set up, would you like to give it a try? Simply create a project in the Activity Browser and start defining uncertainty for your model. Then, you can run a simple Python script like this to get started with MOCA:

import brightway2 as bw
import moca_uncertainty_lca as moca

def main():

   bw.projects.set_current("your_project")  # change to your project

   demand = {
      bw.Database("your_database") # database of your demand activity
         .get("492e35473j74034b6c49b0240ff7800"): 1 # key of your demand activity
   }

   mc_lca = moca.MonteCarloLCA(
      demand=demand,
      lcia_method_name="EF v3.1 no LT", # change to your preferred LCIA method
      run_parallel=False
   )

   # for a quick test, use a small number of iterations
   # for a more thorough analysis, use 2000 or more iterations
   mc_lca.execute_monte_carlo(iterations=100)

   # these lines will save the results and statistics to JSON files
   mc_lca.results_to_json()
   mc_lca.stats_to_json()

if __name__ == "__main__":
   main()

Want more examples? Check out the Code Examples for more code snippets.

I have already used MonteCarloLCA in Brightway2

If you are familiar with the MonteCarloLCA class from Brightway2, getting started with MOCA likely requires only minor changes to your existing code. The main motivation for using MOCA instead of Brightway’s native Monte Carlo implementation is performance.

Compared to Brightway2, MOCA’s MonteCarloLCA implementation:

is significantly faster for large Monte Carlo runs
uses multiprocessing to parallelise across multiple CPU cores
is designed for large-scale uncertainty analysis

Let’s have a look at a simple example to see how the code differs between the two implementations.

1. Setup

Most likely, some part of your code looks similar to this:

import brightway2 as bw

bw.projects.set_current("your_project")  # change to your project

demand = {
   bw.Database("your_database") # database of your demand activity
      .get("492e35473j74034b6c49b0240ff7800"): 1 # key of your demand activity
}

lcia_methods = [
   ('EF v3.1 no LT', 'climate change no LT', 'global warming potential (GWP100) no LT'),
   ('EF v3.1 no LT', 'acidification no LT', 'accumulated exceedance (AE) no LT')
]

2. Brightway2’s MonteCarloLCA

You might then perform a Monte Carlo simulation using Brightway2’s MonteCarloLCA like this:

import brightway2 as bw
import json

# initialize the Monte Carlo LCA
mc_lca = bw.MonteCarloLCA(demand, lcia_methods)

# execute the Monte Carlo simulation
mc_results = []
for _ in range(100):
   mc_results.append(next(mc_lca))

# write the results to a JSON file
json.dump(mc_results, open("bw_mc_results.json", "w"), indent=4)

3. MOCA’s MonteCarloLCA

With MOCA, the same workflow would look like this:

import moca_uncertainty_lca as moca

def main():

   # initialize the Monte Carlo LCA
   mc_lca = moca.MonteCarloLCA(demand, lcia_methods=lcia_methods, run_parallel=False)

   # execute the Monte Carlo simulation
   mc_lca.execute_monte_carlo(iterations=100)

   # retrieve the results and write them to files
   mc_results = mc_lca.mc_results
   mc_lca.results_to_json(filename="moca_mc_results.json")
   mc_lca.stats_to_json(filename="moca_mc_stats.json")

if __name__ == "__main__":
   main()

However, MOCA doesn’t stop here - you can easily run a parallelised calculation by simply setting run_parallel=True when initializing the MonteCarloLCA class. This can significantly reduce the runtime for large Monte Carlo simulations. Additionally, you can set default uncertainty distributions for your model parameters or export a list of all your uncertain parameters.

Want to see more? Check out the Code Examples for more code snippets.