Carbon footprint analysis

Wooden transportation pallets are used massively across the world. In Europe alone, several billion pallets are in circulation. What are their lifecycle emissions? What drives the emissions? What should a policy intervention target to reduce them?

The simulation dataset and a Jupyter notebook to replicate this case are available here.

Case description

The given data represents an estimation of the carbon footprint of wooden pallets, taking into account parameter variation for Finland¹. Total life cycle emissions $E_{\text{total}}$ are comprised of the emissions from different stages of the pallets’ life, taking into account the functional unit size (number of customer trips) and the number of uses of a single pallet $N$:

$$ E_{\text{total}} = \left(E_{\text{production}} + E_{\text{repair}} + E_{\text{end-of-life}}\right) \cdot \frac{F}{N} + E_{\text{transport}} \cdot F $$

For every stage, the amount of emissions is defined by the global warming potential of materials and resources, measured in kilograms of CO₂ equivalent per unit of the corresponding material or resource.

$$ E_{\text{production}} = M_{\text{timber}} \cdot P_{\text{timber}} + M_{\text{nails}} \cdot P_{\text{nails}} + C_{\text{wood}} \cdot P_{C_{\text{wood}}} + H \cdot P_H $$

$$ E_{\text{repair}} = 2 \left( M_{\text{timber}} \cdot P_{\text{timber}} + M_{\text{nails}} \cdot P_{\text{nails}} + C_{\text{wood}} \cdot P_{C_{\text{wood}}} \right) $$

$$ E_{\text{transport}} = \left(0.98 M_{\text{timber}} + M_{\text{nails}} \right) \cdot 2L \cdot P_{\text{truck}} \cdot T_{\text{truck}} $$

$$ E_{\text{end-of-life}} = P_{\text{end-of-life}} \cdot T_{\text{end-of-life}} $$

The nomenclature and numeric assumptions are detailed here:

The input-output dataset obtained with simple random sampling of size 1000 is provided for analysis.

Global sensitivity analysis

We employ variance-based Sobol’ indices implemented within the SimDec package². Although normally Sobol’ indices would require a much larger sample, the simple binning approach within SimDec allows sufficient precision for small samples generated with simple random sampling.

The resulting sensitivity indices highlight the leading pair of input variables: End-of-life and Number-of-uses. This pair is also characterized by a notable second-order effect of 0.22 (not shown in the table).

Sensitivity indices

Visualization

SimDec visualization in stacked histogram form is used to examine how the most influential inputs are mapped onto the distribution of the output values (see short SimDec guide). The automatic output generated by the SimDec dashboard, in this case, benefits from a few adjustments to finetune the visualization: increasing the number of bins to 170, and zooming in on the X-axis to [-5000, 10000]. Colors are edited just to enhance visual association with the corresponding scenarios.

The visualization uncovers an intriguing pattern. The distribution splits into two peaks, each tied to an end-of-life scenario: the right blue peak for landfilling and the left orange peak with partially negative emissions for incineration.

Here’s the twist:

• When pallets are landfilled, more uses naturally lower the carbon footprint (the same emissions spread over more lifecycles).
• Under incineration, the opposite holds true: the fewer times a pallet is used, the better its carbon footprint looks.

There is no mistake, either in the model or in the graph. This puzzling outcome arises from emissions accounting rules: burning wood as a substitute for fossil fuels is counted as producing negative emissions. The fewer times a pallet is reused, the sooner it is burned, and the earlier those negative emissions are recorded, improving the balance on paper.

What are policymakers incentivizing here? Perhaps an intervention to incentivize longer usage of pallets, regardless of their end-of-life, would be beneficial.

Importantly, this insight emerged through visualization. Sensitivity indices alone would not have been enough to guide policy design. At the same time, with eight uncertain variables in play, global sensitivity analysis was essential to focus the visualization on the core effects that matter most in the model.

Implementation

The input-output data can be either uploaded to the dashboard simdec.io or analyzed with the SimDec Python package, see 3_Carbon_footprint.ipynb.

Additional resources

The case is presented in this video and this publication³.