We develop a framework to estimate reference greenhouse gas (GHG) emission levels for French public research units, grounded in unit-specific characteristics such as staff composition, scientific field, geographical location, and supervisory institutions. These reference values offer a more robust and context-sensitive alternative to generic national benchmarks. As such, they are intended to support research units in setting mitigation targets they deem fair and achievable.
This application is based on a study which leverages the GES 1point5 national database, covering approximately 700 or about one-third of French public research units, to analyze GHG emissions from research activities across five main sources: purchases, travels, commutes, electricity and heating, from 2019 to 2022. We develop a framework to identify robust statistical models capable of estimating median emissions and probability distributions for each emission source, based on simple structural characteristics of research units (e.g., staff composition, scientific field). These median values serve as reference GHG emission levels that research units can use to set mitigation targets and strategies.
The present methodology can be found here [ArXiV]. To cite the present work use [reference].
If you conducted your GHG assessment using GES1point5, you can access this information by logging into your personal space
For research domains, you can refer to the Hceres sub-domain classification
Below is an illustration of the output graph you will obtain once all the required information has been provided. The graph displays a comparison between the emissions of your research unit and those of reference research units, calculated with our models. Emissions are normalized per full-time equivalent (FTE), where university and assistant professors are counted as 0.5 FTE to account for the share of their activities devoted to teaching. The per capita emissions of your research unit are shown as red points, while the median emissions estimated by selected models (i.e., reference emissions) are shown as black points. For purchase-related emissions, the graph displays confidence and prediction intervals. For all other emission sources (travel, commuting, electricity, and heating), it shows prediction intervals only. This distinction is due to the fact that the reference emissions for these sources are estimated using Random Forest models, which do not provide confidence intervals.
In this panel, you can compare your research unit’s emissions with those of other research units in the GES1point5 database.
By default, the graph shows research units from the same scientific domain as your own, based on the Hceres scientific domain classification . For instance, if more than 50% of your staff belong to the Human and Social Sciences (HSS), the graph will display emissions from other HSS-dominated research units. If no single domain among HSS, Sciences and Technology (ST) and Life and Health Sciences (LHS) represents a majority, all research units in the database will be displayed. You can change this setting.
The graph displays only the assessments completed in the same year as your selected assessment. If you selected a year before the COVID-19 pandemic, data from 2019 will be shown. If you selected a year after the pandemic, data from 2022 will be shown. If you selected a year during the pandemic, please specify whether you meant 2020 or 2021.
The red points represent your emissions whereas the boxplots represent the distribution of emissions of research units in the database. The black horizontal line indicates the median value; the bottom and top edges of the box represent the first and third quartiles, respectively. The whiskers extend up to 1.5 times the interquartile range, and the dots represent research units whose emissions fall outside this range (outliers).
'wc' in 'travelswc' means 'with contrails'
This graph plots your emissions (red points) and the predicted emissions of the benchmark research unit (black points), as well as the 95% prediction (blue) and confidence (black) intervals. Calculations are performed each time you modify an input parameter. Since they may take some time, please wait for the graph to fully load without transparency. Confidence intervals are not available for travel, commuting, electricity and heating emission source, as their reference per capita GHG emissions have been obtained through Random Forest, which is a non-parametric method. This method doesn't rely on an explicit formula or assume a specific error distribution (like the normal distribution in linear regression), so there is no direct theoretical basis for computing traditional confidence intervals.
Download PNGIf the red point is above the black point, this means that your unit's per capita emissions exceed the modeled average for comparable units. Conversely, if the red point is below the black point, it indicates that your unit's per capita emissions are lower than the modeled average for comparable units.
The table below shows the percentage difference between the GHG emissions of your research unit and the mean values estimated by our models, for emission sources where your emissions exceed the average predicted levels.