Communicating the uncertainty in estimated greenhouse gas emissions from agriculture

Abstract

In an effort to mitigate anthropogenic effects on the global climate system, industrialised countries are required to quantify and report, for various economic sectors, the annual emissions of greenhouse gases from their several sources and the absorption of the same in different sinks. These estimates are uncertain, and this uncertainty must be communicated effectively, if government bodies, research scientists or members of the public are to draw sound conclusions. Our interest is in communicating the uncertainty in estimates of greenhouse gas emissions from agriculture to those who might directly use the results from the inventory. We tested six methods of communication. These were: a verbal scale using the IPCC calibrated phrases such as 'likely' and 'very unlikely'; probabilities that emissions are within a defined range of values; confidence intervals for the expected value; histograms; box plots; and shaded arrays that depict the probability density of the uncertain quantity. In a formal trial we used these methods to communicate uncertainty about four specific inferences about greenhouse gas emissions in the UK. Sixty four individuals who use results from the greenhouse gas inventory professionally participated in the trial, and we tested how effectively the uncertainty about these inferences was communicated by means of a questionnaire. Our results showed differences in the efficacy of the methods of communication, and interactions with the nature of the target audience. We found that, although the verbal scale was thought to be a good method of communication it did not convey enough information and was open to misinterpretation. Shaded arrays were similarly criticised for being open to misinterpretation, but proved to give the best impression of uncertainty when participants were asked to interpret results from the greenhouse gas inventory. Box plots were most favoured by our participants largely because they were particularly favoured by those who worked in research or had a stronger mathematical background. We propose a combination of methods should be used to convey uncertainty in emissions and that this combination should be tailored to the professional group.

Keywords: Communicating statistics; Communication methods; Greenhouse gas emissions; Uncertainty.

Fig. 1

Scenario A: Comparing emissions of nitrous oxide from different countries. The estimated emissions of N₂0 from agriculture in 2010 presented with (a) histograms, with the means shown by the solid black lines and the 95% confidence interval shown by the solid grey lines, (b) shaded arrays, where the intensity of colour indicates the density of the underlying PDF, and (c) boxplots, where the green lines show the median values, the black boxes depicting the lower and upper quartiles and the dotted lines show the extent of the 95% confidence intervals. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

Fig. 2

Scenario B: Comparing emissions of methane to a given threshold value. Estimated emissions of CH₄ from agriculture in 2010 compared with an arbitrarily chosen threshold value of 130 kt CH₄ year⁻¹, presented with (a) histograms, (b) shaded arrays, where the intensity of colour indicates the density of the underlying PDF, and (c) boxplots, where the green lines show the median values, the black boxes depicting the lower and upper quartiles and the solid grey lines show the extent of the 95% confidence intervals. In each case the threshold value is marked by the solid red line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

Fig. 3

Scenario C: Assessing changes in methane emissions over time. The trend in emissions of CH₄ from agriculture between 1990 and 2010, presented with (a) histograms, (b) shaded arrays, where the intensity of colour indicates the density of the underlying PDF, and (c) boxplots, where the green lines show the median values, the black boxes depicting the lower and upper quartiles and the dotted lines show the extent of the 95% confidence intervals. In each case the zero line is marked by the solid red line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

Fig. 4

Scenario D: Assessing mitigation methods for the best opportunity to reduce emissions of nitrous oxide. The reduction in emissions of N₂0 from English grasslands resulting from a mitigation strategy, presented with (a) histograms, (b) shaded arrays, where the intensity of colour indicates the density of the underlying PDF, and (c) boxplots, where the green lines show the median values, the black boxes depicting the lower and upper quartiles and the dotted lines show the extent of the 95% confidence intervals. The solid red line indicates no reduction in emissions. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).

Fig. 5

The percentage of participants from each professional and mathematical group.

Fig. 6

Bar charts showing how participants responded to Question 1 for each of the four scenarios.

Fig. 7

Bar charts showing how participants responded to Question 2 for each of the four scenarios.

Fig. 8

Bar charts showing how participants responded to Questions 3 and 4.

Fig. 9

The black lines show the ranges of values that participants mapped the calibrated phrases to. The red line shows the range that the IPCC define for each phrase.

Fig. 10

The trend in emissions of CH₄ from agriculture in Wales between 1990 and 2010 shown using (a) a boxplot with the expected value (mean), median, 2.5th and 97.5th percentiles annotated on the graph and (b) a shaded array where the intensity of colour indicates the frequency of each observation with darker shading indicating a larger probability of observing that value. The expected value is −0.13 with 95% confidence interval given by [−0.34, 0.08]. The red lines mark the zero line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.).