Global estimates such as those we use in the present work are largely based on sampling from the distribution of biomass worldwide and then extrapolating for areas in which samples are missing. The sampling of biomass in each location can be based on direct biomass measurements or conversion to biomass from other types of measurement, such as number of individuals and their characteristic weight. Some of the main sources of uncertainty for the estimates we present are the result of using such geographical extrapolations and conversion from number of individuals to overall biomass. The certainty of the estimate is linked to the amount of sampling on which the estimate is based. Notable locations in which sampling is scarce are the deep ocean (usually deeper than 200 m) and deep layers of soil (usually deeper than 1 m). For some organisms, such as annelids and marine protists and arthropods, most estimates neglect these environments, thus underestimating the actual biomass. Sampling can be biased toward places that have high abundance and diversity of wildlife. Relying on data with such sampling bias can cause overestimation of the actual biomass of a taxon.
Another source of uncertainty comes from conversion to biomass. Conversion from counts of individuals to biomass is based on either known average weights per individual (e.g., 50 kg of wet weight for a human, which averages over adults and children, or 10 mg of dry weight for a “characteristic” earthworm) or empirical allometric equations that are organism-specific, such as conversion from animal length to biomass. When using such conversion methods, there is a risk of introducing biases and noise into the final estimate. Nevertheless, there is often no way around using such conversions. As such, we must be aware that the data may contain such biases.
In addition to describing the procedures leading to the estimate of each taxon, we quantitatively survey the main sources of uncertainty associated with each estimate and calculate an uncertainty range for each of our biomass estimates. We choose to report uncertainties as representing, to the best of our ability given the many constraints, what is equivalent to a 95% confidence interval for the estimate of the mean. Uncertainties reported in our analysis are multiplicative (fold change from the mean) and not additive (± change of the estimate). We chose to use multiplicative uncertainty as it is more robust to large fluctuations in estimates, and because it is in accord with the way we generate our best estimates, which is usually by using a geometric mean of different independent estimates. Our uncertainty projections are focused on the main kingdoms of life: plants, bacteria, archaea, fungi, protists, and animals.
The general framework for constructing our uncertainties (described in detail for each taxon in theSI Appendix and in the online notebooks) takes into account both intrastudy uncertainty and interstudy uncertainty. Intrastudy uncertainty refers to uncertainty estimates reported within a specific study, whereas interstudy uncertainty refers to variation in estimates of a certain quantity between different papers. In many cases, we use several independent methodologies to estimate the same quantity. In these cases, we can also use the variation between estimates from each methodology as a measure of the uncertainty of our final estimate. We refer to this type of uncertainty as intermethod uncertainty. The way we usually calculate uncertainties is by taking the logarithm of the values reported either within studies or from different studies. Taking the logarithm moves the values to log-space, where the SE is calculated (by dividing the SD by the square root of the number of values). We then multiply the SE by a factor of 1.96 (which would give the 95% confidence interval if the transformed data were normally distributed). Finally, we exponentiate the result to get the multiplicative factor in linear space that represents the confidence interval (akin to a 95% confidence interval if the data were log-normally distributed).
Most of our estimates are constructed by combining several different estimates (e.g., combining total number of individuals and characteristic carbon content of a single organism). In these cases, we use intrastudy, interstudy, or intermethod variation associated with each parameter that is used to derive the final estimate and propagate these uncertainties to the final estimate of biomass. The uncertainty analysis for each specific biomass estimate incorporates different components of this general scheme, depending on the amount of information that is available, as detailed on a case-by-case basis in theSI Appendix .
In cases where information is ample, the procedure described above yields several different uncertainty estimates for each parameter that we use to derive the final estimate (e.g., intrastudy uncertainty, interstudy uncertainty). We integrate these different uncertainties, usually by taking the highest value as the best projection of uncertainty. In some cases, for example, when information is scarce or some sources of uncertainty are hard to quantify, we base our estimates on the uncertainty in analogous taxa and consultation with relevant experts. We tend to round up our uncertainty projections when data are especially limited.
Another source of uncertainty comes from conversion to biomass. Conversion from counts of individuals to biomass is based on either known average weights per individual (e.g., 50 kg of wet weight for a human, which averages over adults and children, or 10 mg of dry weight for a “characteristic” earthworm) or empirical allometric equations that are organism-specific, such as conversion from animal length to biomass. When using such conversion methods, there is a risk of introducing biases and noise into the final estimate. Nevertheless, there is often no way around using such conversions. As such, we must be aware that the data may contain such biases.
In addition to describing the procedures leading to the estimate of each taxon, we quantitatively survey the main sources of uncertainty associated with each estimate and calculate an uncertainty range for each of our biomass estimates. We choose to report uncertainties as representing, to the best of our ability given the many constraints, what is equivalent to a 95% confidence interval for the estimate of the mean. Uncertainties reported in our analysis are multiplicative (fold change from the mean) and not additive (± change of the estimate). We chose to use multiplicative uncertainty as it is more robust to large fluctuations in estimates, and because it is in accord with the way we generate our best estimates, which is usually by using a geometric mean of different independent estimates. Our uncertainty projections are focused on the main kingdoms of life: plants, bacteria, archaea, fungi, protists, and animals.
The general framework for constructing our uncertainties (described in detail for each taxon in the
Most of our estimates are constructed by combining several different estimates (e.g., combining total number of individuals and characteristic carbon content of a single organism). In these cases, we use intrastudy, interstudy, or intermethod variation associated with each parameter that is used to derive the final estimate and propagate these uncertainties to the final estimate of biomass. The uncertainty analysis for each specific biomass estimate incorporates different components of this general scheme, depending on the amount of information that is available, as detailed on a case-by-case basis in the
In cases where information is ample, the procedure described above yields several different uncertainty estimates for each parameter that we use to derive the final estimate (e.g., intrastudy uncertainty, interstudy uncertainty). We integrate these different uncertainties, usually by taking the highest value as the best projection of uncertainty. In some cases, for example, when information is scarce or some sources of uncertainty are hard to quantify, we base our estimates on the uncertainty in analogous taxa and consultation with relevant experts. We tend to round up our uncertainty projections when data are especially limited.