Table 1.

Problems of data quality, collection, treatment, and reproducibility

Category	Examples	Implications
Sum of inconsistently estimated population sizes.	Estimates for Odocoileus virginianus come from US wildlife state agencies using not transparently reported survey methods. Some states include both while-tailed and mule deers (e.g., Colorado). Canada population estimated by direct extrapolation from US numbers based on percent range area. Latin American population based on average density of 1 ind/km2. This figure yields a density estimate of 3.4 ind/km2 (5). Field estimates are 3× higher on average (6): M = 10.2 ind/km2; IQR = 2.177 to 20.5; N = 38.	Inconsistent estimates introduce biases in the population density calculation and model predictions. In this case, biomass is underestimated by −60% (2.7 vs. 8.2 Mt).
Population estimated from mean density, which divided by a different area leads to a different density estimate.	The RL estimates a total population size of ~104 K for the orangutan (Pongo pygmaeus) by multiplying a mean density of 0.67 ind/km2 by 155 K km2. Greenspoon et al. divided 104 K individuals by a habitat area of >82 K km2 obtaining an estimate of 1.26 ind/km2, almost twice the original density estimate.	Introduction of biases in the population density calculation and model predictions.
Density obtained by dividing the population size reported by the RL by area does not match the density reported in the same RL report.	The RL reports a total population of 323 to 955 for the Cozumel raccoon (Procyon pygmaeus), the midrange is 639 but Greenspoon et al. reported 465. The latter was divided by the habitat area resulting in 1.48 ind/km2. However, the RL reports that the species’ population density ranges between 12 and 112 ind/km2.	The mismatch suggests the approach taken is flawed and can lead to important deviations from known density estimates, which are propagated through the models and predictions.
Dividing population size by habitat area results in unrealistic population density estimates.	The mountain pigmy possum (Burramys parvus) is estimated to have a density of 2.69 ind/km2, field density estimates are substantially higher: M = 890 ind/km2, IQR = 510 to 1,335; N = 11 (6). The Tibetan antelope (Pantholops hodgsonii) is estimated to have a density of 1,437 ind/km2, field density estimates are M = 4.3 in/km2; IQR = 1.7 to 6.8; N = 11 (6).	Severe under or overestimation of mammal biomass.
Inconsistent use of RL reported populations, sometimes taken from the description, sometimes from the reported total, sometimes neither.	12 M individuals for the Crab eater seal (Lobodon carcinophaga) in the SI Appendix, 10 M in Table 2, RL reports 4 to 8 M; 27.5 K for the Blue whale (Balaenoptera musculus), RL reports 10 to 25 K including the smaller pygmy blue whale and juveniles; 200 K for the Minke whale (Balaenoptera acutorostrata), RL reports 187 K; 35 K for the Bowhead whale (Balaena mysticetus), RL reports >25 K. 54 K for Hypogeomys antimena, RL reports 5,036 mature individuals.	Impossible to reproduce the results because of the inconsistent use of RL estimates.
Inconsistent inclusion of juveniles in total populations with use of adult body mass.	Methods do not report the juvenile/adult ratio used for marine mammals nor account for differences in body mass between juveniles and adults. Juveniles were added to the RL estimates of adults for Fin whale (Balaenoptera physalus), Sperm whale (Physeter macrocephalus), and Blue whale (B. musculus), but not to the other top 10 marine mammals calculations, even though RL clearly reports adult-only populations.	Excluding juveniles leads to biomass under-estimation, whereas using adult body mass for juveniles leads to over-estimation
Underestimation of uncertainty.	Uncertainty in the RL is expert-based and not meant to represent the real uncertainty around the total estimate. Population size of mature individuals of the African buffalo (Syncerus caffer) between 398 K and 401 K (<1% error). However, the RL clearly states that while some regional estimates are available through aerial counts, the total estimate is highly uncertain since estimates from many regions are absent or biased (7).	Uncertainty of the total population size based on many independent estimates is expected to be much higher than in individual estimates. This approach leads to severe underestimation of uncertainty of the global biomass prediction.

M = Median, IQR = Interquartile range.