Trends and biases in African large carnivore population assessments: identifying priorities and opportunities from a systematic review of two decades of research.

African large carnivores have undergone significant range and population declines over recent decades. Although conservation planning and the management of threatened species requires accurate assessments of population status and monitoring of trends, there is evidence that biodiversity monitoring may not be evenly distributed or occurring where most needed. Here, we provide the first systematic review of African large carnivore population assessments published over the last two decades (2000-2020), to investigate trends in research effort and identify knowledge gaps. We used generalised linear models (GLMs) and generalised linear mixed models (GLMMs) to identify taxonomic and geographical biases, and investigated biases associated with land use type and author nationality. Research effort was significantly biased towards lion (Panthera leo) and against striped hyaena (Hyaena hyaena), despite the latter being the species with the widest continental range. African wild dog (Lycaon pictus) also exhibited a negative bias in research attention, although this was partly explained by its relatively restricted distribution. The number of country assessments for a species was significantly positively associated with its geographic range in that country. Population assessments were biased towards southern and eastern Africa, particularly South Africa and Kenya. Northern, western, and central Africa were generally under-represented. Most studies were carried out in photographic tourism protected areas under government management, while non-protected and trophy hunting areas received less attention. Outside South Africa, almost half of studies (41%) did not include authors from the study country, suggesting that significant opportunities exist for capacity building in range states. Overall, large parts of Africa remain under-represented in the literature, and opportunities exist for further research on most species and in most countries. We develop recommendations for actions aimed at overcoming the identified biases and provide researchers, practitioners, and policymakers with priorities to help inform future research and monitoring agendas.

Optimising the cost of roadkill surveys based on an analysis of carcass persistence.

Reliable estimates of wildlife mortality due to wildlife-vehicle collisions are key to understanding its impact on wildlife populations and developing strategies to prevent or reduce collisions. Standardised approaches for monitoring roadkill are needed to derive robust and unbiased estimates of mortality that are comparable across different study systems and ecological contexts. When designing surveys, there is a trade-off between survey frequency (and hence logistical effort and financial cost) and carcass detection. In this regard, carcass persistence (the period a carcass remains detectable before being removed by decomposition or scavengers) is important; the longer a carcass persists, the greater the likelihood it will be detected with lower survey effort by conducting more infrequent surveys. Using multi-taxon carcass data collected over a month of repeated driven surveys, combined with five covariates (species functional group, body weight, carcass position on road, carcass condition [either flattened or not after impact], and rainfall prior to each survey), we explored the drivers of carcass persistence with the overall aim of providing information to optimise the design of carcass surveys along linear infrastructure. Our methodological approach included a survival analysis to determine carcass persistence, linear regressions to test the effect of covariates, a subsampling analysis (using field data and a simulation exercise) to assess how the proportion of carcasses detected changes according to survey frequency, and an analysis to compare the costs of surveys based on study duration, transect length and survey frequency. Mean overall carcass persistence was 2.7 days and was significantly correlated with position on road and within-functional group body weight. There was no evidence for a significant effect of rainfall, while the effect of carcass condition was weakly non-significant. The proportion of carcasses detected decreased sharply when survey intervals were longer than three days. However, we showed that survey costs can be reduced by up to 80% by conducting non-daily surveys. Expanding on the call for a standardised methodology for roadkill surveys, we propose that carcass persistence be explicitly considered during survey design. By carefully considering the objectives of the survey and characteristics of the focal taxa, researchers can substantially reduce logistical costs. In addition, we developed an R Shiny web app that can be used by practitioners to compare survey costs across a variety of survey characteristics. This web app will allow practitioners to easily assess the trade-off between carcass detection and logistical effort.