Sampling bias in reptile occurrence data for the Kruger National Park

Abstract

o effectively conserve and manage species, it is important to (1) understand how they are spatially distributed across the globe at both broad and fine spatial resolutions and (2) elucidate the determinants of these distributions. However, information pertaining to the distributions of many species remains poor as occurrence data are often scarce or collected with varying motivations, making the resulting patterns susceptible to sampling bias. Exacerbating an already limited quantity of occurrence data with an assortment of biases hinders their effectiveness for research, thus making it important to identify and understand the biases present within species occurrence data sets. We quantitatively assessed occurrence records of 126 reptile species occurring in the Kruger National Park (KNP), South Africa, to quantify the severity of sampling bias within this data set. We collated a data set of 7118 occurrence records from museum, literature and citizen science sources and analysed these at a biologically relevant spatial resolution of 1 km × 1 km. As a result of logistical challenges associated with sampling in KNP, approximately 92% of KNP is data deficient for reptile occurrences at the 1 km × 1 km resolution. Additionally, the spatial coverage of available occurrences varied at species and family levels, and the majority of occurrence records were strongly associated with publicly accessible human infrastructure. Furthermore, we found that sampled areas within KNP were not necessarily ecologically representative of KNP as a whole, suggesting that areas of unique environmental space remain to be sampled. Our findings highlight the need for substantially greater sampling effort for reptiles across KNP and emphasise the need to carefully consider the sampling biases within existing data should these be used for conservation management decision-making. Modelling species distributions could potentially serve as a short-term solution, but a concomitant increase in surveys across the park is needed.Conservation implications: The sampling biases present within KNP reptile occurrence data inhibit the inference of fine-scale species distributions within and across the park, which limits the usage of these data towards meaningfully informing conservation management decisions as applicable to reptile species in KNP.

The role of Indigenous and Community Conservation Areas in herpetofauna conservation: a preliminary list for Santa Cruz Tepetotutla, Oaxaca Mexico

“…The description and delimitation of a species’ geographic distribution is one of the main requirements to propose conservation measures; however, sampling bias is a common problem for reliable delimitation ( Barends et al 2020 ). In general, for rare species, many of which are under extinction risk, these biases often result in information gaps and underestimation of distribution ( Costa et al 2010 ; Fourcade et al 2014 ).…”

“…However, distributional knowledge of some endemic species in this region is still incomplete. This is mainly due to two phenomena: 1) a low number of herpetological studies in the area ( Caldwell 1974 ; Lips et al 2004 ; Delia et al 2013 ; Caviedes-Solis et al 2014; Furbush et al 2017 ) and 2) biased sampling restricted to easily accessible areas, such as along paved roads ( Guedes et al 2017 ; Barends et al 2020 ).…”

occAssess: An R package for assessing potential biases in species occurrence data

“…Within each period, assessRarityBias regresses the number of records for each species on its commonness. Each species’ deviation from the fitted regression - its residual - can be used as an index of how over- or under-sampled it is given its commonness (Barends et al, 2020; Speed et al, 2018). Extending this concept, the coefficient of variation ( r 2 ) of the fitted regression can be used a measure of proportionality between the number of times species have been recorded and their commonness.…”

“…If the PC scores are non-normal, it is simple to substitute the ellipses for the actual data points (see s1 for more details). For similar approaches see Pescott et al (2019b) and Barends et al (2020). Note that this assessment assumes that the spatial resolution of the environmental data are relevant to the responses of the target organism(s) at the spatial scale of the analysis desired.…”

“…There is a growing literature of studies which take species occurrence datasets and screen them for biases (Barends et al, 2020; Boakes et al, 2010; Meyer et al, 2016; Pescott et al, 2019a; Petersen et al, 2021; Ruete, 2015; Speed et al, 2018; Sumner et al, 2019; Troudet et al, 2018); elsewhere, various approaches to visualising the spatial and temporal coverage of occurrence records across large areas have been commonplace in national species atlases for some time (e.g. Preston et al, 2002).…”