Designing an optimal sampling strategy for a national level invasive alien plant assessment: A South African case study

• Published in: Ecological indicators 2020-12, Vol.119, p.106763, Article 106763
• Main Authors: Kotzé, Johann D.F., Beukes, Hein B., Seifert, Thomas
• Format: Article
• Language: English
• Subjects: Invasive trees; Monitoring; Sampling design; Survey; Survey costs; Validation
• ISSN: 1470-160X; 1872-7034
• DOI: 10.1016/j.ecolind.2020.106763

•Scientific rigorous monitoring system of invasive alien plants for South Africa.•Integration of sampling design with total survey costs and accuracy assessment.•Invasive plant species distribution pattern was decisive for efficiency of IAP sampling.•Cost effective monitoring of invasive alien plant indicators in support of management. The availability of accurate spatial distribution data of invasive alien plant (IAP) species is a shortcoming in most programmes and strategies that deal with biological invasions. Such data are required at the relevant spatial and temporal scales to assist decision making ranging from high level policy to local level implementation and the assessment of biological invasion status as part of the monitoring of the range and abundance of invasive species’ indicators. Because the complete inventories of IAPs are not feasible due to budget constraints, sampling of a representative proportion is necessary. The selection of an adequate sampling design is critical to ensure cost effective assessments that will also result in accurate and reliable IAP distribution data across extensive and environmental varying landscapes. This paper presents the development of an objective and scientifically accurate monitoring system at a regional level for major invaders for an extensive area such as South Africa by comparing three standard statistical sampling strategies, namely, simple random, systematic and proportionally stratified sampling. The comparison is based on the prediction accuracy of the three designs and their sensitivity to sample size. Sampling was repeated in simulations for each of the sampling designs based on an incremental increase in sampling intensity and compared to an actual field survey result. A generalised linear model was used to predict how accurately the distribution and density of species were reflected by the different sampling designs. Comparisons show that the proportionally stratified design provides the most precise results, followed by systematic and finally random sampling. Further sampling simulations showed that a proportionally stratified design at a national level requires approximately 61 000 sample points to be able to detect IAP species’ distribution and distinguish between IAP density levels of five percent and more with a maximum coefficient of variation of 15% at a specified cost threshold of R10 million (600000 €).