BORIS Theses

BORIS Theses
Bern Open Repository and Information System

Occupancy vs. detection: Estimating changes in epiphytic lichen communities over 20 years

von Hirschheydt, Gesa (2023). Occupancy vs. detection: Estimating changes in epiphytic lichen communities over 20 years. (Thesis). Universität Bern, Bern

23vonhirschheydt_g.pdf - Thesis
Available under License Creative Commons: Attribution (CC-BY 4.0).

Download (11MB) | Preview


The growing global human population is exerting increasing pressure on the natural environment. Habitat destruction and anthropogenic climate change are causing species to decline or to shift their distribution ranges, but some species cannot keep up with the unprecedented speed of these changes and go extinct. As a result, we are losing biodiversity at the pace of a mass extinction. Already now, this loss has entailed unwanted effects on human well-being by negatively affecting ecosystem services like food provisioning, climate regulation, or pest control. Increased political pressure has urged governments to take action towards the conservation of diversity of life on Earth. To be effective, however, actions aimed at the protection of species require the evaluation of the current status of the species and how the populations change over time. Like many others, the government of Switzerland uses national Red Lists to identify the most threatened species and to set priorities for conservation actions at the national scale. The data for these Red List assessments come from large-scale surveys or monitoring programs that were established for the purpose of observing and inferring changes over time. However, ecological surveys are subject to detection errors, i.e., failing to detect species where they occur. These errors can lead to biases in the estimation of species distributions, habitat associations, or population changes, potentially resulting in an inappropriate threat category and a misassignment of resources for conservation measures. It is the purpose of this thesis to obtain estimates of population change for epiphytic lichen species in Switzerland that are less affected by detection errors, using data collected within the scope of the national Red List assessment. To estimate detection errors, it was first necessary to test the applicability of the available statistical methods to the lichen data (Chapter 1). Given the scarcity of published literature on the subject of detection errors in lichens, it also made sense to investigate the extent and the causes of such errors in greater detail (Chapter 2). The insights from these investigations then allowed me to analyse the ecological patterns behind population changes of epiphytic lichens in Switzerland over the last 20 years (Chapter 3). In Chapter 1, I tested whether the structure of the lichen data was generally suitable for the type of statistical models that are most often used to account for detection errors. They are called occupancy models and they require data from sites that were surveyed multiple times over a short period. The model uses the differences and similarities between the observations of the repeated visits to estimate the detection probability. In the standardised lichen data from the national Red List survey, only a small subset of all sites was surveyed a second time, while the others were surveyed only once. To find out whether these single-visit sites could contribute information to parameter estimation in an occupancy model, I simulated data under different designs but with the same mixed structure as the lichen data, i.e., with some repeated-visit sites (with two or four visits) and many single-visit sites. I first fitted an occupancy model to only the repeated-visit portion of the data and extracted the precision of the parameter estimates. I then successively added more single-visit sites, reran the analysis, and checked whether the precision of the parameter estimates improved. Precision did improve with additional single-visit sites, both for the parameter occupancy and the parameter detection probability. This shows that single-visit sites contribute to parameter estimation, when they are combined with repeated-visit data, and that it is beneficial to include single-visit data in an occupancy analysis. When the number of repeated visits was raised from two to four, precision was not only generally better, but also the contribution of single-visit sites improved. This finding is of limited relevance for the analysis of the currently available lichen data, but it could be useful to make adjustments to the design in the future. In Chapter 2, I explored the magnitude of and variation in detection probability in the lichen data that were collected during the first Red List assessment (1995–2000). I included the conspicuousness and the taxonomic identifiability of the species as covariates to detection probability, supposing that conspicuous and easy-to-identify species may have a higher detectability. The experience of observers with individual species was also likely to affect detectability in a positive way. Average detection probability across all observers was estimated at 49%, with substantial differences between observers and species, some of which were due to people’s experience or to the conspicuousness or identifiability of the species. As observer experience changed over time, detectability was slightly higher towards the end of the sampling period than at the beginning. The result that detection success was estimated to be almost a fifty-fifty chance was rather surprising. The standardised circumstances would have suggested a higher detectability: the size of the sampled area was limited, survey time almost unlimited, and all observers had prior experience with lichen surveys. In contrast to animals, lichens cannot run away or hide, and while most plants and fungi exhibit seasonality in their morphology, lichens do not. It is therefore likely that such low detection probabilities ⎯ in other words, such high detection errors ⎯ occur in most datasets of sessile organisms. Ignoring them would lead to a severe understimation of frequencies of occurrence and area of occupancy. The variation between species and differences between observers in combination with a potential spatial clustering of observers is expected to result in a stronger bias for some species than for others, an effect that is difficult to assess without the explicit estimation of detection probability. In Chapter 3, I estimated how occupancy changed for 329 epiphytic lichen species in Switzerland between the first and the second national Red List assessment conducted over the periods 1995–2000 and 2018–2022. Although the model estimates occupancy at the species level, I took a more community-based approach in this chapter and grouped species into 18 ecological guilds. Three guilds described a preference for free-standing trees, humid forests, or old trees, two guilds represented specialized photobionts (trentepohlioid and cyano), and twelve guilds were derived from high and low ecological indicator values for temperature, precipitation, continentality, eutrophication, pH, and light availability. With this guild-based approach, I was able to find potentially meaningful correlations with environmental change in Switzerland over the same time scale. An ongoing decline in species associated with old trees suggests that the low abundance of such trees, though increasing, has not yet allowed specialist lichens to recover from the severe loss they experienced due to unsustainable forestry practices in the last century. A strong increase in species indicative of high pH and tolerant to eutrophication in combination with a decline in eutrophication-sensitive and acidophytic species suggests a continuing effect of environmental pollutants on lichen communities. While acid deposition decreased to a very low level over the last decades, critical levels for nitrogen deposition are still exceeded in two thirds of the country. Some guild changes could also potentially be attributed to climate change. Species of high temperatures and low precipitation tended to increase, whereas species with a preference for low temperatures or high precipitation tended to decline. If these simultaneous environmental changes were indeed the driving force behind the observed changes, they are likely to continue in the near future. In the three chapters, I have consequently shown that it was possible to use the mixed structure of the lichen data to obtain detection-corrected estimates of frequency of occurrence and population changes. I showed how large the detection error was despite many favourable circumstances, and how it can be accounted for in an ecological study. Limitations of this thesis include model assumptions that may not be entirely fulfilled, and the restrictions imposed by data scarcity on the number of covariates that could be included in the model. In the future, I see potential in combining the standardised data with the countless individual observations recorded by volunteers or in other projects. Including multiple sources in one integrated model could improve both accuracy and precision of estimates of population changes. At a larger scale, e.g., for standardised species distribution modelling for global Red List assessments, it would be valuable to find a set of readily available and reliable predictor variables to model lichen occurrences. It is important to keep in mind, however, that estimates of species frequency or population changes will not reduce the risk of extinction a species may be facing, however precise these estimates may be. Ultimately, conservation actions will be necessary to ensure the persistence of many species. Nevertheless, this thesis lays the foundation for a more accurate, data-based Red List assessment. As such, I hope it can direct conservation efforts to where they are most needed.

Item Type: Thesis
Dissertation Type: Cumulative
Date of Defense: 12 October 2023
Subjects: 500 Science > 570 Life sciences; biology
500 Science > 580 Plants (Botany)
Institute / Center: 08 Faculty of Science > Department of Biology > Institute of Plant Sciences (IPS)
Depositing User: Sarah Stalder
Date Deposited: 17 Jan 2024 10:16
Last Modified: 17 Jan 2024 10:16

Actions (login required)

View Item View Item