This page will be used to document past community of practice meetings.
CANCELLED: Community member Rob Cameron (Dalhousie University) will visit us!
Exploring Ecological Questions with SDMs: 3 examples using lichens
SDMs are not only useful for predicting species distribution but also for answering important ecological questions. We will look at 3 examples of lichen studies that use SDMs to explore ecological patterns and processes. The first example uses SDMs to look at niche differentiation in 4 sympatric species of lichen. The second example looks at assessing threats to conservation of 2 lichens and third example looks at predicting lichen rich ecosystems.
Jordy Thomson from DFO (Fisheries and Oceans Canada) walked us through the eelgrass SDM his team developed using the biomod2 package for R! He described the environmental data layers, presence-absence observations and model context.
On April 14th, Maggie MacPherson from the Boreal Avian Modelling Project discussed how to predict bird distributions under global change.
Kellina Higgins helped us understand how to design surveys for effective models.
- You can find her presentation here.
Field data to predictions with statistics in between: linking Eastern waterfan observations to landscape attributes
Ecological field data can come in many forms: systematic experimental designs with set variables recorded (ecological studies), opportunistic observations (citizen science), surveys to detect species (botanical surveys) and more. Each form of data is valuable in its own way yet has its own limits. Sometimes data recorded for one purpose can be used in another context. Here, data collected with the objective of detecting populations of Eastern Waterfan (Peltigera hydrothyria) was used in predictive modelling to estimate the probability of its occurrence and its density across the landscape in and around Fundy National Park. Methods: Field data was collected by the Atlantic Canada Conservation Data Centre (ACCDC) and Parks Canada supplemented by other occurrences contributed to ACCDC. In addition to observations of populations, GPS tracks representing surveys were used to estimate absence records. Landscape metrics derived from remote-sensing and photo-interpreted forest inventory data were used in the analysis to determine key habitat attributes: stream size, elevation, slope, stream aspect, canopy height, stand composition, stand age, crown closure, distance to roads, and distance to harvest cut-blocks. Conditional inference trees, logistic regression and multiple regression were used to analyze the data with the statistical software R in order to predict its presence and abundance. Results: Waterfan was found in medium-size streams far from roads and harvest cut-blocks at higher elevations. Its presence was also influenced by the stand composition. Larger colonies were found in steeper sections of streams with northern and eastern flow orientations. The probability of water occurrence and its abundance was inferred based on these landscape attributes to identify potential waterfan hotpots in other streams. Discussion: The field notes recording variables such as dominant canopy species, stream substrate, stream speed and percent canopy cover could not be included in any of the analyses due to inconsistencies in data collection between field crews and large data gaps. In addition, biases may have been introduced in the density estimates given that the methodology differed between field crews and by size of colonies. Thus, field data collection could be improved to adopt a common methodology to estimate population sizes and to record field attributes in order to better understand habitat requirements and develop more accurate predictive models.
The Julie Lee-Yaw lab visited us to talk about the limitations of SDMs and their work testing the impacts of modeling decisions on predictions and accuracy.
Meeting theme: building pseudo SAR habitat models.
Skills Building Shop: Lionel Leston from the University of Alberta demonstrated boosted regression trees for Canada Warbler and/or Wood Thrush in Nova Scotia in R.
- You can find the workshop materials here. Ensure the Google Drive folder structure is replicated on your machine.
- Optional: download the raw data and large model outputs for Canada Warbler. Once downloaded, move the data into the correct locations in the folder structure created above (put data in 0_data/0_raw and outputs in 2_BRT_outputs)
September 2022: CANCELLED
Westwood Lab workshop: an approach to ‘grading’ data quality and quantity for environmental covariates for modelling
Riley and Caitlin have asked shared a link where you can add any additional thoughts about the method and the workshop today: here
Remember to fill out the two question CoP Winter & Spring exit survey.
Meeting theme: spread awareness of existing best practices in peer discussion groups
We did not have time for the analytical working group task:
- Continue community building through several short talks and networking
Establish working groups. See working group agenda here.
- Data Sharing working group meeting notes
Meeting theme: Connecting people
Meeting theme: Establishing community of practice priorities