This project will develop a modeling framework for integrating professional and citizen-science data, leading to smarter surveillance and improved estimates of AIS distribution that account for imperfect detection and sampling biases.
Probabilistic, structured surveys performed by professionals provide high quality data for invasive species surveillance, but the high costs associated with these surveys limit their use. Thus, citizen scientists play a key and growing role in detecting and monitoring invasive species in Minnesota and elsewhere. For example, volunteers participating in the MAISRC/Extension program, Starry Trek, have discovered 3 of the 13 known starry stonewort populations in Minnesota, and AIS Detectors have performed thousands of hours of surveillance for AIS. Yet, estimating the spatial distribution of invasive species from citizen-science data alone is challenging, because the probability of detecting an invasive species depends on both its abundance and on how sampling effort is allocated—typically in a manner that is spatially heterogeneous and biased towards areas that are easy-to-access (see visual). Further, species are not always detected when present (“imperfect detection”). In these cases, it is difficult to estimate occurrence risk, i.e., the probability a target species is truly absent versus present but not yet observed.
To address these challenges, and thus maximize the value of data collected by AIS professionals and citizen scientists, the research team will:
- Develop an integrated modeling framework that leverages the complementary strengths of data collected opportunistically by citizen scientists (widespread coverage in space and time) and structured survey data collected by AIS professionals (known sampling effort, reduced sampling biases);
- Apply this framework to two high-priority aquatic invasive plant species, Eurasian watermilfoil and starry stonewort. These species differ in their statewide abundance, stages of invasion, biology, and detectability. This framework will lead to improved estimates of the distributions of these species and improved understanding of the factors that influence their occurrence and detectability;
- Explore ways to improve AIS surveillance more generally (e.g., by strategically allocating sampling effort spatially, temporally, and between structured and unstructured survey efforts).
Progress
The research team has developed: 1) a landscape-level model for the distribution and spread of Eurasian watermilfoil (EWM) in Minnesota lakes; and 2) a within-lake distribution model for starry stonewort (SSW). Their work during 2022 primarily focused on the latter model. They finalized the modeling framework, including the choice of predictors to include in the different components of the model (the component used to describe the distribution of SSW and the component used to model detection given SSW presence). In addition, the research team evaluated the sensitivity of model results to various data limitations and modeling assumptions. Lastly, they submitted a paper describing the model and results to the journal Biological Invasions.
The probability of SSW occurrence was higher near boat accesses and in areas with high local access density, highlighting the role of boater movement in spreading starry stonewort to new lakes. Fetch and water depth were also important predictors of within-lake occurrence probabilities. The probability of detecting SSW with a single rake throw, given SSW was present, was around 32%, with detection probabilities varying by time of year and decreasing with greater density of other plant species detected via the rake sample.
The results can be used to identify when and where to sample for SSW. The research team recommends that aquatic invasive surveyors search for starry stonewort preferentially in areas of shallow depth and with closer proximity to and higher density of accesses. They also recommend focusing these searches in late summer/early fall, when detection probabilities are highest due to starry stonewort’s phenology.