Welcome to Statistics Seminar! Professor Jarno Vanhatalo (Department of Mathematics and Statistics, Faculty of Science Organismal and Evolutionary Biology Research Programme, Faculty of Biological and Environmental Sciences, University of Helsinki) will address briefly the bayesian integrated data models for baltic ringed seal dynamics and fish distribution.
Welcome to seminar on Friday, 13th December at 14.15. The lecture will take place at Mattilanniemi Campus in lecture hall MaA210.
Anyone interested is warmly welcome!
Content of the statistics seminar
Bayesian inference provides compelling framework to integrate multiple, complementary, data sets into one analysis. However, the challenge in
applying this theory is how to derive such models in practice so that all data are sensibly linked to the model parameters. In this talk, I
present three examples from our ongoing work in the area and discuss the lessons learned.
My first example is an integrated population model (IPM) for the Baltic ringed seals, which integrates multiple data sets collected between 1988 and 2023. In addition to analyzing the past and future development of the Baltic ringed seal population, we studied the relative importance of alternative data sets to the final inference on essential model parameters. Our results also provided support to the hypothesis that a greater proportion of seals haul-out on ice under lower ice cover circumstances -- a result that would not have been attainable with any single data set alone.
The two other examples are integrated species distribution models (ISDMs). For the Baltic sea spawning whitefish we integrated local scale scientific survey data on fish larvae, and areal data from commercial fisheries, informing about a spawning stock size. We analyzed environmental drivers of distribution of whitefish reproduction areas, spawner density, and maximum proliferation rate along the Finnish coast of the Gulf of Bothnia. Our results show that all these processes behind whitefish reproduction success are strongly dependent on local environmental conditions, having implications for climate change predictions.
In the second ISDM, we integrated scientific survey data on pike perch larvae and expert elicited larvae distribution maps collected from local fishermen. Since surveys are expensive, we aimed to improve larvae distribution predictions with inexpensive expert information. However, since expert knowledge is inherently subjective and prone to biases, our model includes components to calibrate experts' assessments and to measure their reliability. The expert information improved species distribution predictions compared to predictions conditioned on survey data only. However, experts' reliability also varied considerably, and even generally reliable experts had spatially structured biases in their assessments.
The works presented are collaborative efforts with Murat Ersalman, Mervi Kunnasranta, Markus Ahola, Anja M. Carlsson, Sara Persson, Britt-Marie Bäcklin, Inari Helle, Linnea Cervin (the first example), Ilaria Pia, Lari Veneranta, Elina Numminen (the second example), Karel Kaurila, Sanna Kuningas, Antti Lappalainen (the third example).
