Search for: All records

A key problem in computational sustainability is to understand the distribution of species across landscapes over time. This question gives rise to challenging large-scale prediction problems since (i) hundreds of species have to be simultaneously modeled and (ii) the survey data are usually inflated with zeros due to the absence of species for a large number of sites. The problem of tackling both issues simultaneously, which we refer to as the zero-inflated multi-target regression problem, has not been addressed by previous methods in statistics and machine learning. In this paper, we propose a novel deep model for the zero-inflated multi-target regression problem. To this end, we first model the joint distribution of multiple response variables as a multivariate probit model and then couple the positive outcomes with a multivariate log-normal distribution. By penalizing the difference between the two distributions’ covariance matrices, a link between both distributions is established. The whole model is cast as an end-to-end learning framework and we provide an efficient learning algorithm for our model that can be fully implemented on GPUs. We show that our model outperforms the existing state-of-the-art baselines on two challenging real-world species distribution datasets concerning bird and fish populations. 

Abstract Dispersal drives diverse processes from population persistence to community dynamics. However, the amount of temporal variation in dispersal and its consequences for metapopulation dynamics is largely unknown for organisms with environmentally driven dispersal (e.g., many marine larvae, arthropods and plant seeds). Here, we used genetic parentage analysis to detect larval dispersal events in a common coral reef fish,Amphiprion clarkii, along 30 km of coastline consisting of 19 reef patches in Ormoc Bay, Leyte, Philippines. We quantified variation in the dispersal kernel across seven years (2012–2018) and monsoon seasons with 71 parentage assignments from 791 recruits and 1,729 adults. Connectivity patterns differed significantly among years and seasons in the scale and shape but not in the direction of dispersal. This interannual variation in dispersal kernels introduced positive temporal covariance among dispersal routes that theory predicts is likely to reduce stochastic metapopulation growth rates below the growth rates expected from only a single or a time‐averaged connectivity estimate. The extent of variation in mean dispersal distance observed here among years is comparable in magnitude to the differences across reef fish species. Considering dispersal variation will be an important avenue for further metapopulation and metacommunity research across diverse taxa. 

Abstract Determining metapopulation persistence requires understanding both demographic rates and patch connectivity. Persistence is well understood in theory but has proved challenging to test empirically for marine and other species with high connectivity that precludes classic colonisation–extinction dynamics. Here, we assessed persistence for a yellowtail anemonefish (Amphiprion clarkii) metapopulation using 7 years of annual sampling data along 30 km of coastline. We carefully accounted for uncertainty in demographic rates. Despite stable population abundances through time and sufficient production of surviving offspring for replacement, the pattern of connectivity made the metapopulation unlikely to persist in isolation and reliant on immigrants from outside habitat. To persist in isolation, the metapopulation would need higher fecundity or to retain essentially all recruits produced. This assessment of persistence in a marine metapopulation shows that stable abundance alone does not indicate persistence, emphasising the necessity of assessing both demographic and connectivity processes to understand metapopulation dynamics.