Fish modeling in complex environments is critical for understanding drivers of population dynamics in aquatic systems. This paper proposes a Bayesian network method for modeling fish survival and growth over multiple connected rivers. Traditional fish survival models capture the effect of multiple environmental drivers (e.g., stream temperature, stream flow) by adding different variables, which increases model complexity and results in very long and impractical run times (i.e., weeks). We propose a coupled survival-growth model that leverages the observations from both sources simultaneously. It also integrates the Bayesian process into the neural network model to efficiently capture complex variable relationships in the system while also conforming to known survival processes used in existing fish models. To further reduce the performance disparity of fish body length across cohorts, we propose two approaches for enforcing fairness by the adjustment of training priorities and data augmentation. The results based on a real-world fish dataset collected in Massachusetts, US demonstrate that the proposed method can greatly improve prediction accuracy in modeling survival and body length compared to independent models on survival and growth, and effectively reduce the performance disparity across cohorts. The fish growth and movement patterns discovered by the proposed model are also consistent with prior studies in the same region, while vastly reducing run times and memory requirements.

Multi-modal time series data in real applications often contain data of different dimensionalities, e.g., high-dimensional modality such as image data series, and low-dimensional univariate time series. Multi-modal time series data with missing high-dimensional modal values are ubiquitous in real-world classification and regression applications. To accurately predict the target labels, it is important to appropriately impute the high-dimensional modal missing values. However, most existing imputation methods focus on multivariate time series, fail to simultaneously consider temporal dependencies within each series and the correlations across the series, and also lack a probabilistic interpretation. In this paper, we propose a novel method, which uses a new structured variational approximation technique for the imputation of missing values in multi-modal time series. Instead of directly imputing high-dimensional modal missing values, we use the variational approximation technique to impute intermediate lower-dimensional feature representations of high-dimensional modal missing values from simple modalities related to high-dimensional modality and then feed them into a dynamical model. The dynamical model captures the temporal dependencies of the feature representations and finally predicts the target labels. In order to address the optimization difficulties caused by the lack of ground truth values of lower-dimensional feature representations, we also propose a two-stage isolated optimization strategy for better convergence. We evaluate our method on a real-world stream monitoring dataset. Our extensive experiments demonstrate that the proposed method outperforms several state-of-the-art methods in both data imputation and prediction performance.

Brook trout (Salvelinus fontinalis) is a freshwater fish species of ecological, economic, and cultural importance in eastern North America. Estimating the abundance, movement, and survival of brook trout in the wild is an important task for environmental management, and current methods often involve physical tagging or collection of DNA samples for each of the fish as their unique identifier. However, this process is expensive and inefficient. Meanwhile, although deep learning methods have proven effective for individual recognition of humans, it remains challenging to apply this system to wildlife biology due to fewer available images, different biometric patterns, and relatively poor image quality. In this paper, we develop a framework to automate the process of individual recognition of brook trout. Distinguished from simply adopting traditional feature descriptors (e.g., SIFT and HOG) or using deep neural networks on the raw images, our framework utilizes multiple modalities consisting of the region of interest and gray-scaled pigmentation patterns. We use these multiple modality features in a Convolutional Neural network to generate feature vectors as fish descriptors. These descriptors are then used to distinguish individual brook trout by ranking their relative distance in latent space. Our experimental framework demonstrates better results than baseline methods such as SIFT and HOG while being more robust to distortions characteristic of large imagery datasets collected through crowdsourcing and citizen science.

Effective water resources management depends on monitoring the volume of water flowing through streams and rivers, but collecting continuous discharge measurements using traditional streamflow gages is prohibitively expensive. Time-lapse cameras offer a low-cost option for streamflow monitoring, but training models for predicting streamflow directly from images requires streamflow data to use as labels, which are often unavailable. We address this data gap by proposing the alternative task of Streamflow Rank Estimation (SRE), in which the goal is to predict relative measures of streamflow such as percentile rank rather than absolute flow. In particular, we use a learning-to-rank framework to train SRE models using pairs of stream images ranked in order of discharge by an annotator, obviating the need for discharge training data and thus facilitating monitoring streamflow conditions at streams without gages. We also demonstrate a technique for converting SRE model predictions to stream discharge estimates given an estimated streamflow distribution. Using data and images from six small US streams, we compare the performance of SRE with conventional regression models trained to predict absolute discharge. Our results show that SRE performs nearly as well as regression models on relative flow prediction. Further, we observe that the accuracy of absolute discharge estimates obtained by mapping SRE model predictions through a discharge distribution largely depends on how well the assumed discharge distribution matches the field observed data.