Identification of Patterns over Regional Scales Using Self-Organising Maps on Images from Marine Modelling Outputs

Souza, P. A. de; Williams, RN; Jones, Emlyn

doi:10.1109/dicta.2012.6411677

Cited by 1 publication

(1 citation statement)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Different network topologies are presented in SOM with neurons, which are arranged in a 2-D grid with dimensions M × N. The, the number of clusters equals to the product M × N (e.g., if the grid is 2 × 3, then the number of clusters equals to 6). SOM has been proven efficient for the analysis of spatiotemporal data and examples of its application include meteorology, oceanography, and water resource research (Kalteh et al 2008;Liu and Weisberg 2011;Vilibi c et al 2011;Souza et al 2012;Jahan et al 2013).…”

Section: Spatial Clusteringmentioning

confidence: 99%

MovCLUfish: A data mining tool for discovering fish movement patterns from individual‐based models

Politikos

Kleftogiannis

Tsiaras

et al. 2021

Limnology & Ocean Methods

View full text Add to dashboard Cite

Spatially explicit individual-based models (IBMs) are useful tools for simulating the movement of discrete fish individuals within dynamic and heterogeneous environments. However, processing the IBM outputs is complicated because fish individuals are continuously adjusting their behavior in response to changing environmental conditions. Here, we present a new analysis tool, called MovCLUfish, that uses data mining to identify patterns from the trajectories of the individuals generated from IBMs. MovCLUfish is configured to identify features of fish behavior related to occupation (area of fish presence), dynamics of aggregation (how fish individuals are distributed within the area of presence), and mobility (how fish move between subregions). MovCLUfish receives as input the fish locations (longitude, latitude) at fixed times during a specific time period and performs spatial clustering on consecutive timestamps, considering them as moving objects. Fish locations are grouped into clusters whose features (centroid, shape, size, density) are used to provide further information about the spatial distributions. The clusters are analyzed using three built-in pattern mining methods: tracking moving centroids (TMC), aggregating moving clusters (AMC), and tracking fish mobility (TFM). TMC detects shifts in the distribution of fish over time, AMC visualizes the way fish aggregations change geographically over time, and TFM provides quantitative information on the patterns of exchange and connectivity of individuals among regions within the domain. We describe the workflow of MovCLUfish and illustrate its applicability using output from an IBM model configured for anchovy in the Eastern Mediterranean Sea. Further avenues for improvement and expansion of MovCLUfish are discussed.

show abstract