From 1 - 4 / 4
  • Model describes the potential distribution range of Zostera marina in the Finnish coast. Model was produced using extensive data (~140,000 samples) on the Finnish Inventory Programme for Underwater Marine Environment (VELMU). Model was built using Boosted regression trees (BRT), and resulting models describe the probability of detecting a habitat-forming species in a cell. Environmental predictors include for instance (and are not only restricted to): bathymetry, euphotic depth, salinity, substrate, and wave exposure. As more accurate information is gained by diving than from video methods, dive data was used as the primary source for modelling with 75–90% for model training and 10–25% for validation. The secondary source, video data, was used only for species clearly identifiable from videos with additional subsets (25%) from targeted inventories. Dive and video data are limited to rather shallow depths (typically 20–30 m), leading to a situation where there are not enough samples from deep areas (below 50 m). To avoid artefacts in the models, a randomized absence dataset for areas deeper than 50 m was used during the modelling process. These points were used only as absences in macrophytes models, based on the knowledge that macrophytes do not live at such depths in the Baltic Sea due to habitat constraints and lack of light.

  • Model describes the potential distribution range of Potamogeton perfoliatus in the Finnish coast. Model was produced using extensive data (~140,000 samples) on the Finnish Inventory Programme for Underwater Marine Environment (VELMU). Model was built using Boosted regression trees (BRT), and resulting models describe the probability of detecting a habitat-forming species in a cell. Environmental predictors include for instance (and are not only restricted to): bathymetry, euphotic depth, salinity, substrate, and wave exposure. As more accurate information is gained by diving than from video methods, dive data was used as the primary source for modelling with 75–90% for model training and 10–25% for validation. The secondary source, video data, was used only for species clearly identifiable from videos with additional subsets (25%) from targeted inventories. Dive and video data are limited to rather shallow depths (typically 20–30 m), leading to a situation where there are not enough samples from deep areas (below 50 m). To avoid artefacts in the models, a randomized absence dataset for areas deeper than 50 m was used during the modelling process. These points were used only as absences in macrophytes models, based on the knowledge that macrophytes do not live at such depths in the Baltic Sea due to habitat constraints and lack of light.

  • Model describes the potential distribution range of Mytilus trossulus x edulisin the Finnish coast. Model was produced using extensive data (~140,000 samples) on the Finnish Inventory Programme for Underwater Marine Environment (VELMU). Model was built using Boosted regression trees (BRT), and resulting models describe the probability of detecting a habitat-forming species in a cell. Environmental predictors include for instance (and are not only restricted to): bathymetry, euphotic depth, salinity, substrate, and wave exposure. As more accurate information is gained by diving than from video methods, dive data was used as the primary source for modelling with 75–90% for model training and 10–25% for validation. The secondary source, video data, was used only for species clearly identifiable from videos with additional subsets (25%) from targeted inventories. Dive and video data are limited to rather shallow depths (typically 20–30 m), leading to a situation where there are not enough samples from deep areas (below 50 m). To avoid artefacts in the models, a randomized absence dataset for areas deeper than 50 m was used during the modelling process. These points were used only as absences in macrophytes models, based on the knowledge that macrophytes do not live at such depths in the Baltic Sea due to habitat constraints and lack of light.

  • Model describes the potential distribution range of Fucus spp in the Finnish coast. Model was produced using extensive data (~140,000 samples) on the Finnish Inventory Programme for Underwater Marine Environment (VELMU). Model was built using Boosted regression trees (BRT), and resulting models describe the probability of detecting a habitat-forming species in a cell. Environmental predictors include for instance (and are not only restricted to): bathymetry, euphotic depth, salinity, substrate, and wave exposure. As more accurate information is gained by diving than from video methods, dive data was used as the primary source for modelling with 75–90% for model training and 10–25% for validation. The secondary source, video data, was used only for species clearly identifiable from videos with additional subsets (25%) from targeted inventories. Dive and video data are limited to rather shallow depths (typically 20–30 m), leading to a situation where there are not enough samples from deep areas (below 50 m). To avoid artefacts in the models, a randomized absence dataset for areas deeper than 50 m was used during the modelling process. These points were used only as absences in macrophytes models, based on the knowledge that macrophytes do not live at such depths in the Baltic Sea due to habitat constraints and lack of light.