| snippet:
|
This grid represents the predicted presence probability of small gorgonian corals using random forest on presence and absence data collected from DFO multispecies trawl surveys in the Maritimes region between 2002 and 2014 with additional in situ benthic imagery records collected between 1997 and 2014. |
| summary:
|
This grid represents the predicted presence probability of small gorgonian corals using random forest on presence and absence data collected from DFO multispecies trawl surveys in the Maritimes region between 2002 and 2014 with additional in situ benthic imagery records collected between 1997 and 2014. |
| accessInformation:
|
Fisheries and Oceans Canada, Bedford Institute of Oceanography, P.O. Box 1006, Dartmouth, NS, Canada B2Y 4A2 |
| thumbnail:
|
|
| maxScale:
|
5000 |
| typeKeywords:
|
[] |
| description:
|
<DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>Species distribution modelling using a random forest (RF) machine learning approach was used to predict the probability of occurrence and biomass of sponges, sea pens, large and small gorgonians and Vazella pourtalesi in the Maritimes Region. A suite of 66 environmental predictor variables from different data sources were used. Species occurrence was predicted using all presence and absence data (unbalanced model), and a balanced species prevalence model (i.e. an equal number of presences and absences). Also, for such taxonomic groups whose distribution was felt was not fully sampled by the multispecies stock assessment surveys, or when the number of trawl records for a group was insufficient for producing accurate predictions of distribution, additional random forest models were run using trawl survey data augmented with data from other sources: 1) in situ benthic imagery observations from scientific surveys, 2) DFO scallop stock assessment surveys, and 3) commercial records from the Fisheries Observer Program (FOP). The model unbalanced presence and absence catch data from DFO multispecies trawl surveys and in addition with other sources was chosen for the sea pens, small and large gorgonian corals and Vazella pourtalesi as the better prediction surface and the model produced from the balanced data was chosen for sponges.</SPAN></P><P STYLE="margin:0 0 0 0;"><SPAN>Three measures of accuracy were used to assess model performance: sensitivity, specificity, and AUC, or Area Under the Receiver Operating Curve. The accuracy measures for the random forest model using all small gorgonian presence and absence data, in situ benthic imagery observations and a threshold equal to species prevalence (0.06)were AUC= 0.949, sensitivity= 0.876and specificity=0.916; indicating excellent model performance.</SPAN></P><P><SPAN /></P><P><SPAN /></P></DIV></DIV></DIV> |
| licenseInfo:
|
<DIV STYLE="text-align:Left;"><DIV><DIV><P><SPAN>Please refer to the Open Government Licence - Canada</SPAN></P><P><A href="http://open.canada.ca/en/open-government-licence-canada"><SPAN>http://open.canada.ca/en/open-government-licence-canada</SPAN></A></P><P><SPAN /></P></DIV></DIV></DIV> |
| catalogPath:
|
|
| title:
|
Maritimes_SmallGorgonians_PresenceProbability_Model3 |
| type:
|
|
| url:
|
|
| tags:
|
["Scotian Shelf","Maritimes Region","Small Gorgonian Coral","Deep-Sea Coral","Presence Probability","Model 3","Random Forest."] |
| culture:
|
en-CA |
| name:
|
|
| guid:
|
|
| minScale:
|
625000 |
| spatialReference:
|
|