MoveApps
Github repository: https://github.com/dmpstats/Cluster_Importance_Scoring
Assigns importance scores to spatial clusters of track locations based on various cluster properties, including the incidence of feeding events. The underlying importance scoring methodology is still a work-in-progress and is expected to be updated with a more advanced approach in the near future.
This App calculates importance scores for clusters of spatially aggregated location points by combining cluster attributes such as the incidence of feeding events, the number of days it was active, number of member tracks and average duration of visits during daytime. Clusters are scored into one of four classes of incremental importance, indicated by the column importance_label in output: "Low", "Medium", "High" and "Critical".
Cluster importance scores are key for identifying points of interest to e.g. inform ground patrolling decision-making.
The process for calculation of the importance score for each cluster uses four binomial models. Two models to determine carcass vs no carcass and two models to determine large carcass vs small/medium carcass. Each model uses a variety of the cluster variables calculated from the "Generate Avian Cluster Metrics" MoveApp.
- p(Carcass given a single bird cluster)
- p(Carcass given a multibird cluster)
- p(Large Carcass given a single bird cluster)
- p(Large Carcass given a multibird cluster)
Fitted values for each cluster are converted back to binary scores using a ROC threshold and the importance determined as follows:
| p(Carcass) | p(Large) | Importance | Importance Label |
|---|---|---|---|
| 0 | 0 | 0 | Low |
| 0 | 1 | 1 | Medium |
| 1 | 0 | 2 | High |
| 1 | 1 | 3 | Critical |
In addition, the App uses a simple duration threshold (via parameter nest_thresh_days) to flag long-lasting clusters that may represent nesting. These clusters are automatically assigned a "Low" importance score. While many are likely to correspond to nests, such clusters may also represent water-drinking sources, food 'restaurants', or other prolonged non-carcass spatial aggregations.
This App has been developed specifically for Vulture species.
This App is strictly dependent on the prior deployment of the App Generate Avian Cluster Metrics (GitHub) in the workflow.
A move2::move2_loc object.
A move2::move2_loc object.
-
"clusters_map.html": if option Generate Interactive Map is selected, an interactive map will be produced showing the centroid locations and key properties of the analysed clusters. Points are colour-coded according to their assigned importance score, and point size is proportional to the number of location points comprised in the cluster.
-
"clusters_tbl.csv": a dataset with key metrics and the calculated importance scores of the analysed clusters.
Generate Interactive Map (map_output): Select this option to create an interactive map of the output as an App artefact.
Potential Nest Threshold (days) (nest_thresh_days): Consecutive days after which an active cluster is flagged as a potential nest. Such clusters are automatically given a low importance score. Leave blank to skip nest identification.
The app will halt processing an throw an error if the dependency App 'Generate Avian Cluster Metrics' is not deployed earlier in the workflow, as the importance score relies on the availability of specific cluster properties calculated in that App.
Not applicable.