groups.qmd

## Clustering of Locations, Events, and Source Records

#### Clusters in Surveillance Activities

The units **Location, Event, and Source Record could be enough to record the data structure of a specific Surveillance Activity** gathering information during patrols (Field Visits) in Protected Areas (Locations) at points (Events) where dead animals (Source Records) are found at any given time. But another Surveillance Activity could have the following data structure: 

- Visit (**Field Visit**)
  - Protected area (Spatial cluster level 1), 
    - Zones within protected area (**Location**), 
      - Grid cells within each zone (Spatial cluster level 2), 
        - Capture site (**Event**) 
          - Mist nets at the capture site (**Collection**), 
            - Bats captured (**Source Records**) 

Another example is surveillance of wild animals in a live markets. One of the potential options to structure these data is:

- Visit (**Field Visit**)
    - City (Spatial cluster level 1)
      - Neighborhood (Spatial cluster level 2)
        - Market (**Location**)
          - Vendor (Spatial cluster level 3)
            - Stalls (Spatial cluster level 4)
              - Cage (**Event**)
                - Animals (**Source Records**) 
                  - Rectal Swab (**Specimens**)

<!-- Looking at the examples it is easy to understand that **what an Event represents depends on the Surveillance Activity** (See Event section above) as another Surveillance Activity can establish each Vendor as an Event. However, certain guidelines can be followed to promote standardization of Events (see Recommendations for Standardization below).  -->

Looking at the examples above, it is straightforward to understand that **the units Location, Event, and Collection might not be enough to accommodate the data of all Surveillance Activities and other units ("Clusters") can be needed**. 

A second layer of complexity is the potential need for nested Clusters and non-nested Clusters. The last example above corresponds to a series of nested Clusters where the stalls are within vendors and neighborhoods are within cities. But it is also possible to need non-nested spatial Clusters, for example, "Zip Code". Zip Code A can include portions of cities 1 and 2, and Zip Code B can also include portions of cities 1, 2, and 3.

And a third layer of complexity is the potential need for spatial and temporal clusters. For example, it is possible that the data structure of the live market Surveillance Activity presented above has to be categorized in decades, year, season, season-year, etc. The following example adds the categorization of the data by season, no matter the markets visited or how many years the Surveillance Activity lasts: 

  - **Season (Temporal cluster 1)**
  - Visit (**Field Visit**)
    - City (Spatial cluster level 1)
      - Neighborhood (Spatial cluster level 2)
        - Market (**Location**)
          - Vendor (Spatial cluster level 3)
            - Stalls (Spatial cluster level 4)
              - Cage (**Event**)
                - Animals (**Source Records**) 
                  - Rectal Swab (**Specimens**)

Finally, the number of categories within each Cluster can also be different across Surveillance Activities. For example, two Surveillance Activities may include grid cells as one of their Clusters, however, one Surveillance Activity can include grid cells A - R, whilst the another Surveillance Activity can include grid cell A - W (more categories). In summary:

- Clusters units can be needed
- The Clusters across Surveillance Activities can be different
- What these Clusters are grouping across Surveillance Activities can be different
- The number of categories can be different among Clusters, within and between Surveillance Activities 
- Nested, Non-nested or both types of Clusters can be needed
- Spatial, Temporal or both Clusters types of Clusters can be needed

#### Clusters in the Data Model

To accommodate these needs in data structure, **the data model allows the inclusion of Clusters** between the Source Record and Event (in the absence of Collection) or Collection and Event, Event and Location, and Location and Field Visit. The data model also allows the inclusion of an undetermined number of nested and non-nested Clusters at each of these levels, and the inclusion of spatial and temporal Clusters. 
<!-- **Environmental and Arthropod Source Records cannot be clustered in units smaller than Event** but the Locations and Events containing them can be clustered.  -->

**Unavoidably, the number of Clusters, what Clusters represent, what they cluster, the number of categories per Cluster, and the data to be collected from each of these extra units will vary among Surveillance Activities. Therefore, Cluster properties must be reported in the Surveillance Activity metadata. In the data model, the only default property for each Cluster are the identifier, the cross identifier, the origin of the cross identifier, and a description**. Other potential properties must be documented in a separate file (e.g., an excel sheet) with common identifiers to allow joining the clusters with the corresponding data.

An interesting case study is the collection of Specimens from reindeer carcasses when they are found dead and also from the soil underneath each carcass. In this Surveillance Activity, an Event is a site where dead animals (at least one) are found. So, if two or more carcasses are found at the same site, then the Event contains two Animal Sources, but also two Environmental Sources providing soil Specimens. However, the interest of the researchers is to maintain the connection between the soil Specimens collected below each Animal Source and the Specimens from each Animal Source. Basically, they want to keep track of the soil Specimen that was below each Animal Source. In this case, Cluster unit can be used to group the corresponding Soil Specimen and Animal Specimen in two pairs while keeping the four specimens under the same Event. 



<!-- Further characteristics of each clustering unit, whether static or time dependent through the Surveillance Activity period, should be prepared in a file, such as an excel sheet, with columns with the identifiers of all relevant units (Location grouping units. Event grouping units that are not the Location, and Source Record grouping units that are not the Event) and  attached it at the corresponding level. A recommendation is to prepared a sheet for the features of all spatio temporal grouping units visited during a Field Visit and attach it at this level (Field Visit). The columns with the identifiers will allow joining the data of the corresponding Field Visit (Locations, Events, Source Records, grouping units) with the attached and spatial data. -->