Analysis of Ocean Data Science Initiatives’ Geospatial Data Visualizations

VGDP Identification:

We first conducted an exploratory analysis on the websites of the ODSI in our purposively selected sample to determine how ODSIs visualize data (Melvin et al., 2025; see also Drakopulos et al., 2022). Based on this analysis, we defined VGDPs as any file, image, or user interface that is used to convey data through a visual, spatial depiction of oceans without independent cleaning, processing, or analysis. This definition included only those visuals that each ODSI website framed as primary outputs or data products of the ODSI. We excluded visuals that are exemplary of the ways ODSI data may be used.  We also excluded images contained in reports, memos, articles, or white papers, and those that are external to the ODSI but linked within the ODSI.

Database Creation and Coding Protocol:

We conducted our analysis using QSR NVivo, capturing categorical information regarding the characteristics of the VGDP using case attributes, and using qualitative coding to capture themes reflected in the VGDPs. We established the following protocol to build our NVivo Database:

1. Download the map image file from the ODSI or take screenshots of the API.  Where a map is an API (see ReadMe for definition), the screenshot should be the default map that appears on the website. If no default map appears (i.e. the API has no initial image but instead allows a user to select options to build an image), capture an exemplary output. Save each image in the Teams database.

2. Import each captured image into NVivo.  Label each VGDP with a map identifier (XXX-XX) where the 3 digits before the dash are the ODSI unique identifier, and the two digits after the dash are numbered sequentially for each VGDP (e.g. maps from ODSI 2 are 002-01, 002-02, etc.)

3. For each new VGDP, create a case for the VGDP using the map identifier.  Classify each case as either an API or static image.

4. Fill out the case classification sheet for each attribute. 

5. [Initial 30 VGDPs only; see below] Create a memo for each VGDP.  The naming scheme for the memo should be Map ID- Initials-Date the memo was written in YYYYYMMDD format (e.g. 001-01-ECM-20230626).  Code the memo to the VGDP case.  In this memo, capture your reactions to the VGDP in the broadest terms.  You do not need to repeat information coded in the attributes, but to the extent there are reactions that are not captured by those attributes, add them here.  Consider but do not limit your reflections to the following questions: What purpose is this map seeking to serve? What story is it telling?  How might it be used? What does the basemap look like and how is it depicting the oceans?  What kinds of data is it presenting?  What are the general impressions it is giving?  What questions are you left with? What information does the map obscure?  What is the spatial question the map is trying to answer?  Is it successful in doing so?  What are the visual spatial elements in the map?

6. Create a new file for the ODSI description in the “ODSI Map Descriptions” folder. Name the file XXX-XX-ODSIDescription (where XXX-XX is the map ID number.) Copy the text from the ODSI website that describes the map.  Code the file to the map case.  (If there is no description, create the file and indicate that there is no description.) 

Inductive Qualitative Coding Process

Qualitative themes were identified using an iterative coding process.  We identified themes based both on the image captured in our database and, for APIs, our impressions as we interacted with them on the ODSI database. As noted above, in the early stages of the coding process, we drafted open-ended memos for thirty VGDPs that captured our impressions and understandings.  We identified a combination of deductive and inductive codes (described in the attached ReadMe) throughout this process and coded the relevant portions of the memos for these themes. After drafting memos for thirty VGDPs, we reached saturation of our codebook, ceased writing memos, and directly coded VGDPs for qualitative themes.  After saturation, we also directly re-coded the initial 30 VGDPs, to ensure all codes were captured.

Data Export and Cleaning

We exported case classification sheets of categorical attributes in Excel format. We also conducted a matrix coding query to indicate the presence or absence of each qualitative code within each VGDP case (whether coded to images or memos) and exported the query results in Excel format. We combined these outputs and cleaned the data using RStudio. We refined and recategorized some initial case attributes (e.g. renamed some initial attributes for clarity; combined some attributes that were previously separate); our final classification scheme is found in the attached ReadMe. We also exported memos and ODSI Descriptions for inclusion in the database.

This dataset includes (1) Map Dataset (the dataset containing all coding results); (2) ReadMe (descriptions of codes in dataset); and (3) child datasets corresponding to each Map ID, which include images, ODSI Descriptions, and (where applicable) memos. (Due to constraints on child datasets, VGDPs have been combined in a single child for each of ODSIs 086 and 090.)

Researchers are welcome to use the data in this dataset with proper attribution.

References:

Drakopulos, Lauren Alexandra; Havice, Elizabeth; Crisp, Katie; Zurita Posas, Ana; Campbell, Lisa M.. 2022. "Catalog of Ocean Data Science Initiatives". Qualitative Data Repository. https://doi.org/10.5064/F6ZQWQJS. QDR Main Collection. V1.