Written guide to an imported profile. Engineering Signal Processing’s Underwater Propagation Editor can inspect depth and sound-speed columns, calculate descriptive gradients and export the complete selected profile. This example covers Input Contract and Sound-Speed Profile. It does not run a propagation solver or establish field calibration.
1. Identify the published file and its units
The example comes from Roger Flood’s MGDS sound-velocity collection for central Long Island Sound, DOI 10.60521/332590. The repository describes depth in metres and sound speed in metres per second; its individual dataset license is CC BY 4.0. Retain the author, dataset citation and selected filename with your input.
The selected file is 2022-08-12_. The saved preparation retains all 19 published depth/speed pairs in their original order and decimal text, adding only the CSV headers depth_ and sound_. It performs no local thinning, interpolation, filtering, clipping or unit conversion. The prepared file is mgds-lis4a-3510044-full.csv.
Keep the provenance limitation attached. “Full” here means all rows of this published, already thinned file. The rows span 0–12,000 m, although the collection is from coastal Long Island Sound. The saved review does not independently establish the deep extension’s origin or that every row was measured in situ. Do not present this file as an unprocessed sensor cast, a calibrated local water column or a navigation product.
This is a row table, not a time series. No sample rate was supplied in the reviewed workflow. A source-row index is an address in the table; it is not elapsed seconds, sample frequency or distance along a survey track.
2. Import through the app and choose the two routes
- Open Underwater Propagation Editor in Engineering Signal Processing (
underwater_). Use the app’s Import/Upload action for the prepared CSV. The saved native workflow used this app-specific import; the shared launcher’s waveform importer is a separate path.propagation_ editor - Set Input schema → Depth and sound speed. This explicit schema expects two columns in order: depth in metres, then sound speed in metres per second. Keep the unit-bearing headers even though this selection assigns the columns by position.
- Set Solver provenance → Review imported data and Profile order → Preserve source order. Leave SSP profile ID at
auto_firstfor this single-profile input. It resolves to profile ID 0 here; with multiple IDs, the current default chooses the smallest valid ID, not necessarily the first row’s ID. - Select only Input Contract and Sound-Speed Profile, then run. Confirm successful route status and inspect the input contract before reading the plots.
The expected input contract is 19 rows, 2 columns, 19 valid rows and 1 profile ID. The present fields are depth and sound speed. The saved contract explicitly reports that no solver was executed and no external solver output was declared or verified. Changing the provenance declaration to “External solver output” would not authenticate a solver result.
For another input, inspect valid-row counts and the selected profile. The current profile route requires at least two valid rows and unique depths within that profile; repeated depths are refused for the gradient calculation. Do not remove conflicting observations merely to obtain a result without recording why they differ.
3. Follow row identity through the plots and numerical table
The profile route sorts the selected valid profile by depth before calculating its gradients, even when the import’s order setting is “Preserve source order.” Its numerical export retains source_, the zero-based row in the uploaded table, alongside each depth, speed and gradient. This example was already ordered by depth, so its exported source indices are 0 through 18.
The profile figure contains four views: speed against depth, the spatial gradient dc/dz, row-depth samples and a speed histogram. Depth increases downward in the first three views. This input has no range column, so the sample plot uses source row index. The histogram counts published rows; unequal depth spacing means it is not a depth-weighted distribution or a probability model for the ocean.
| Exported column | How to read it |
|---|---|
source_ | Zero-based address in the uploaded table, retained after depth sorting. |
depth_ | Imported depth in metres. |
sound_ | Imported sound speed in metres per second. |
gradient_ | Change in sound speed per metre of depth: (m/s)/m = 1/s. This is a spatial derivative, not a time derivative. |
The exported numeric table has 19 rows × 4 columns = 76 values. Those rows describe this selected valid profile. They do not represent every profile in a larger upload or restore excluded invalid rows.
4. Read descriptive results without upgrading their meaning
| Saved result | Meaning for this file |
|---|---|
| Minimum speed: 1,483 m/s “Sound Channel Depth m”: 1,000 m | The minimum among the imported profile’s rows. The field name does not establish a measured coastal sound channel. |
| Maximum Gradient Depth m: 7.21 m | Depth of the largest absolute saved spatial gradient. |
| Gradient Span 1/s: about 0.121539 | Maximum gradient minus minimum gradient across the selected rows; not a temporal rate or an uncertainty estimate. |
| Surface Speed m/s: 1,518.39 Deep Speed m/s: 1,669 | Speeds at the first and last depth-sorted rows, respectively. Here those depths are 0 and 12,000 m; the deep row retains the provenance limitation above. |
| Linear trend: about 0.0114078 (m/s)/m | A fitted linear summary of this imported profile. It is not a propagation model or evidence that the profile is physically linear. |
The native numeric export retains, for example, gradient span 0.1215389458928345 and linear trend 0.011407757852089457. Rounded values in this guide and the app’s log are display summaries. Use the exported numeric fields for a comparison.
The earlier independent receipt compared 13 scalar fields and all 76 profile values with scalar reference arithmetic, using an absolute tolerance of 1e−10. Its largest gradient difference was about 5.68e−14. This guide reopens the saved results and checks their agreement with that receipt; it does not recalculate gradients, refit a trend or rerun the app.
The saved “Method Reference” includes a TEOS-10 sound-speed link. This selected route reads already supplied speeds and calculates descriptive summaries; it does not execute TEOS-10, derive speed from temperature and salinity, trace rays, estimate transmission loss or verify an external solver.
5. Save both result representations and retain the input separately
- After the selected routes finish, use Open Viewer → Export Package. The reviewed viewer ZIP contains two PNG figures, three CSV/JSON table pairs, summary text and a manifest. The table sizes are 17 input-contract fields, 20 profile-summary fields and the full 19-by-4 profile table.
- Reopen the CSV and JSON tables, not just the PNGs. The saved check compares all 150 table cells, including 89 numeric cells, across the native CSV/JSON, historical server JSON and app-specific table export. All 19 source depth/speed pairs remain present.
- The app also has EXPORT PRIVATE EVIDENCE PACKAGE. Choose an existing destination directory. For this reviewed example it creates six files: two PNGs,
receipt.json,settings.json,tables.jsonandunderwater_. The NPZ contains four complete 19-value arrays for source row, depth, speed and gradient.propagation_ evidence.npz - The settings describe the completed run, including its schema, profile selection and provenance declaration. Later control edits do not change that saved run’s settings snapshot. Keep the two export formats distinctly named; the private app package and viewer ZIP have different contents.
The viewer manifest says reloadable session: false. Its figures are images and it does not separately include numeric plot arrays; the complete table is the numerical representation for this route. The app-specific NPZ contains selected-profile coordinates and derived values, not a full raw-source archive. Neither package replaces the input CSV, its ASVP source and the provenance receipt.
The app-specific package withholds source filenames, paths and logs. That does not anonymize the imported scientific values or certify content for sharing; the viewer manifest separately requires content review. The current source refuses an oversized complete numeric export instead of silently dropping rows. No large-input or new export execution was performed for this guide.