Seafloor Geodetic Monitoring at Axial Seamount
Processing multi-year acoustic ranging data from cabled FETCH instruments to measure tectonic baseline distances on the Juan de Fuca Ridge
AxialFetch processes pressure, temperature, salinity, sound speed, and acoustic travel-time time series from three ocean-bottom stations at the summit caldera of Axial Seamount. The two-stage pipeline produces calibrated baseline distance measurements between all station pairs, enabling detection of volcanic deformation at centimeter-scale precision.
Northern (2504)
/ \
NW / \ NE
/ \
Western (2503) ---- Eastern (2502)
WE / EW
| Station | ID | Coordinates | Caldera Position |
|---|---|---|---|
| Northern | 2504 | 45.959 N, 130.011 W | North side |
| Western | 2503 | 45.947 N, 130.027 W | West side |
| Eastern | 2502 | 45.963 N, 129.991 W | East side |
| Baseline | Direction | Length |
|---|---|---|
| NW / WN | North ↔ West | 1.765 km |
| NE / EN | North ↔ East | 1.642 km |
| WE / EW | West ↔ East | 3.260 km |
# Step 1 — Build calibrated station DataFrames and harmonic velocities
python -m src.create_dataframes
# Step 2 — Compute distances, apply corrections, generate figures
python -m src.calculate_rangePipeline 1 must complete before Pipeline 2 can run (Pipeline 2 loads pickle files produced by Pipeline 1).
Ingests raw instrument data and produces calibrated, corrected per-station DataFrames.
Processing Steps
| Step | Description |
|---|---|
| 1. Load FETCH CSVs | Parse multi-code instrument files (TMP, DQZ, SSP, BSL, INC) for all three stations |
| 2. Salinity correction | Align CTD salinity from OOI cabled array (MJ03F), apply calibration residuals, correct data gaps via mean-shift |
| 3. Build combined DataFrames | Merge pressure, temperature (+0.351 °C offset for station 2504), and salinity per station |
| 4. Tidal pressure correction | Fit tidal predictions via least-squares optimization, remove tidal signal from pressure |
| 5. Velocity computation | Compute two sound speed estimates: SSP (onboard velocimeter) and TEOS-10 (via gsw) |
| 6. Pressure post-processing | Demean pressure and compute 15-day moving average, incorporating external NANO sensors |
| 7. Harmonic mean velocities | Compute pairwise harmonic means for both SSP-based and TEOS-10-based velocities |
| 8. Tilt perturbation | Project inclinometer pitch/roll through instrument headings onto baseline directions |
Outputs — src/output/
| File | Contents |
|---|---|
combined_df4.pkl |
Northern station (2504) — pressure, temperature, salinity, velocity, tidal correction |
combined_df3.pkl |
Western station (2503) |
combined_df2.pkl |
Eastern station (2502) |
harmonic_mean_dfs.pkl |
SSP-based harmonic mean velocities (column: Harmonic Mean) |
harmonic_df_dict.pkl |
TEOS-10-based harmonic mean velocities (column: HMean) |
baseline_perturb.pkl |
Tilt-induced baseline perturbations (2504-2502_dL, 2504-2503_dL, 2502-2503_dL) |
Extracts acoustic range measurements, computes inter-station distances, applies corrections, and generates figures.
Processing Steps
| Step | Description |
|---|---|
| 1. Load inputs | Re-parse FETCH CSVs for BSL data; load pickle files from Pipeline 1 |
| 2. Extract & filter BSL | Apply station-specific outlier filtering (see table below) |
| 3. Compute distances | Interpolate harmonic-mean velocity onto timestamps, calculate one-way distance |
| 4. Tilt correction | Subtract tilt-induced path-length perturbation from each distance |
| 5. 30-day moving average | Compute rolling 30-day mean on each baseline time series |
| 6. Save distance CSVs | Export 6 CSVs with Record Time, Distance (cm), Moving Average (cm) |
| 7. Perturbation correction | Apply tilt correction via merge_asof (nearest, 1-hour tolerance), save corrected CSVs |
| 8. Generate figures | Produce 21 PNG plots: time series, binned stats, bidirectional comparisons, precision analysis |
Outlier Filtering Rules:
| Baseline | Method | Parameters |
|---|---|---|
| NE (2504 → 2502) | IQR | Quantiles (0.15, 0.85) |
| NW (2504 → 2503) | Hard threshold | Range(ms) ≥ 2618 |
| WE (2503 → 2502) | IQR | Quantiles (0.25, 0.75) |
| WN (2503 → 2504) | IQR | Quantiles (0.25, 0.75) |
| EW (2502 → 2503) | Hard threshold | Range(ms) ∈ [4636.2, 4638] |
| EN (2502 → 2504) | Hard threshold | Range(ms) ≥ 2400 |
Outputs — Distance CSVs
All saved to src/output/distances/:
| Raw | Corrected | Baseline |
|---|---|---|
NW_distance.csv |
NW_distance_corrected.csv |
North → West |
NE_distance.csv |
NE_distance_corrected.csv |
North → East |
WN_distance.csv |
WN_distance_corrected.csv |
West → North |
WE_distance.csv |
WE_distance_corrected.csv |
West → East |
EN_distance.csv |
EN_distance_corrected.csv |
East → North |
EW_distance.csv |
EW_distance_corrected.csv |
East → West |
Outputs — Figures
All saved to src/output/figures/:
| Figure | Description |
|---|---|
{NW,NE,WE,WN,EW,EN}_distance.png |
Distance time series with scatter + 30-day moving average |
{NW,NE,WE,WN,EW,EN}_binned.png |
30-day binned mean ± 1 std error bars |
North_West_vs_West_North.png |
Bidirectional comparison: NW vs WN |
East_North_vs_North_East.png |
Bidirectional comparison: EN vs NE |
West_East_vs_East_West.png |
Bidirectional comparison: WE vs EW |
10day_std_all.png |
10-day binned standard deviation, all 6 baselines |
10day_std_normalized.png |
10-day STD normalized by baseline length (cm/km) |
10day_std_subplots.png |
10-day STD in 3×1 subplot panels |
10day_std_normalized_subplots.png |
Normalized 10-day STD in 3×1 subplot panels |
rolling_std_normalized.png |
30-day rolling STD normalized by baseline length |
rolling_std_3baselines.png |
30-day rolling STD for WN, NE, WE |
AxialFetch/
├── src/
│ ├── __init__.py
│ ├── config.py # Paths, constants, station metadata, baseline mappings
│ ├── data_loader.py # FETCH CSV parsing (multi-code instrument files)
│ ├── salinity.py # CTD salinity correction pipeline
│ ├── station_data.py # Per-station combined DataFrame assembly
│ ├── tidal.py # Tidal pressure correction (least-squares fit)
│ ├── velocity.py # Sound speed: SSP interpolation + TEOS-10 via gsw
│ ├── pressure.py # Pressure demeaning + 15-day moving average
│ ├── harmonic.py # Harmonic mean computation (SSP + TEOS-10 based)
│ ├── tilt.py # Inclinometer → baseline perturbation + tilt correction
│ ├── create_dataframes.py # Pipeline 1 orchestration
│ ├── range_filter.py # BSL extraction, outlier filtering, distance computation
│ ├── range_correction.py # Moving average, CSV export, perturbation correction
│ ├── range_figures.py # All figure generation (21 plots)
│ ├── calculate_range.py # Pipeline 2 orchestration
│ ├── VSM_forward.py # Okada dislocation forward model (used by Slead notebooks)
│ ├── *.ipynb # Source notebooks + modeling/analysis work
│ └── output/
│ ├── *.pkl # Pickle files from Pipeline 1
│ ├── distances/ # CSV files from Pipeline 2
│ ├── figures/ # PNG plots from Pipeline 2
│ └── drift/ # Drift-correction outputs from drift2.ipynb
├── Data/ # All static inputs (referenced via config.DATA_DIR)
│ ├── Data_*_2502_.csv # FETCH instrument data (Eastern)
│ ├── Data_*_2503_.csv # FETCH instrument data (Western)
│ ├── Data_*_2504_.csv # FETCH instrument data (Northern)
│ ├── model3a.h5, model3b.h5 # Slead et al. slip-inversion models
│ └── *_drifcor.csv, *.mat # Drift corrections, misc inputs
├── Extra_files_fetch/ # Archived/exploratory work (PINN, common-mode, older baselines)
├── MJ03{E-B,F-B,F-E}/ # Differential pressure records
└── README.md
Paths to
Data/are centralised insrc/config.py(DATA_DIR). Notebooks insrc/andExtra_files_fetch/reach model files via the relative path../Data/, so run them from the directory they live in.
The .py modules above are the maintained pipeline. The notebooks are the research record —
where each processing step was worked out, plus modeling and figure work that never moved into
the pipeline.
| Notebook | Description |
|---|---|
Create_Dataframes.ipynb |
Source notebook for Pipeline 1; the modules were refactored out of this |
Calculate_Range.ipynb |
Source notebook for Pipeline 2 |
FETCH StreamLine Final.ipynb |
Monolithic ancestor (172 cells) — full load → salinity → velocity → harmonic mean → range workflow in one notebook |
FETCH Range Calculation.ipynb |
Later revision of the same monolith (140 cells); the two-notebook split came from these |
Okada dislocation models from the Slead et al. slip inversions (Data/model3{a,b}.h5), used to
predict what range change a given magma-source geometry should produce.
| Notebook | Description |
|---|---|
Slead 3a.ipynb |
Loads model 3a → 574 patches → VSM_forward.okada on the full grid → validates UZ against saved Z → station displacements → predicted Δrange per baseline |
Slead 3b.ipynb |
Same for model 3b |
Slead_3a_station_only_fast.ipynb |
Fast path: evaluates Okada only at station coordinates, skipping the grid. No map or UZ validation |
Least-squares fits of baseline range against bottom pressure plus an annual sine/cosine, to separate real deformation from seasonal and pressure-correlated signal.
| Notebook | Description |
|---|---|
Axial Modeling West Final.ipynb |
West↔North baseline vs central pressure + annual terms; reports R² and de-seasonalised residuals |
Axial Modeling East-West Final.ipynb |
East↔West baseline against three z-scored pressures (west, east, central) + annual terms |
These two label station 2504 as "Central" (co-located with NodeF), not "Northern" as elsewhere in this README.
| Notebook | Description |
|---|---|
drift2.ipynb |
Piecewise instrument-drift fits P ≈ a + b·Pc + c·t' over manually chosen segments; converts F2025_all.dat to MJ03F format; writes F_node_pressure.csv |
drift_removal.ipynb |
Successor to drift2 — same fit, cleaner workflow, plus depth-change extension. Writes per-station *_driftcorr_daily_*.csv |
seasonality removal.ipynb |
Fits L = a·Pc + b·sin + c·cos + d to 30-day-MA baselines and subtracts the seasonal term. Writes baselines_fit_outputs.csv and the 4-panel summary figure |
plots.ipynb |
Publication figures: cumulative seismic moment east vs west from the ForMaleenV4.mat earthquake catalogue, and baseline projections onto station geometry |
Sound Speed Computation
Two independent velocity estimates are computed for each station:
1. SSP (onboard velocimeter) Direct measurement from the FETCH instrument's sound speed profiler.
Station 2503 SSP data is unreliable and set to NaN.
2. TEOS-10 (derived) Calculated from temperature, pressure, and salinity using the Gibbs SeaWater library:
SA = gsw.SA_from_SP(S, P_dbar, lon, lat) # Absolute Salinity
CT = gsw.CT_from_t(SA, T, P_dbar) # Conservative Temperature
V = gsw.sound_speed(SA, CT, P_dbar) # Sound speed (m/s)Pairwise harmonic means of velocities are used for distance calculation, representing the average velocity along the acoustic path between two stations.
Distance Formula
Distance (m) = V_harmonic × (Range_ms − TAT_ms) / 2000
| Symbol | Description |
|---|---|
V_harmonic |
Pairwise harmonic mean sound speed (m/s) |
Range_ms |
Measured two-way acoustic travel time (ms) |
TAT_ms |
Instrument turnaround time (ms) |
/ 2000 |
Converts ms → s (÷ 1000) and two-way → one-way (÷ 2) |
Special Handling
| Condition | Treatment |
|---|---|
| Station 2503 SSP | Set to NaN (no reliable velocimeter). harmonic_mean_dfs['2502_2503'] uses 2502_2504 data; harmonic_mean_dfs['2503_2504'] uses only station 2504 SSP |
| Station 2504 temperature | +0.351 °C offset applied to TMP sensor readings |
| Tilt correction | Inclinometer pitch/roll projected onto baseline directions to remove apparent distance changes from instrument tilting |
| Data | Source | Format |
|---|---|---|
| FETCH instrument data | 3 CSV files in Data/ |
Multi-header CSV with codes (TMP, DQZ, SSP, BSL, INC) |
| CTD salinity | OOI Cabled Array (RS03CCAL, MJ03F) | CSV |
| Tidal predictions | Pre-computed (2022–2025) | Text (Year, Month, Day, Hour, Min, Sec, Value) |
| NANO pressure | Bottom pressure recorders (MJ03F, MJ03E, MJ03B) | .Data files |
python >= 3.8
pandas
numpy
scipy
matplotlib
gsw