x̄ - > 🌐 Unlocking Ocean Mysteries: Missing Salinity in NOAA WOD

🌐 Unlocking Ocean Mysteries: Missing Salinity in NOAA WOD

The NOAA World Ocean Database (WOD) is one of the most comprehensive oceanographic datasets ever compiled. It contains millions of vertical “casts” measuring temperature, salinity, and other key variables across the global ocean.

Yet even this gold-standard dataset contains inconsistencies. One particularly intriguing issue: CTD profiles that completely lack salinity data.

Key Insight: Missing salinity is not random—it follows geographic, institutional, and depth-related patterns.

🌍 Spatial Patterns

Rather than being evenly distributed, missing salinity clusters in specific regions:

• Northwest Pacific (near Japan)
• Arctic fringe (Canada/Alaska)
• Northwest Atlantic

This strongly suggests systematic causes—likely tied to regional surveys or instrumentation practices.

Read More

🏛️ Institutional Signals

Using statistical diagnostics (Chi-square + residuals):

• Canada: +7.51 → strong excess missing data
• Japan: +1.65 → moderate excess
• USA: −4.56 → highly complete

A few cruises contribute disproportionately, indicating localized data issues.

📏 Depth Bias

Result: Shallower profiles are significantly more likely to miss salinity (p < 0.0001)

Profile Type	Avg Depth (m)
With Salinity	569.85
Missing Salinity	405.24

Difference ≈ 165 m. Spearman correlation: −0.518.

🔬 CTD Profile Behavior

• Temperature profiles remain complete
• Salinity may be entirely absent
• Gaps often persist across consecutive casts

🌊 Implications

• Not Missing Completely At Random (MCAR)
• Regional bias affects climate models
• Data provenance is critical
• Cloud tools enable rapid QA/QC

Core Takeaway: Data gaps are signals—not noise—and must be explicitly modeled.

⏭️ Next Step

Future work will explore animal-borne ocean sensors (e.g., seals) to assess whether they can fill these observational gaps.