Data Quality
Understanding and ensuring good seismic data quality from your Grillo sensors.
What is data quality?
Data quality refers to how well your seismic recordings represent actual ground motion, free from noise and artifacts.
Good quality data
- Clear earthquake signals when events occur
- Low background noise
- Consistent, continuous recording
- Accurate timing
Poor quality data
- Excessive noise obscuring signals
- Gaps in recording
- Timing errors
- Artifacts from installation issues
Why quality matters
For earthquake detection
| Quality level | Detection capability |
|---|---|
| High quality | Detect smaller events, accurate parameters |
| Medium quality | Detect moderate events, some errors |
| Poor quality | Miss events, false triggers, wrong parameters |
For early warning
Fast, accurate detection requires:
- Clear P-wave onsets
- Low false trigger rate
- Reliable timing
For research
Scientific applications need:
- Consistent data characteristics
- Documented installation
- Known instrument response
Noise sources
Natural noise
| Source | Characteristics |
|---|---|
| Wind | Variable, affects exposed sites |
| Ocean waves (microseisms) | Continuous, low frequency |
| Weather | Rain, thunder |
| Biological | Animals near sensor |
Anthropogenic (human-caused) noise
| Source | Characteristics |
|---|---|
| Traffic | Variable by time of day |
| Machinery | Often periodic |
| Construction | Intermittent, intense |
| HVAC | Continuous when operating |
| Walking | Impulsive, irregular |
Sensor/installation noise
| Source | Characteristics |
|---|---|
| Poor coupling | Resonances, weak signal |
| Loose mounting | Spikes, instability |
| Tilting | DC offset changes |
| Electronics | High-frequency noise |
Assessing quality
Visual inspection
Look at waveforms for:
- Background noise level
- Unusual patterns
- Gaps or spikes
- Consistency
Quality metrics
Common measures:
| Metric | Description |
|---|---|
| RMS noise | Root-mean-square of background |
| PSD | Power spectral density |
| Data completeness | Percentage of expected data |
| Timing quality | Clock accuracy |
Comparative analysis
Compare sensors:
- Similar sites should have similar noise
- Outliers indicate problems
- Consistent characteristics across network
Improving data quality
Site improvements
-
Move away from noise sources
- Relocate sensor if possible
- Address source if controllable
-
Better coupling
- Direct contact with solid surface
- Remove soft materials underneath
-
Environmental control
- Stable temperature
- Protected from drafts
- Away from direct sunlight
Installation improvements
-
Level the sensor
- Use bubble level
- Adjust mounting
-
Secure mounting
- No wobble
- Won't shift over time
-
Cable management
- No tension on sensor
- Protected from disturbance
Operational practices
-
Regular monitoring
- Check data quality dashboards
- Compare to baseline
-
Prompt issue resolution
- Investigate anomalies
- Fix problems quickly
-
Documentation
- Record installation details
- Note any changes
Common quality issues
High noise floor
Symptoms: Background level higher than expected
Possible causes:
- HVAC or machinery nearby
- Traffic vibration
- Poor installation site
- Electrical interference
Solutions:
- Relocate sensor
- Address noise source
- Improve installation
Spikes/glitches
Symptoms: Sudden jumps in data
Possible causes:
- Loose sensor
- Cable issues
- Electrical interference
- Nearby impacts
Solutions:
- Secure sensor
- Check cables
- Shield from interference
- Identify impact source
Data gaps
Symptoms: Missing data periods
Possible causes:
- Network connectivity issues
- Power outages
- Sensor malfunction
- Server issues
Solutions:
- Improve network reliability
- Add power backup
- Check sensor health
- Contact support
High-frequency noise
Symptoms: Excessive noise at high frequencies
Possible causes:
- Electrical interference
- Sensor electronics
- Nearby equipment
Solutions:
- Check power source
- Add filtering (if available)
- Relocate sensor
Quality and network performance
Detection threshold
Higher noise = higher detection threshold
- Good quality sites detect M2-3
- Noisy sites may only detect M4+
Location accuracy
Quality affects location:
- Clear arrivals = precise timing
- Noisy data = uncertain picks
- Network average determines accuracy
False triggers
Noise causes false triggers:
- Looks like earthquake signal
- Wastes processing resources
- May cause false alerts
Monitoring quality over time
Establish baseline
When sensor is installed:
- Record typical noise levels
- Document expected characteristics
- Set quality thresholds
Track changes
Monitor for:
- Increasing noise (new source?)
- Sudden changes (installation issue?)
- Seasonal variations (weather, HVAC)
Regular review
Schedule periodic review:
- Weekly spot checks
- Monthly quality reports
- Quarterly comprehensive review
Quality vs quantity trade-off
Dense networks
More sensors can compensate for some quality issues:
- Redundancy covers gaps
- Bad sensors can be excluded
- Statistics improve with numbers
Minimum quality standards
Even in dense networks, maintain minimums:
- Sensors must detect target events
- Timing must be accurate
- Data must be usable