databaseData in the ClassroomFor Educators
waves

Level 3 — AI vs Manual

Comparing Analysis Approaches

The Traditional Approach to Sea Level Analysis

Before machine learning and artificial intelligence became standard tools, oceanographers analyzed sea level data using statistical methods developed over decades of practice. These traditional approaches are rigorous, transparent, and well-understood, forming the foundation for modern AI-enhanced analysis.

The classical approach begins with data preparation. Scientists would manually examine raw measurements from tide gauges, looking for obvious errors caused by equipment malfunction. They would flag suspicious values and either remove them or estimate what the correct value should have been based on surrounding data.

This approach worked well for analyzing data from one or two locations. But by the 1990s, scientists had satellite altimetry data from thousands of locations worldwide. Traditional methods of manual quality control and single-location analysis simply could not scale to this massive volume of measurements.

Traditional Analysis Steps

  1. 1
    Data Collection:Download measurements from tide gauges or satellite databases
  2. 2
    Quality Control:Manually identify and remove erroneous readings
  3. 3
    Apply Corrections:Adjust for atmospheric pressure, land movement, etc.
  4. 4
    Calculate Averages:Compute monthly/yearly means to smooth short-term variations
  5. 5
    Regression Analysis:Fit a trend line and calculate rate of change
  6. 6
    Peer Review:Other scientists verify methods and conclusions

What AI Brings to the Table

speed

Speed

AI can process 30 years of satellite data in minutes. Manual analysis of the same dataset might take months of researcher time.

hub

Scale

Machine learning handles millions of daily measurements across the entire global ocean — a scale impossible for human analysts.

pattern

Pattern Detection

AI identifies subtle patterns, correlations, and anomalies that might be invisible to human analysis of averaged data.

Case Study: AI Processing 30 Years of Satellite Data

In 2020, researchers at NASA's Jet Propulsion Laboratory used machine learning to reanalyze the complete satellite altimetry record from 1993 to 2020. The AI system processed over 2 billion individual measurements.

The results revealed something manual analysis had missed: sea level rise is not constant. The rate of rise has been accelerating — each decade, the ocean rises faster than the decade before.

This discovery has major implications for coastal planning. If acceleration continues, sea level by 2100 could be significantly higher than older projections suggested.

Key Findings

  • trending_upSea level rise accelerated from 2.5mm/year (1993) to 3.9mm/year (2020)
  • mapRegional variations identified — some areas rising 3x faster than global average
  • timelineEl Nino/La Nina effects separated from long-term climate signal
  • bug_reportIdentified and corrected systematic errors in early satellite data

Strengths and Limitations of AI Analysis

thumb_upStrengths

  • check_circleProcesses massive datasets quickly and consistently
  • check_circleIdentifies patterns too subtle for human detection
  • check_circleCan be retrained as new data becomes available
  • check_circleReduces human bias in data interpretation
  • check_circleEnables real-time monitoring and early warning systems

warningLimitations

  • errorCan only learn patterns present in training data
  • errorMay miss unprecedented events or novel phenomena
  • error'Black box' decisions can be hard to explain
  • errorRequires high-quality input data to produce reliable results
  • errorCannot replace scientific domain expertise

Activity: Compare the Results

Below are two analyses of the same sea level dataset — one performed manually by researchers in 2010, and one using AI in 2020. Compare the findings and discuss the differences.

Manual Analysis (2010)

  • Finding: Global sea level rising at 3.1mm/year
  • Method: Monthly averages from 23 reference stations
  • Conclusion: Rise appears constant since 1993
  • Time to complete: 8 months

AI Analysis (2020)

  • Finding: Rise accelerating from 2.5 to 3.9mm/year
  • Method: All 2 billion satellite measurements
  • Conclusion: Acceleration detected with high confidence
  • Time to complete: 3 days

Discussion Questions:

  1. Why might the 2010 analysis have missed the acceleration that AI detected?
  2. Does the AI result mean the 2010 researchers were wrong? Why or why not?
  3. What role should human scientists play when AI performs the analysis?
  4. How might the AI analysis be limited in ways the manual analysis was not?
arrow_backBack to Level 2Next: Level 4arrow_forward

Next: Level 4

You now understand how AI and traditional analysis approaches differ, and how each has particular strengths and limitations. In Level 4, you'll explore how AI is used to forecast future sea level changes. You'll learn how machine learning models make predictions, how scientists quantify prediction uncertainty, and how policymakers use these predictions to plan adaptation strategies.

Continue to Level 4: How AI Models Forecast Sea Level Change.