Chapter 12 — Data Evaluation, Presentation, and Interpretation

12.1 Turning numbers into understanding

Data becomes useful only when evaluated: plotted, compared with the baseline and design, and interpreted. This chapter is about separating signal from noise.

12.2 Presentation

Effective plots reveal behavior at a glance:

• Time-series — readings vs. date, with reservoir level overlaid.
• Scatter plots — seepage vs. reservoir, piezometric head vs. reservoir.
• Profiles — piezometric head vs. depth, deformation vs. distance along the crest.
• Cumulative — settlement vs. time to show rate.

12.3 The baseline and the band

The baseline is the historical normal range. Plotting a band around it makes departures obvious. A reading outside the band is not automatically a problem — but it is a prompt to investigate.

12.4 Separating signal from noise

Real trends must be distinguished from:

• Temperature effects — many sensors drift with temperature; correct or isolate them.
• Seasonal effects — reservoir and pore pressure cycle annually.
• Tidal / barometric effects — coastal or shallow sensors.
• Instrument noise and faults — frozen, dropping, or drifting sensors.

12.5 Correlating variables

The strongest interpretations correlate measurements: a piezometric rise with a reservoir rise is normal; a piezometric rise without one is not. Deformation that tracks pore-pressure change points to a specific mechanism.

12.6 From anomaly to hypothesis

When a departure appears, the engineer forms and tests hypotheses:

1. Is it a sensor fault? (Check validation, neighbors, manual reading.)
2. Is it an environmental effect? (Season, temperature, event.)
3. Is it real dam behavior? (Correlate with other instruments and inspection.)

12.7 Documentation of evaluation

Every evaluation — routine or triggered — should be recorded: the plots used, the conclusion, and any action. This builds an institutional memory of how the dam "speaks."

12.8 Key takeaways

• Plot against time, reservoir, depth, and distance to reveal behavior.
• Use a baseline band; departures prompt investigation, not panic.
• Remove temperature, seasonal, and sensor-noise effects before judging a trend.
• Correlate variables and document every evaluation.