Plant monitoring via Micro:bit

WP3

Before – an invitation

• Symposium at SGME
• Google Meet
• April 24, starting at 11 a.m., Romanian time

The plan

3 600 000 ms = 1 hour

Anturio, Funchal

• Our colleagues from Madeira even wet the plant automatically.

EJH-Alexandroupolis

• Many batteries
• Visual confirmation that data is recorded

Silves

• With visual confirmation for data recording, as well.

Craiova

• Two batteries
• No visual confirmation
• Recording hourly

Number of records per school

Madeira has the most…

However…

Madeira recorded data every 10 seconds

• We had to use one record out of each 360…

​

• 3151 records divided by 360…
• We ended up with 9 records we could use in the comparison.

Next issue - humidity

• EJH has some huge values – as outliers in the plot
• The largest value: 832
• All the others have values in [0, 100].

Normalized moisture for EJH

• EJH uses raw analog values (0–1023+)
• the others use a 0–100 “percentage-like” scale
• For comparable results, we need to normalize EJH data

Normalized moisture for EJH

• The level of moisture decreased heavily for EJH
• Silves and Craiova had very constant values.

Why the

large span

in Madeira?

Madeira – self watering on duty

Silves data

• Recorded every 15 minutes, instead of 60.
• Still, out of 301 records, we reduce to around a quarter.
• Some issues with the date were manually repaired.

Number of samples per school (again)

Before

After

Overall temperature comparison

Overall light comparison

Humidity comparison over time

Temperature comparison over time

Light comparison over time

Conclusions

• We harmonized the data into a common hourly format, for visual comparisons.
• The plots show site-specific patterns in temperature, light and moisture
• reflecting both local climate and indoor placement of the plants.

​

• Inconsistent date formats and different input sensors significantly influenced initial results
• careful preprocessing is very important

​

• After cleaning, most apparent “outliers” were special events (isolated days or hours), not sensor errors.

THANK YOU!

Key message before showing results

The comparison required preprocessing before interpretation

Raw sample counts were not directly meaningful.

Sampling intervals had to be harmonized to hourly data.

Humidity from EJH had to be normalized.

Only after these corrections could the school data be compared in a meaningful way.