Skip to main content
Learn · Climate & Atmosphere

Microclimate Monitoring as Scientific Practice

From single station to publishable dataset — rigour, uncertainty and contribution

Grades UG 90 min research live IESH data

Undergraduate students design a formal monitoring study using the sensor hub, grapple with calibration uncertainty, compare their readings to reanalysis data, and understand what it means to contribute to an open scientific record.

Learning objectives

  • Design a measurement protocol specifying sampling frequency, logging duration and quality control procedures.
  • Quantify uncertainty in sensor measurements from known specifications and environmental sources.
  • Compare station data with ERA5 reanalysis gridded data and explain discrepancies.
  • Write a data methods section meeting the standard expected in an environmental science paper.

Background

A single sensor reading is not the same as scientific data. Scientific data requires documentation of instrument specifications, calibration status, measurement uncertainty, site metadata (location, altitude, land cover, nearby heat sources) and a quality-control procedure for flagging or removing anomalous readings. Without this documentation, a dataset cannot be critically evaluated by others or compared to data from different stations.

ERA5 is the European Centre for Medium-Range Weather Forecasts (ECMWF) global atmospheric reanalysis — a model-derived dataset that blends historical observations with atmospheric models to produce gridded estimates of weather variables at approximately 31 km horizontal resolution for every hour since 1940. It is freely available via the Copernicus Climate Data Store. Comparing your station's readings to ERA5 values for the same grid cell and time period is a standard way to check for systematic biases in low-cost sensors.

Open data practices require that datasets be deposited in accessible repositories with persistent identifiers (DOIs), complete metadata, and clear licensing. The FAIR principles (Findable, Accessible, Interoperable, Reusable) are the current standard for research data management. The himalayansciences.org platform this hub can upload to is designed with these principles in mind.

Procedure

  1. Document the station metadata: GPS coordinates, altitude (sensor-derived and map-derived), surrounding land cover (forest / cropland / urban / other), distance to nearest building, sensor model and firmware version.
  2. Note the sensor specifications: temperature accuracy ±0.5°C, humidity accuracy ±2%, pressure accuracy ±1 hPa. Calculate the resulting altitude uncertainty using error propagation.
  3. Design a sampling protocol: decide on logging frequency (current: every 60s), data retention, and quality-control flags (e.g. flag readings where temperature changes >5°C in 60s as suspect).
  4. Access ERA5 data for the nearest grid point. Compare ERA5 2 m temperature and surface pressure to this station's readings for the same time period.
  5. Calculate the mean bias (station minus ERA5) and the root mean square error (RMSE) for temperature and pressure.
  6. Write a 400-word data methods section describing the station, instrumentation, sampling protocol and quality-control approach, following the structure of a published journal article.

Discussion

  1. ERA5 represents an average over a 31 km grid cell. Why might a single station always disagree with ERA5 to some degree, even if both are perfectly accurate?
  2. What would you need to do to this dataset before submitting it to a data repository? What metadata is essential?
  3. A journalist cites your station's temperature reading as evidence of a record-breaking heat event. What caveats would you want them to include?

Worksheet

1. The BMP280 pressure sensor has a stated accuracy of ±1 hPa. Using error propagation through the barometric formula, estimate the resulting uncertainty in the calculated altitude in metres. [6 marks]

Answer guide (for teachers)

At ~866 hPa and 1,289 m, dA/dP ≈ −8.4 m/hPa. So uncertainty ≈ ±8.4 m. Accept answers in range ±5 to ±15 m with reasoning shown.

2. List THREE potential sources of systematic bias in a low-cost temperature sensor mounted inside a building or in direct sunlight. [3 marks]

Answer guide (for teachers)

Heat island effect from building; solar radiation heating sensor enclosure; insufficient ventilation around sensor; proximity to heat sources (computers, humans); ground radiation at night.

Vocabulary

Reanalysis Data
A gridded climate dataset produced by combining historical observations with a weather model to create a continuous record with no gaps.
Uncertainty
The range within which the true value of a measurement is expected to lie, arising from limitations in instruments, methods and environmental conditions.
FAIR Principles
A framework for scientific data management: data should be Findable, Accessible, Interoperable and Reusable.
Metadata
Data that describes other data — in a sensor context, this includes location, instrument type, calibration status and measurement conditions.

Extension

Submit a week of quality-controlled data to himalayansciences.org and write a 200-word data note describing the station. Consider submitting this as a data paper to a student journal.

← Climate & Atmosphere