How to Do Real Earth and Environmental Sciences Research at Home: A High School Student’s Guide to Free Tools, Affordable Kits, and Public Databases

Earth and Environmental Sciences used to mean a university lab, a research vessel, or a field station with a six-figure budget. That world has cracked open. A student with a laptop, a $40 sensor, and a free Google account can now measure, model, and map the planet at a level that would have been a PhD thesis 15 years ago.

This guide is your starting point. It covers three things: the kit you can put together for under the cost of a school Chromebook, the free software that runs the same workflows used at NASA and NOAA, and the public datasets that already contain the answers to thousands of unasked questions.

Why this is possible now

Three shifts changed what a high schooler can do from a bedroom.

First, satellites went open. Sentinel-1, Sentinel-2, Sentinel-5P, Landsat, MODIS, VIIRS, GRACE-FO, and TEMPO all stream their data to the public for free. You can pull 10 years of imagery for any county on Earth in a single afternoon.

Second, the sensor market collapsed in price. A BME280 measures temperature, humidity, and pressure for under $10. A PMS5003 measures fine particulate pollution for around $25. An MH-Z19 reads CO₂. An RTL-SDR dongle picks up radio signals from meteor trails. None of these existed at hobby prices a decade ago.

Third, the compute caught up. Google Earth Engine processes petabytes of satellite data in your browser for free. Google Colab gives you a real GPU for training neural networks. QGIS and Python libraries like xarray, rasterio, and Flopy run the same analyses used in government and academic labs.

Put together: a kitchen counter plus a laptop is now a small Earth-observation lab.

The Earth and Environmental Sciences home kit

You do not need everything in this list. Pick the items that match the question you want to ask.

Air and atmosphere sensors

• PMS5003 or SDS011 particulate sensor for PM2.5 and PM10 (~$25)
• MH-Z19 or SCD41 CO₂ NDIR sensor for indoor and outdoor CO₂ (~$25 to $60)
• MQ-series gas sensors for CO, NO₂, VOCs (~$5 each)
• BME280 for temperature, humidity, and barometric pressure (~$10)
• MLX90614 IR thermometer for surface temperatures and urban heat mapping (~$15)

Water sensors

• TDS / EC pen and pH pen for quick water-quality readings (~$15 each)
• Dissolved-oxygen probe (hobby-grade, under $40)
• Turbidity sensor (LDR + LED through a cuvette, DIY for ~$10)
• Hall-effect flow meters for stream and rainfall flow (~$10)
• Nitrate, ammonia, and phosphate test strips from any aquarium store

Earth and weather sensors

• Capacitive soil-moisture sensors (~$3 each)
• Anemometer kit and tipping-bucket rain gauge (~$30 each)
• Geophone or piezo element paired with a Raspberry Pi for a backyard seismometer
• RTL-SDR dongle for meteor radar reflections and GNSS experiments (~$30)

Microcontrollers, cameras, and structure

• ESP32 or Arduino for any sensor you want logged with WiFi or LoRa (~$10)
• Raspberry Pi 4 or 5 for vision projects, all-sky cameras, and seismometers
• USB microscope for sediment, microplastic, and soil work (~$30)
• Smartphone camera and GoPro / timelapse for photogrammetry and field surveys
• 3D-printed Stevenson screens, sensor mounts, mini flumes, and wave tanks

A full single-question kit usually lands between $50 and $250.

The signature technique: turn your phone into a field instrument

Smartphones are the single most underused tool in environmental science. The camera, GPS, microphone, and compass already in your pocket can do real measurement work. Here is a 5-step workflow that applies to dozens of projects, from turbidity to aerosol optical depth to cloud classification.

1. Pick the signal. Decide what you are measuring (water clarity, sky color at sunset, cloud type, sediment grain size, bird collisions on a window). Identify the physical quantity behind it.
2. Calibrate against a known reference. Use formazin for turbidity, AERONET for aerosols, MODIS for clouds, USGS gages for streamflow. Your phone reading must be tied to a trusted instrument so judges trust your numbers.
3. Standardize the capture. Lock white balance, exposure, and ISO if your camera app allows. Mount the phone the same way every time. Take 10+ frames per sample and average.
4. Extract numbers with Python. Use OpenCV or PIL to pull RGB values, brightness, edge counts, or object detections from each frame. Save to a CSV with timestamp and GPS.
5. Fit a model. Build a regression or small neural network that maps your phone numbers to the calibrated reference, then deploy it across a transect, a neighborhood, or a season.

This is the spine of citizen-science Earth research right now. One phone, one calibration, and a clear question.

The dry-lab side: free software you can install today

Geographic information systems and remote sensing

• Google Earth Engine runs decades of satellite data in your browser with Python or JavaScript
• QGIS is the free desktop GIS that does what ArcGIS does
• geemap wraps Earth Engine into Jupyter notebooks
• SNAP is the European Space Agency's toolbox for Sentinel data
• rasterio and xarray are the Python workhorses for raster and gridded climate data

Hydrology and groundwater

• HEC-RAS and HEC-HMS are the US Army Corps tools for flood and watershed modeling
• MODFLOW with Flopy runs groundwater simulations from a Python script
• SWAT+ and SWAT-CUP model whole watersheds for nutrients and sediment
• LANDLAB and ParFlow simulate landscape evolution and surface-subsurface flow

Atmosphere, climate, and fire

• WRF is the research-grade weather model (runs on Colab for short cases)
• OpenFOAM does shallow CFD for wind around buildings
• FDS simulates wildfire smoke dispersion
• PCRaster and CLM-FATES handle land-surface and vegetation dynamics

Photogrammetry and 3D

• OpenDroneMap and Meshroom turn smartphone photos into 3D models of outcrops, stream banks, and tree canopies

Machine learning and statistics

• scikit-learn, PyTorch, JAX for the core ML stack
• PyMC for Bayesian regression and hierarchical models
• Prophet, ARIMA, PatchTST, TimesNet for time-series forecasting
• U-Net, ConvLSTM, graph neural nets, PINNs, SINDy for spatial and physics-informed problems

Running the same software that government scientists run changes the project. You stop summarizing what others did and start producing your own results.

Public databases that count as real data

Satellites and imagery

• Sentinel-1 radar, sees through clouds, perfect for flood mapping and subsidence
• Sentinel-2 10 m optical imagery, ideal for land cover and vegetation
• Sentinel-5P TROPOMI for NO₂, CH₄, SO₂, HCHO, and CO columns
• TEMPO for hourly air-quality data over North America
• Landsat 5 through 9 for 40+ years of surface change
• MODIS and VIIRS for daily land surface temperature, snow, fire, and night lights

Climate and weather

• ERA5 and ERA5-Land reanalysis on Copernicus CDS
• NASA POWER for solar and meteorology at any point on Earth
• GHCN for ground station temperature and precipitation records
• NOAA NCEI for storm reports, IBTrACS hurricane tracks, and historical climate
• GPM IMERG for half-hourly global precipitation

Water and hydrology

• USGS NWIS for stream gages, well levels, and water quality
• GRACE-FO mascons for total water storage anomalies
• GLDAS and SMAP for soil moisture and land-surface fluxes
• HydroSHEDS for global river networks and watersheds

Air quality

• PurpleAir API and OpenAQ for crowdsourced ground PM and gas data
• AERONET for sun-photometer aerosol optical depth
• EPA AirNow for regulatory monitor data

Earth and seismic

• USGS earthquake catalog for global seismicity
• IRIS for raw seismic waveforms
• Nevada Geodetic Lab and UNAVCO for GNSS time series
• SRTM and Copernicus DEM for global elevation

Ecology and biodiversity

• GBIF, iNaturalist, eBird for species occurrence
• Global Forest Watch and Hansen forest-loss for deforestation
• NEON open data for long-term ecological measurements

Re-analyzing public data with a fresh question is itself a legitimate research path. Some of the strongest student projects never collect a new measurement and still discover something the original analysts missed.

How to combine wet and dry: the strongest project shape

Pattern A: ground truth a satellite. Pick a satellite product (Sentinel-5P NO₂, MODIS LST, Sentinel-2 chlorophyll) and validate it against your own sensor readings on a transect across your city. The contribution is the local correction model.

Pattern B: extrapolate a backyard experiment. Run a controlled mesocosm or sensor experiment (road-salt on duckweed, first-flush runoff, soil compaction) and use a public dataset (OSM, Sentinel-2, USGS) to project the result across an entire region.

Judges respond to hybrid projects because they prove you understand both the measurement and the model. One side without the other looks like a demo. The two together look like science.

Choosing a phenomenon that has not been done

1. Open Google Scholar and search the specific phrase you would use to describe your project. Read the abstracts of the top 10 results. Note which variables, regions, or methods have already been covered.
2. Search the Society for Science abstracts archive for the past 5 years of ISEF and Regeneron projects in Earth and Environmental Sciences. Skim titles and abstracts.
3. Search NASA ADS, AGU journals, and Web of Science (or your library's database) for any recent paper that touches your phenomenon. Look at the "future work" or "limitations" section of one strong paper. That is often a gift list of unanswered questions.

If you find five papers that are close to your idea but none that asks your exact question, you are in the right place. Adjacent prior work is good news. It means the field cares, and your question fits.

A realistic timeline

• 1 to 2 weeks: Pull one public dataset, ask one focused question, produce one figure and one statistic. Great for replication or a first measurement campaign.
• 1 to 2 months: A full hybrid project for a regional fair. One sensor build or one mesocosm, paired with one satellite or reanalysis dataset, plus a model.
• Full year: An ISEF-track project. Multi-season data collection, a published-quality model, uncertainty quantification, and an open-source code release.

If this is your first research project, start with the 1 to 2 week version. Finishing a small project teaches you more than abandoning a big one.

A starter checklist

1. A clean workspace with power, internet, and room for sensors or a small tank.
2. A free Google account with Colab and Earth Engine access enabled.
3. A local Python environment with xarray, rasterio, geemap, scikit-learn, and PyTorch installed.
4. QGIS installed on your laptop.
5. A lab notebook (paper or digital) where you log every change, sensor reading, and code run with timestamps.
6. A GitHub account with one private repo for your project code.
7. A one-line written research question that names the phenomenon, the variable, and the location.

If you have all seven, you are ready to pick a phenomenon.

Where to go next

ISEF organizes Earth and Environmental Sciences into five subcategories. Each one has its own MehtA+ project guide that builds on the kit and software described above. Pick the one that pulls you in.

• Atmospheric Science (AIR): air quality, urban heat, clouds, aerosols, ozone, and methane plumes.
• Climate Science (CLI): downscaling, extreme events, paleoclimate reconstruction, and long-term trend detection.
• Environmental Effects on Ecosystems (ECS): pollution dose-response, light pollution, bioacoustics, pollinator networks, and habitat fragmentation.
• Geosciences (GES): seismology, geomorphology, volcanic monitoring, subsidence, and analog tectonic experiments.
• Water Science (WAT): streamflow, groundwater, lake stratification, harmful algal blooms, and stormwater quality.
• Other (OTH): coupled human-natural systems, environmental justice, climate-health links, and policy modeling.

Earth science used to live behind a locked lab door. The door is open now, and the next project on the list could be yours.

Project ideas in this category (68)