Every variable below is "one estimate of one thing, over one patch of ground, at one time." The skill is knowing, for each: what it really measures, in what units, from which sensor, and the one way it lies. That last column — the gotcha — is what separates a plausible analysis from a correct one.
Land & vegetation
NDVI / EVI
measuresGreenness / vegetation vigour
unitsunitless −1…1
fromMODIS, Sentinel-2, HLS
⚠ Saturates over dense canopy; cloud-contaminated pixels must be masked.
LST — land-surface temperature
measuresSkin temperature of the ground
units°C / K
fromMODIS MOD11, ECOSTRESS, Landsat TIRS
⚠ NOT air temperature — can be 10–20 °C hotter on a sunny day.
Land cover
measuresWhat's on the ground (cropland, forest, urban…)
unitscategories
fromMODIS, ESA WorldCover
⚠ A classification, so it has an accuracy — check the confusion matrix.
Atmosphere & greenhouse gases
NO₂ / SO₂ / CO / O₃ column
measuresTotal amount in the air column
unitsmolecules/cm²
fromOMI (2004–), TROPOMI (2018–, sharper)
⚠ A column ≠ what you breathe at the surface; mind the OMI row-anomaly.
CH₄ / CO₂ (XCO₂)
measuresMethane / carbon-dioxide concentration
unitsppb / ppm
fromTROPOMI, EMIT, OCO-2/3, GOSAT
⚠ Albedo/aerosol artefacts; coarse pixels mix many sources.
AOD — aerosol optical depth
measuresHaziness / particulate load
unitsunitless
fromMODIS
⚠ A proxy for PM, not a PM measurement; needs ground calibration.
Water & cryosphere
Precipitation
measuresRainfall / snowfall amount
unitsmm
fromGPM IMERG (~10 km), CHIRPS (land)
⚠ Satellite precip is biased, especially for extremes & mountains — always cross-check.
Soil moisture
measuresWater in the top ~5 cm of soil
unitsm³/m³
fromSMAP (microwave)
⚠ Surface-only; coarse (~9–36 km); dense-vegetation signal is weak.
TWS — terrestrial water storage
measuresTotal water (incl. groundwater) as mass change
unitscm equiv. water
fromGRACE / GRACE-FO (gravity)
⚠ Very coarse (~300 km), monthly; groundwater = TWS − soil − surface − snow.
SST — sea-surface temperature
measuresOcean skin temperature
units°C
fromGHRSST/MUR, MODIS
⚠ Skin vs. bulk temperature differ; cloud gaps in IR products.
Sea level / altimetry
measuresSea-surface height
unitscm
fromJason, SARAL/AltiKa
⚠ Along-track; coastal values are noisier than open ocean.
Snow / SWE, ice elevation
measuresSnow extent / water equiv. / ice height
unitsfraction, mm, m
fromMODIS snow, ICESat-2 (ATL06/ATL15)
⚠ Optical snow is cloud-blind; SWE retrievals are uncertain.
Human activity & hazards
Nightlights
measuresCity-light radiance (power/activity proxy)
unitsnW/cm²/sr
fromVIIRS Black Marble (VNP46)
⚠ Moonlight, cloud and snow distort it; mask to clear nights.
SAR backscatter (σ⁰)
measuresRadar brightness — floods, soil, structure
unitsdB
fromSentinel-1, NISAR
⚠ Smooth water is dark (→ floods) but so are roads/sand — look-alikes.
FRP / active fire
measuresFire detections & radiative power
unitsMW
fromVIIRS/MODIS (FIRMS)
⚠ Misses fires under cloud/canopy; thermal threshold can miss small fires.
dNBR — burn severity
measuresHow badly a fire burned
unitsindex
fromNBR pre vs. post (NIR/SWIR)
⚠ Confused by pre-fire conditions; pair with field severity classes.
The four gotchas that catch everyone(1) a column (NO₂, CH₄) is not a surface concentration; (2) LST is not air temperature; (3) satellite rainfall is biased — cross-check it; (4) every variable has fill values + a scale factor + a quality flag you must apply before you average. Get these and you've avoided most published-paper mistakes.
Two threads run through this whole list and the rest of the course: compare to a baseline (a raw value rarely means anything), and cross-check before you trust (especially the biased ones). Next up: the stats that turn these numbers into "something changed."