Skip to main content

Plain-language guide

How we know

Every number on Terra Earth comes from NASA satellites and is checked with fixed statistical tests before it reaches a chart. This guide explains each idea in everyday words — no equations required.

What the satellites measure

Two NASA instruments feed every location page: MODIS Terra for vegetation and land-surface temperature, and NASA POWER for air temperature, rainfall and sunlight.

Vegetation — NDVI

Satellite cameras measure how strongly plants reflect light. The score, called NDVI, runs from about 0 (bare soil) to 1 (dense green canopy). Rising NDVI usually means more or healthier vegetation.

Land surface temperature

Infrared sensors measure how hot the ground itself is — pavements, fields, rooftops. On sunny afternoons the surface is often much hotter than the air, which is why this number can sit well above the air temperature on the same page.

Air temperature

NASA POWER reports the air temperature about two metres above the ground, averaged over the month — the everyday temperature you would feel standing outside.

Rainfall & sunlight

The same NASA POWER record carries corrected monthly rainfall and the sunlight energy reaching the surface. All three come from the same satellite-era archive, 2001 onward.

The detective method, step by step

  1. 1

    Keep a fixed copy of the record

    Before any analysis, Terra Earth saves a copy of the NASA observations for each location. Every page and every visitor reads the same saved numbers, so results cannot quietly change between visits.

  2. 2

    Ask: is there really a direction?

    The Mann–Kendall test compares every year with every later year and asks whether rising pairs clearly outnumber falling pairs. It answers “does this series keep moving one way?” — not “how fast?”.

  3. 3

    Measure the pace with Theil–Sen

    Theil–Sen draws a line between every pair of years and keeps the middle slope. One strange heatwave year barely moves the median, so the rate of change — shown per decade — stays trustworthy.

  4. 4

    Let the p-value judge

    The p-value is the chance of seeing a pattern this strong if the series were actually flat. When it drops below 0.05 — a 1-in-20 coincidence — we call the trend significant and show the badge.

  5. 5

    Show the receipt

    Every chart keeps its provenance beside it: the dataset ID, the source link, and when the record was retrieved. You can always trace a number back to NASA itself.

  6. 6

    Admit the gaps

    When a location has no cached record, the page says “Data not yet available” instead of guessing. An honest blank teaches more than an invented number.

Put together: Mann–Kendall says whether a real direction exists, Theil–Sen says how fast it moves, and the p-value says how confident we may be. A trend gets shown as significant only when all three agree.

Words on every card

Trend
A long, steady direction across many years of observations — not one hot summer or one rainy month.
Significance
Whether the pattern passed the statistical test. A significant trend is one we can distinguish from random year-to-year noise.
Rate per decade
The Theil–Sen slope multiplied by ten — how much the value changes across ten years, in the variable's own unit.
Confidence band
On comparison charts, the shaded region where the true trend line most plausibly lies (95% confidence). A narrow band means a well-pinned trend.
Provenance
The receipt for every number: dataset ID, source URL, and retrieval time. Open it from any chart's View evidence control.
Cache
A pre-saved copy of NASA records stored with the site, so analysis always runs on the same tested values.
Biome verdict
A plain-language label — Stable, Greening, Drying, Warming stress — computed by fixed rules from the tested variables. No model invents it.

Our honesty rules

No AI-made numbers

Statistics come only from tested code — Mann–Kendall and Theil–Sen. The AI assistant explains computed results; it never supplies climate values.

Pictures are not evidence

Comic art, mascots and location scenery exist to explain and welcome. Only the charts and provenance panels carry measured claims.

Places, not national averages

Comparisons use named representative capital points, each with its own record. They show how places differ — never a national average.

One hot year is not a trend

Weather jumps around every year. That is exactly why we test many years together before calling anything a trend.

Ready to investigate yourself? Open any location from the globe, or see the full dataset list on Data & Methods. Data & Methods