NASA Space Apps 2026 · Earth System Trend Detective
Click anywhere on the Atlantic, pull real NASA POWER (MERRA-2) records, strip out the seasons, and test whether a trend is a genuine signal or just noise.
The 3D globe could not load (offline or WebGL unavailable).
Use the 2D map, presets or coordinates — the analysis still works.
Decode the data
Method
For every calendar month we average all Januaries, all Februaries and so on, then subtract that monthly mean from each value. What remains is the anomaly: how unusual a month was. Trend statistics are always computed on anomalies, so summer-to-winter swings cannot hide or fake a trend.
Ordinary least squares gives the slope (shown per decade), R² and a 95% confidence interval from Student’s t distribution. Monthly data are serially correlated, so we also apply an effective-sample-size correction, neff = n(1−r₁)/(1+r₁); the verdict uses the cautious, corrected value.
Counts how many later values are higher than earlier ones (S), converts to a Z-score with tie and continuity corrections, and reports a two-tailed p-value from the normal distribution. Sen’s slope gives a robust rate of change. The same autocorrelation correction is applied.
p < 0.01 statistically significant · p < 0.05 significant · p < 0.10 emerging · otherwise not significant. The gauge shows the more cautious of the two tests. A small p-value means the trend is unlikely to be chance, not that it is large or that we know its cause.