The first rung of the Cloud’s analytics ladder: before classifying, correlating or simulating anything, look at how your pixel values are distributed. The histogram tells you whether an insight separates populations, where a threshold should sit, and whether the data is clean enough to trust.
Running a histogram #
In Cloud → Exports, open an Area’s files and click the chart icon (Compute Histogram) on any exported GeoTIFF. It works on files from saved Areas up to 5 million pixels — the tooltip tells you when a file doesn’t qualify.
The form has two parameters:
- Number of Bins (24–128, default 32, with 32/64/128 presets) — more bins show finer detail but get noisier; fewer are smoother but less precise.
- Percentiles — the robust range for the statistics: [2, 98] (default) excludes extreme outliers, [5, 95] is more conservative, [1, 99] keeps almost everything.
An optional execution log attaches a step-by-step trace of the computation for diagnostics.
Reading the results #
- One tab per band, each with its pixel count, full range and robust range.
- The chart switches between Bars and Steps, with marker lines for the mean, median and your chosen percentiles.
- On very large files the platform samples rather than reads every pixel — the result is flagged approximate, so treat the numbers as estimates.
The Statistical Analysis #
One more click (Statistical Analysis) computes the full statistical profile, organized in six cards:
- Central Tendency — mean, median, percentile values.
- Variability — standard deviation, coefficient of variation, IQR, range.
- Distribution Shape — skewness and kurtosis: is the population symmetric, or does it drag a tail?
- Outliers & Quality — outlier count and percentage, valid pixels, Z-score range.
- Texture Analysis — contrast, homogeneity, energy and edge density: how spatially structured the values are.
- Information Complexity — entropy and band correlation.
A General Interpretation line summarizes it in plain words — it’s AI-generated, so verify before acting on it.
Exporting the analysis #
The Export button produces a PDF report (one page per band: chart, statistics, interpretation) or a CSV with the raw histogram table and every metric — ready for your own tooling.
When to reach for it #
- Setting thresholds: find the valley between two populations before building a Mask.
- Sanity-checking an insight: a distribution collapsed on one value means the index isn’t discriminating anything on this Area.
- Comparing dates: run the same insight’s histogram on two acquisitions to see if the population truly shifted or just the illumination.