Appearance
From signals to a drought index
This is the step that turns raw drought signals into a comparable drought measure: the bridge between satellite data and triggers and payouts. It is what lets the platform judge very different areas on the same scale.
The problem: "normal" differs everywhere
The same level of greenness might be a great season for an arid area and a terrible one for a lush area. You cannot compare raw readings across areas, and you cannot set one universal trigger. The fix is to measure each season relative to its own area's history.
Step 1: Sum up the season
For each area and year, the monthly greenness readings are accumulated over the growing months into a seasonal total, cumulative greenness (CumNDVI). The model uses the wet season.
Step 2: Build the baseline
The baseline is the area's long-term "normal," computed over a baseline window (a range of historical years):
- the long-term average seasonal greenness, and
- its variability (how much it swings from year to year).
The reference project uses 22 years (2003-2024) as the window.
IMPORTANT
The baseline window is a real lever. A longer window is more stable but slower to reflect a changing climate; a shorter one is more responsive but noisier. The window used is recorded on every analysis, so results are reproducible.
Step 3: Express the season as a standard score
Each year's seasonal total is turned into a standard score (a z-score): how many standard deviations it sits from the area's baseline average.
standard score = (this season's total − area's average) / area's variability- A negative score means the season was worse than the area's normal, the drought signal.
- A positive score means better than normal.
- Because every area is divided by its own average and variability, scores are directly comparable across areas.
A small worked example
An area whose normal seasonal greenness is 5.0 with a variability of 0.8 records a wet-season total of 3.4:
score = (3.4 − 5.0) / 0.8 = −2.0Two standard deviations below normal, a severe season for that area, regardless of how its absolute greenness compares to anywhere else.
The standardised drought measures
The platform stores two standardised forms per area and season:
- ZCumNDVI: the standard score of cumulative seasonal greenness (the value above).
- PZCumNDVI: its percentile form: where the season ranks in the area's own history.
The map colours areas by the percentile form (red below the 10th percentile, green above) for an at-a-glance drought view.
From scores to thresholds
Pricing thresholds are set as percentiles of each area's historical scores, then resolved to a concrete value per area, for example, a trigger at the 10th percentile (drought begins) and an exit at the worst observed season (maximum payout). Setting thresholds as percentiles keeps them meaningful for each area's own variability. How they become payouts is the subject of Triggers & payouts.
Climate signals work the same way
Rainfall and soil moisture carry the same kind of anomaly scores, at monthly, seasonal, and annual horizons, so a dry season shows up consistently across vegetation, rainfall, and soil moisture. The drought index (SPEI) is already standardised by construction.
Where to go next
- Triggers & payouts, turning thresholds into payouts.
- Pricing the risk, pricing the simulated payouts.
- The drought signals, where the raw measures come from.