1 · What the indicators measure
SPI (Standardized Precipitation Index; McKee et al., 1993) transforms accumulated precipitation over a chosen window into a probability-based anomaly expressed in standard deviations relative to that calendar month's climatological distribution — negative values indicate precipitation deficit. SPEI (Standardized Precipitation Evapotranspiration Index; Vicente-Serrano et al., 2010) applies the same idea to the climatic water balance (precipitation minus potential evapotranspiration), making it sensitive to both rainfall deficits and atmospheric evaporative demand. The soil-moisture anomaly expresses root-zone-equivalent moisture as a percent departure from its seasonal climatology, and the Vegetation Condition Index (VCI; Kogan, 1995) rescales vegetation greenness between historical stress and optima. SPI/SPEI are dimensionless z-scores; SMA and VCI are percentages.
2 · Accumulation periods and reference period
Accumulation windows of 1, 3, 6 and 12 months are provided for SPI and SPEI: SPI-1/SPEI-1 track immediate meteorological fluctuations; 3-month values approximate a monsoon-season signal; 6-month values align with hydrological storage; 12-month values capture inter-annual drought. Standardization uses the per-calendar-month distribution over the reference period —. A fixed, stated reference period is essential: SPI/SPEI values are only comparable within it.
3 · Standardization and classification
Indicator values are classified with the conventional SPI/SPEI thresholds below. Class edges are inclusive on the dry side (e.g., −2.00 is extreme drought). SMA and VCI use their own documented thresholds because their distributions are bounded percentages — reusing SPI thresholds on them would be invalid.
4 · Spatial aggregation and area computation
Regional means are area-weighted: the mean of district values weighted by each district's precomputed surface area. Affected-area statistics sum the equal-area surfaces of districts whose value falls in a drought class. Areas were precomputed in a Lambert Azimuthal Equal-Area projection centred on India (22.5°N, 79.5°E) — never from degree-coordinate arithmetic. District geometries are served simplified for web performance; statistical areas are computed from the same simplified geometry, so reported areas are internally consistent with the served map.
5 · Drought events
Events are identified by runs analysis (Yevjevich, 1967) on the continuous monthly series: a run of at least 3 consecutive months with the indicator at or below the −1.0 truncation level. Duration = run length in months; severity = cumulative deficit below −1.0 over event months (Σ(−1.0 − x)); mean intensity = severity ÷ duration. Isolated single-month excursions below −1.0 do not constitute events under this definition.
6 · Missing observations
Missing district-months are never imputed, interpolated, or silently treated as normal. Charts show gaps; maps style missing units explicitly as no data; summary statistics use only valid observations (pairwise deletion for comparisons) and always report the valid count; tables show a completeness column and flag sparse rows. KPI cards display N/A rather than derived placeholders when no valid observations exist.
7 · Data lineage in this build
Demonstration dataset (synthetic). All indicator values were generated by a seeded stochastic model (numpy default_rng(42)) documented in scripts/generate_demonstration_data.py: a latent monthly anomaly combining national and state-scale autoregressive factors with district noise, monsoon-weighted variance, illustrative dry/wet episodes (2009, 2012, 2015–16, 2018–19), and deliberately injected missing observations. Marginal distributions, cross-scale persistence and SPI–SPEI correlations mimic real indicator behaviour so that every analytical code path is exercised realistically. These values are not observations and must not be cited as hydroclimatic evidence. Boundaries are community-maintained 2011-census-demarcation district polygons, generalised for web use; they follow the source's depiction and carry no political endorsement.
8 · Resolution and limitations
The analysis unit is the administrative district (724 units, 36 states/UTs as supplied); indicator values represent district-level aggregates, not point measurements. Monthly resolution; 240 months (2005-01 to 2024-12). Limitations: simplified boundary geometry; demonstration values carry no observation error model beyond the generator's design; event statistics depend on the stated truncation level and minimum duration; district splits/mergers after the boundary vintage are not reflected. Trend lines (optional, OLS) require ≥ 120 valid observations and are shown with sample size.
9 · Version and citation
—
—